diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU')
41 files changed, 4077 insertions, 676 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll new file mode 100644 index 0000000..e117200 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll @@ -0,0 +1,612 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=hawaii < %s | FileCheck -check-prefix=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck -check-prefix=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX11 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX12 %s + +define i16 @s_add_i16(i16 inreg %a, i16 inreg %b) { +; GFX7-LABEL: s_add_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s16 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s16 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s16 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i16 %a, %b + ret i16 %c +} + +define i16 @v_add_i16(i16 %a, i16 %b) { +; GFX7-LABEL: v_add_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_u16_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u16_e32 v0, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_nc_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_nc_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i16 %a, %b + ret i16 %c +} + +define i32 @s_add_i32(i32 inreg %a, i32 inreg %b) { +; GFX7-LABEL: s_add_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s16 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s16 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s16 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i32 %a, %b + ret i32 %c +} + +define i32 @v_add_i32(i32 %a, i32 %b) { +; GFX7-LABEL: v_add_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i32 %a, %b + ret i32 %c +} + +define <2 x i16> @s_add_v2i16(<2 x i16> inreg %a, <2 x i16> inreg %b) { +; GFX7-LABEL: s_add_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s18 +; GFX7-NEXT: s_add_i32 s17, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: v_mov_b32_e32 v1, s17 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_lshr_b32 s4, s16, 16 +; GFX9-NEXT: s_lshr_b32 s5, s17, 16 +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: s_add_i32 s4, s4, s5 +; GFX9-NEXT: s_pack_ll_b32_b16 s4, s16, s4 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_lshr_b32 s4, s16, 16 +; GFX8-NEXT: s_lshr_b32 s5, s17, 16 +; GFX8-NEXT: s_add_i32 s4, s4, s5 +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: s_and_b32 s4, 0xffff, s4 +; GFX8-NEXT: s_and_b32 s5, 0xffff, s16 +; GFX8-NEXT: s_lshl_b32 s4, s4, 16 +; GFX8-NEXT: s_or_b32 s4, s5, s4 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_lshr_b32 s4, s16, 16 +; GFX10-NEXT: s_lshr_b32 s5, s17, 16 +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: s_add_i32 s4, s4, s5 +; GFX10-NEXT: s_pack_ll_b32_b16 s4, s16, s4 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_lshr_b32 s2, s0, 16 +; GFX11-NEXT: s_lshr_b32 s3, s1, 16 +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: s_add_i32 s2, s2, s3 +; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s2 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_lshr_b32 s2, s0, 16 +; GFX12-NEXT: s_lshr_b32 s3, s1, 16 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: s_add_co_i32 s2, s2, s3 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: s_pack_ll_b32_b16 s0, s0, s2 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add <2 x i16> %a, %b + ret <2 x i16> %c +} + +define <2 x i16> @v_add_v2i16(<2 x i16> %a, <2 x i16> %b) { +; GFX7-LABEL: v_add_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_add_i32_e32 v1, vcc, v1, v3 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u16_e32 v2, v0, v1 +; GFX8-NEXT: v_add_u16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add <2 x i16> %a, %b + ret <2 x i16> %c +} + +define i64 @s_add_i64(i64 inreg %a, i64 inreg %b) { +; GFX7-LABEL: s_add_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_u32 s4, s16, s18 +; GFX7-NEXT: s_addc_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: v_mov_b32_e32 v1, s5 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_u32 s4, s16, s18 +; GFX9-NEXT: s_addc_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_u32 s4, s16, s18 +; GFX8-NEXT: s_addc_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: v_mov_b32_e32 v1, s5 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s4, s16, s18 +; GFX10-NEXT: s_addc_u32 s5, s17, s19 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: v_mov_b32_e32 v1, s5 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s0, s0, s2 +; GFX11-NEXT: s_addc_u32 s1, s1, s3 +; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_nc_u64 s[0:1], s[0:1], s[2:3] +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i64 %a, %b + ret i64 %c +} + +define i64 @v_add_i64(i64 %a, i64 %b) { +; GFX7-LABEL: v_add_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i64 %a, %b + ret i64 %c +} + +define void @s_uaddo_uadde(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: s_uaddo_uadde: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_u32 s4, s16, s18 +; GFX7-NEXT: s_addc_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v4, s4 +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_cselect_b32 s8, 1, 0 +; GFX7-NEXT: v_mov_b32_e32 v5, s5 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: buffer_store_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 +; GFX7-NEXT: v_mov_b32_e32 v0, s8 +; GFX7-NEXT: buffer_store_dword v0, v[2:3], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_uaddo_uadde: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_u32 s4, s16, s18 +; GFX9-NEXT: s_addc_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v4, s4 +; GFX9-NEXT: s_cselect_b32 s6, 1, 0 +; GFX9-NEXT: v_mov_b32_e32 v5, s5 +; GFX9-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX9-NEXT: v_mov_b32_e32 v0, s6 +; GFX9-NEXT: global_store_dword v[2:3], v0, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_uaddo_uadde: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_u32 s4, s16, s18 +; GFX8-NEXT: s_addc_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v4, s4 +; GFX8-NEXT: s_cselect_b32 s6, 1, 0 +; GFX8-NEXT: v_mov_b32_e32 v5, s5 +; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; GFX8-NEXT: v_mov_b32_e32 v0, s6 +; GFX8-NEXT: flat_store_dword v[2:3], v0 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_uaddo_uadde: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s4, s16, s18 +; GFX10-NEXT: s_addc_u32 s5, s17, s19 +; GFX10-NEXT: s_cselect_b32 s6, 1, 0 +; GFX10-NEXT: v_mov_b32_e32 v4, s4 +; GFX10-NEXT: v_mov_b32_e32 v5, s5 +; GFX10-NEXT: v_mov_b32_e32 v6, s6 +; GFX10-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX10-NEXT: global_store_dword v[2:3], v6, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_uaddo_uadde: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s0, s0, s2 +; GFX11-NEXT: s_addc_u32 s1, s1, s3 +; GFX11-NEXT: s_cselect_b32 s2, 1, 0 +; GFX11-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX11-NEXT: v_mov_b32_e32 v6, s2 +; GFX11-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX11-NEXT: global_store_b32 v[2:3], v6, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_uaddo_uadde: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_u32 s0, s0, s2 +; GFX12-NEXT: s_add_co_ci_u32 s1, s1, s3 +; GFX12-NEXT: s_cselect_b32 s2, 1, 0 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX12-NEXT: v_mov_b32_e32 v6, s2 +; GFX12-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX12-NEXT: global_store_b32 v[2:3], v6, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %uaddo = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) + %add = extractvalue {i64, i1} %uaddo, 0 + %of = extractvalue {i64, i1} %uaddo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %add, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +define void @v_uaddo_uadde(i64 %a, i64 %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: v_uaddo_uadde: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX7-NEXT: buffer_store_dwordx2 v[0:1], v[4:5], s[4:7], 0 addr64 +; GFX7-NEXT: buffer_store_dword v2, v[6:7], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_uaddo_uadde: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX9-NEXT: global_store_dword v[6:7], v2, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_uaddo_uadde: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1] +; GFX8-NEXT: flat_store_dword v[6:7], v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_uaddo_uadde: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX10-NEXT: global_store_dword v[6:7], v2, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_uaddo_uadde: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: global_store_b32 v[6:7], v2, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_uaddo_uadde: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: global_store_b32 v[6:7], v2, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %uaddo = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) + %add = extractvalue {i64, i1} %uaddo, 0 + %of = extractvalue {i64, i1} %uaddo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %add, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll new file mode 100644 index 0000000..e440bee --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll @@ -0,0 +1,165 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=-real-true16 -mcpu=gfx1100 -o - %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-FAKE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=+real-true16 -mcpu=gfx1100 -o - %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-TRUE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=-real-true16 -mcpu=gfx1200 -o - %s | FileCheck -check-prefixes=GCN,GFX12,GFX12-FAKE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=+real-true16 -mcpu=gfx1200 -o - %s | FileCheck -check-prefixes=GCN,GFX12,GFX12-TRUE16 %s + +define amdgpu_ps half @fadd_s16_uniform(half inreg %a, half inreg %b) { +; GFX11-FAKE16-LABEL: fadd_s16_uniform: +; GFX11-FAKE16: ; %bb.0: +; GFX11-FAKE16-NEXT: v_add_f16_e64 v0, s0, s1 +; GFX11-FAKE16-NEXT: ; return to shader part epilog +; +; GFX11-TRUE16-LABEL: fadd_s16_uniform: +; GFX11-TRUE16: ; %bb.0: +; GFX11-TRUE16-NEXT: v_add_f16_e64 v0.l, s0, s1 +; GFX11-TRUE16-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_s16_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f16 s0, s0, s1 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd half %a, %b + ret half %fadd +} + +define amdgpu_ps half @fadd_s16_div(half %a, half %b) { +; GFX11-FAKE16-LABEL: fadd_s16_div: +; GFX11-FAKE16: ; %bb.0: +; GFX11-FAKE16-NEXT: v_add_f16_e32 v0, v0, v1 +; GFX11-FAKE16-NEXT: ; return to shader part epilog +; +; GFX11-TRUE16-LABEL: fadd_s16_div: +; GFX11-TRUE16: ; %bb.0: +; GFX11-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l +; GFX11-TRUE16-NEXT: ; return to shader part epilog +; +; GFX12-FAKE16-LABEL: fadd_s16_div: +; GFX12-FAKE16: ; %bb.0: +; GFX12-FAKE16-NEXT: v_add_f16_e32 v0, v0, v1 +; GFX12-FAKE16-NEXT: ; return to shader part epilog +; +; GFX12-TRUE16-LABEL: fadd_s16_div: +; GFX12-TRUE16: ; %bb.0: +; GFX12-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l +; GFX12-TRUE16-NEXT: ; return to shader part epilog + %fadd = fadd half %a, %b + ret half %fadd +} + +define amdgpu_ps float @fadd_s32_uniform(float inreg %a, float inreg %b) { +; GFX11-LABEL: fadd_s32_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f32_e64 v0, s0, s1 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_s32_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f32 s0, s0, s1 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd float %a, %b + ret float %fadd +} + +define amdgpu_ps float @fadd_s32_div(float %a, float %b) { +; GCN-LABEL: fadd_s32_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_add_f32_e32 v0, v0, v1 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd float %a, %b + ret float %fadd +} + +define amdgpu_ps void @fadd_s64_uniform(double inreg %a, double inreg %b, ptr addrspace(1) %ptr) { +; GFX11-LABEL: fadd_s64_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f64 v[2:3], s[0:1], s[2:3] +; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off +; GFX11-NEXT: s_endpgm +; +; GFX12-LABEL: fadd_s64_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: v_add_f64_e64 v[2:3], s[0:1], s[2:3] +; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off +; GFX12-NEXT: s_endpgm + %fadd = fadd double %a, %b + store double %fadd, ptr addrspace(1) %ptr + ret void +} + +define amdgpu_ps void @fadd_s64_div(double %a, double %b, ptr addrspace(1) %ptr) { +; GFX11-LABEL: fadd_s64_div: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f64 v[0:1], v[0:1], v[2:3] +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: s_endpgm +; +; GFX12-LABEL: fadd_s64_div: +; GFX12: ; %bb.0: +; GFX12-NEXT: v_add_f64_e32 v[0:1], v[0:1], v[2:3] +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: s_endpgm + %fadd = fadd double %a, %b + store double %fadd, ptr addrspace(1) %ptr + ret void +} + +define amdgpu_ps <2 x half> @fadd_v2s16_uniform(<2 x half> inreg %a, <2 x half> inreg %b) { +; GFX11-LABEL: fadd_v2s16_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_pk_add_f16 v0, s0, s1 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_v2s16_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_lshr_b32 s2, s0, 16 +; GFX12-NEXT: s_lshr_b32 s3, s1, 16 +; GFX12-NEXT: s_add_f16 s0, s0, s1 +; GFX12-NEXT: s_add_f16 s1, s2, s3 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) | instskip(NEXT) | instid1(SALU_CYCLE_1) +; GFX12-NEXT: s_pack_ll_b32_b16 s0, s0, s1 +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd <2 x half> %a, %b + ret <2 x half> %fadd +} + +define amdgpu_ps <2 x half> @fadd_v2s16_div(<2 x half> %a, <2 x half> %b) { +; GCN-LABEL: fadd_v2s16_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_pk_add_f16 v0, v0, v1 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd <2 x half> %a, %b + ret <2 x half> %fadd +} + +define amdgpu_ps <2 x float> @fadd_v2s32_uniform(<2 x float> inreg %a, <2 x float> inreg %b) { +; GFX11-LABEL: fadd_v2s32_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f32_e64 v0, s0, s2 +; GFX11-NEXT: v_add_f32_e64 v1, s1, s3 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_v2s32_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f32 s0, s0, s2 +; GFX12-NEXT: s_add_f32 s1, s1, s3 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd <2 x float> %a, %b + ret <2 x float> %fadd +} + +define amdgpu_ps <2 x float> @fadd_v2s32_div(<2 x float> %a, <2 x float> %b) { +; GCN-LABEL: fadd_v2s32_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd <2 x float> %a, %b + ret <2 x float> %fadd +} diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll new file mode 100644 index 0000000..588802c --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll @@ -0,0 +1,66 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx700 < %s | FileCheck -check-prefixes=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx803 < %s | FileCheck -check-prefixes=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GFX11 %s + +define amdgpu_kernel void @fcmp_uniform_select(float %a, i32 %b, i32 %c, ptr addrspace(1) %out) { +; GFX7-LABEL: fcmp_uniform_select: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x9 +; GFX7-NEXT: s_load_dword s3, s[4:5], 0xb +; GFX7-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd +; GFX7-NEXT: s_mov_b32 s2, -1 +; GFX7-NEXT: s_waitcnt lgkmcnt(0) +; GFX7-NEXT: v_cmp_eq_f32_e64 s[4:5], s6, 0 +; GFX7-NEXT: s_or_b64 s[4:5], s[4:5], s[4:5] +; GFX7-NEXT: s_cselect_b32 s4, 1, 0 +; GFX7-NEXT: s_and_b32 s4, s4, 1 +; GFX7-NEXT: s_cmp_lg_u32 s4, 0 +; GFX7-NEXT: s_cselect_b32 s3, s7, s3 +; GFX7-NEXT: v_mov_b32_e32 v0, s3 +; GFX7-NEXT: s_mov_b32 s3, 0xf000 +; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7-NEXT: s_endpgm +; +; GFX8-LABEL: fcmp_uniform_select: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX8-NEXT: s_load_dword s6, s[4:5], 0x2c +; GFX8-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x34 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_cmp_eq_f32_e64 s[4:5], s0, 0 +; GFX8-NEXT: s_cmp_lg_u64 s[4:5], 0 +; GFX8-NEXT: s_cselect_b32 s0, 1, 0 +; GFX8-NEXT: s_and_b32 s0, s0, 1 +; GFX8-NEXT: s_cmp_lg_u32 s0, 0 +; GFX8-NEXT: s_cselect_b32 s0, s1, s6 +; GFX8-NEXT: v_mov_b32_e32 v0, s2 +; GFX8-NEXT: v_mov_b32_e32 v2, s0 +; GFX8-NEXT: v_mov_b32_e32 v1, s3 +; GFX8-NEXT: flat_store_dword v[0:1], v2 +; GFX8-NEXT: s_endpgm +; +; GFX11-LABEL: fcmp_uniform_select: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 +; GFX11-NEXT: s_load_b32 s6, s[4:5], 0x2c +; GFX11-NEXT: s_load_b64 s[2:3], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v1, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_cmp_eq_f32_e64 s0, s0, 0 +; GFX11-NEXT: s_cmp_lg_u32 s0, 0 +; GFX11-NEXT: s_cselect_b32 s0, 1, 0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1) +; GFX11-NEXT: s_and_b32 s0, s0, 1 +; GFX11-NEXT: s_cmp_lg_u32 s0, 0 +; GFX11-NEXT: s_cselect_b32 s0, s1, s6 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: global_store_b32 v1, v0, s[2:3] +; GFX11-NEXT: s_endpgm + %cmp = fcmp oeq float %a, 0.0 + %sel = select i1 %cmp, i32 %b, i32 %c + store i32 %sel, ptr addrspace(1) %out + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir new file mode 100644 index 0000000..b6652f6 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir @@ -0,0 +1,37 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=amdgcn -mcpu=gfx700 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GFX7 %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx803 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GF8 %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GFX11 %s + +--- +name: test_copy_scc_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.0: + ; GFX7-LABEL: name: test_copy_scc_vcc + ; GFX7: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF + ; GFX7-NEXT: [[S_OR_B64_:%[0-9]+]]:sreg_64 = S_OR_B64 [[DEF]], [[DEF]], implicit-def $scc + ; GFX7-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GFX7-NEXT: $sgpr0 = COPY [[COPY]] + ; GFX7-NEXT: S_ENDPGM 0, implicit $sgpr0 + ; + ; GF8-LABEL: name: test_copy_scc_vcc + ; GF8: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF + ; GF8-NEXT: S_CMP_LG_U64 [[DEF]], 0, implicit-def $scc + ; GF8-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GF8-NEXT: $sgpr0 = COPY [[COPY]] + ; GF8-NEXT: S_ENDPGM 0, implicit $sgpr0 + ; + ; GFX11-LABEL: name: test_copy_scc_vcc + ; GFX11: [[DEF:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF + ; GFX11-NEXT: S_CMP_LG_U32 [[DEF]], 0, implicit-def $scc + ; GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GFX11-NEXT: $sgpr0 = COPY [[COPY]] + ; GFX11-NEXT: S_ENDPGM 0, implicit $sgpr0 + %0:vcc(s1) = G_IMPLICIT_DEF + %1:sgpr(s32) = G_AMDGPU_COPY_SCC_VCC %0 + $sgpr0 = COPY %1 + S_ENDPGM 0, implicit $sgpr0 +... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll index 02d0e52..6facdfd 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll @@ -104,109 +104,110 @@ define amdgpu_cs <4 x i32> @abs_sgpr_v4i32(<4 x i32> inreg %arg) { ret <4 x i32> %res } -define amdgpu_cs i16 @abs_vgpr_i16(i16 %arg) { +define i16 @abs_vgpr_i16(i16 %arg) { ; GFX6-LABEL: abs_vgpr_i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u16_e32 v1, 0, v0 ; GFX8-NEXT: v_max_i16_e32 v0, v0, v1 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u16 v1, 0, v0 ; GFX10-NEXT: v_max_i16 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_sub_nc_u16 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_max_i16 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i16 @llvm.abs.i16(i16 %arg, i1 false) ret i16 %res } -define amdgpu_cs i32 @abs_vgpr_i32(i32 %arg) { +define i32 @abs_vgpr_i32(i32 %arg) { ; GFX6-LABEL: abs_vgpr_i32: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i32: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u32_e32 v1, vcc, 0, v0 ; GFX8-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i32: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u32_e32 v1, 0, v0 ; GFX10-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i32: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_sub_nc_u32_e32 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i32 @llvm.abs.i32(i32 %arg, i1 false) ret i32 %res } -define amdgpu_cs i64 @abs_vgpr_i64(i64 %arg) { +define i64 @abs_vgpr_i64(i64 %arg) { ; GFX6-LABEL: abs_vgpr_i64: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v2 ; GFX6-NEXT: v_addc_u32_e32 v1, vcc, v1, v2, vcc ; GFX6-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX6-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i64: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 ; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v2, vcc ; GFX8-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX8-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i64: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v2, vcc_lo ; GFX10-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i64: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) ; GFX1250-NEXT: v_mov_b32_e32 v3, v2 @@ -214,17 +215,15 @@ define amdgpu_cs i64 @abs_vgpr_i64(i64 %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX1250-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i64 @llvm.abs.i64(i64 %arg, i1 false) ret i64 %res } -define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { +define <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX6-LABEL: abs_vgpr_v4i32: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v4 ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v1 @@ -233,14 +232,11 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX6-NEXT: v_max_i32_e32 v2, v2, v4 ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v3 ; GFX6-NEXT: v_max_i32_e32 v3, v3, v4 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: v_readfirstlane_b32 s3, v3 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v4i32: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v0 ; GFX8-NEXT: v_max_i32_e32 v0, v0, v4 ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v1 @@ -249,14 +245,11 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX8-NEXT: v_max_i32_e32 v2, v2, v4 ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v3 ; GFX8-NEXT: v_max_i32_e32 v3, v3, v4 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: v_readfirstlane_b32 s2, v2 -; GFX8-NEXT: v_readfirstlane_b32 s3, v3 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v4i32: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u32_e32 v4, 0, v0 ; GFX10-NEXT: v_sub_nc_u32_e32 v5, 0, v1 ; GFX10-NEXT: v_sub_nc_u32_e32 v6, 0, v2 @@ -265,14 +258,12 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX10-NEXT: v_max_i32_e32 v1, v1, v5 ; GFX10-NEXT: v_max_i32_e32 v2, v2, v6 ; GFX10-NEXT: v_max_i32_e32 v3, v3, v7 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: v_readfirstlane_b32 s2, v2 -; GFX10-NEXT: v_readfirstlane_b32 s3, v3 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v4i32: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_dual_sub_nc_u32 v4, 0, v0 :: v_dual_sub_nc_u32 v5, 0, v1 ; GFX1250-NEXT: v_dual_sub_nc_u32 v6, 0, v2 :: v_dual_sub_nc_u32 v7, 0, v3 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) @@ -281,13 +272,7 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) ; GFX1250-NEXT: v_max_i32_e32 v2, v2, v6 ; GFX1250-NEXT: v_max_i32_e32 v3, v3, v7 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX1250-NEXT: v_readfirstlane_b32 s2, v2 -; GFX1250-NEXT: v_readfirstlane_b32 s3, v3 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %arg, i1 false) ret <4 x i32> %res } @@ -304,44 +289,43 @@ define amdgpu_cs <2 x i8> @abs_sgpr_v2i8(<2 x i8> inreg %arg) { ret <2 x i8> %res } -define amdgpu_cs <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { +define <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { ; GFX6-LABEL: abs_vgpr_v2i8: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v2 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v1 ; GFX6-NEXT: v_max_i32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v2i8: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v2, 0 ; GFX8-NEXT: v_sub_u16_sdwa v3, v2, sext(v0) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_sub_u16_sdwa v2, v2, sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v0, sext(v0), v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX8-NEXT: v_max_i16_sdwa v1, sext(v1), v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v2i8: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX10-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX10-NEXT: v_sub_nc_u16 v2, 0, v0 ; GFX10-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX10-NEXT: v_max_i16 v0, v0, v2 ; GFX10-NEXT: v_max_i16 v1, v1, v3 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v2i8: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) @@ -350,10 +334,7 @@ define amdgpu_cs <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_max_i16 v0, v0, v2 ; GFX1250-NEXT: v_max_i16 v1, v1, v3 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <2 x i8> @llvm.abs.v2i8(<2 x i8> %arg, i1 false) ret <2 x i8> %res } @@ -372,9 +353,10 @@ define amdgpu_cs <3 x i8> @abs_sgpr_v3i8(<3 x i8> inreg %arg) { ret <3 x i8> %res } -define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { +define <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX6-LABEL: abs_vgpr_v3i8: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 8 @@ -384,13 +366,11 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX6-NEXT: v_max_i32_e32 v1, v1, v3 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v2 ; GFX6-NEXT: v_max_i32_e32 v2, v2, v3 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v3i8: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v3, 0 ; GFX8-NEXT: v_sub_u16_sdwa v4, v3, sext(v0) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v0, sext(v0), v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD @@ -398,13 +378,11 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX8-NEXT: v_sub_u16_sdwa v3, v3, sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v1, sext(v1), v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX8-NEXT: v_max_i16_sdwa v2, sext(v2), v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: v_readfirstlane_b32 s2, v2 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v3i8: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX10-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX10-NEXT: v_bfe_i32 v2, v2, 0, 8 @@ -414,13 +392,12 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX10-NEXT: v_max_i16 v0, v0, v3 ; GFX10-NEXT: v_max_i16 v1, v1, v4 ; GFX10-NEXT: v_max_i16 v2, v2, v5 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: v_readfirstlane_b32 s2, v2 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v3i8: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v2, v2, 0, 8 @@ -433,12 +410,7 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) ; GFX1250-NEXT: v_max_i16 v1, v1, v4 ; GFX1250-NEXT: v_max_i16 v2, v2, v5 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) -; GFX1250-NEXT: v_readfirstlane_b32 s2, v2 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <3 x i8> @llvm.abs.v3i8(<3 x i8> %arg, i1 false) ret <3 x i8> %res } @@ -485,44 +457,44 @@ define amdgpu_cs <2 x i16> @abs_sgpr_v2i16(<2 x i16> inreg %arg) { ret <2 x i16> %res } -define amdgpu_cs <2 x i16> @abs_vgpr_v2i16(<2 x i16> %arg) { +define <2 x i16> @abs_vgpr_v2i16(<2 x i16> %arg) { ; GFX6-LABEL: abs_vgpr_v2i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 16 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v2 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v1 ; GFX6-NEXT: v_max_i32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v2i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v2, 0 ; GFX8-NEXT: v_sub_u16_e32 v1, 0, v0 ; GFX8-NEXT: v_sub_u16_sdwa v2, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX8-NEXT: v_max_i16_e32 v1, v0, v1 ; GFX8-NEXT: v_max_i16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_e32 v0, v1, v0 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v2i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_pk_sub_i16 v1, 0, v0 ; GFX10-NEXT: v_pk_max_i16 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v2i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_pk_sub_i16 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_pk_max_i16 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <2 x i16> @llvm.abs.v2i16(<2 x i16> %arg, i1 false) ret <2 x i16> %res } @@ -576,9 +548,10 @@ define amdgpu_cs <3 x i16> @abs_sgpr_v3i16(<3 x i16> inreg %arg) { ret <3 x i16> %res } -define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { +define <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX6-LABEL: abs_vgpr_v3i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 16 @@ -588,13 +561,11 @@ define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX6-NEXT: v_max_i32_e32 v1, v1, v3 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v2 ; GFX6-NEXT: v_max_i32_e32 v2, v2, v3 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v3i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v3, 0 ; GFX8-NEXT: v_sub_u16_e32 v2, 0, v0 ; GFX8-NEXT: v_sub_u16_sdwa v3, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 @@ -603,31 +574,27 @@ define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX8-NEXT: v_max_i16_sdwa v0, v0, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 ; GFX8-NEXT: v_max_i16_e32 v1, v1, v4 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v3i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_pk_sub_i16 v2, 0, v0 ; GFX10-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX10-NEXT: v_pk_max_i16 v0, v0, v2 ; GFX10-NEXT: v_max_i16 v1, v1, v3 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v3i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_pk_sub_i16 v2, 0, v0 ; GFX1250-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_pk_max_i16 v0, v0, v2 ; GFX1250-NEXT: v_max_i16 v1, v1, v3 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <3 x i16> @llvm.abs.v3i16(<3 x i16> %arg, i1 false) ret <3 x i16> %res } diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll index 7714c03..d3e2118 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll @@ -113,9 +113,9 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -161,16 +161,17 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 ; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -208,11 +209,7 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -258,17 +255,13 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -310,14 +303,12 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_le_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -372,16 +363,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 ; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll index 7b81669..250fbc7 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll @@ -116,9 +116,9 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -164,16 +164,17 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -211,11 +212,7 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -261,17 +258,13 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -313,14 +306,12 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_le_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -375,16 +366,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir new file mode 100644 index 0000000..097372a --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir @@ -0,0 +1,524 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s +--- +name: add_s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: add_s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC]](s16) + ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[ANYEXT]], [[ANYEXT1]] + ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[ADD]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[COPY2]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: add_s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: add_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s32_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: add_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s64_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $sgpr2_sgpr3 + ; CHECK-LABEL: name: add_s64_ss + ; CHECK: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr2_sgpr3 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s64) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 255 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s64) = G_AND [[ADD]], [[ADD]] + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $sgpr2_sgpr3 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_CONSTANT i64 255 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: add_s64_sv + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY]](s64) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $vgpr0_vgpr1 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: add_s64_vs + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY1]](s64) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-LABEL: name: add_s64_vv + ; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr2_vgpr3 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $vgpr2_vgpr3 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: uaddo_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: uaddo_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:sgpr(s32), [[UADDO1:%[0-9]+]]:sgpr(s32) = G_UADDO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[UADDO1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[SELECT]], [[UADDO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %4, %2 +... + +--- +name: uaddo_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1 + ; CHECK-LABEL: name: uaddo_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uaddo_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1 + ; CHECK-LABEL: name: uaddo_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uaddo_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: uaddo_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uadde_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s32) = G_UADDE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND1]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND2:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY3]], [[COPY1]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY]], [[COPY3]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-LABEL: name: uadde_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[ICMP:%[0-9]+]]:vcc(s1) = G_ICMP intpred(ne), [[AND]](s32), [[C1]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY]], [[COPY1]], [[ICMP]] + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $vgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_ss_scc_use +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_ss_scc_use + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s32) = G_UADDE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND1]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND2:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %8:_(s32) = G_AND %4, %6 +... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir index 54ee69f..30c958f 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir @@ -1,6 +1,5 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s --- name: add_s16_ss legalized: true @@ -19,13 +18,13 @@ body: | ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[ANYEXT]], [[ANYEXT1]] ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[ADD]](s32) - ; CHECK-NEXT: S_ENDPGM 0, implicit [[TRUNC2]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -44,13 +43,13 @@ body: | ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[COPY2]], [[TRUNC1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $vgpr0 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -69,13 +68,13 @@ body: | ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[COPY2]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $sgpr0 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -93,11 +92,11 @@ body: | ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[TRUNC1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir index 97018fa..01eb391 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir @@ -1,6 +1,5 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s -# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s --- name: add_v2s16_ss @@ -18,16 +17,19 @@ body: | ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 ; CHECK-NEXT: [[LSHR:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY1]](<2 x s16>) - ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 - ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) + ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[BITCAST]], [[BITCAST1]] ; CHECK-NEXT: [[ADD1:%[0-9]+]]:sgpr(s32) = G_ADD [[LSHR]], [[LSHR1]] ; CHECK-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ADD]](s32), [[ADD1]](s32) - ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR_TRUNC]](<2 x s16>) + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR [[C1]](s16), [[C1]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR]] %0:_(<2 x s16>) = COPY $sgpr0 %1:_(<2 x s16>) = COPY $sgpr1 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(<2 x s16>) = G_BUILD_VECTOR %3, %3 + %5:_(<2 x s16>) = G_AND %2, %4 ... --- @@ -44,11 +46,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY]](<2 x s16>) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY2]], [[COPY1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](<2 x s16>) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $sgpr0 %1:_(<2 x s16>) = COPY $vgpr0 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(<2 x s16>) = G_AND %2, %2 ... --- @@ -65,9 +67,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY1]](<2 x s16>) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $sgpr0 %2:_(<2 x s16>) = G_ADD %0, %1 + %3:_(<2 x s16>) = G_AND %2, %2 ... --- @@ -83,9 +87,9 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr1 ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY]], [[COPY1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](<2 x s16>) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(<2 x s16>) = G_AND %2, %2 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir index 7378c93..e0e783e 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir @@ -77,10 +77,14 @@ body: | ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s1) = G_ICMP intpred(eq), %0, %1 %3:_(s16) = G_SEXT %2 + %4:_(s16) = G_CONSTANT i16 255 + %5:_(s16) = G_AND %3, %4 ... --- @@ -215,9 +219,13 @@ body: | ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]] %0:_(s32) = COPY $sgpr0 %1:_(s1) = G_TRUNC %0 %2:_(s16) = G_SEXT %1 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(s16) = G_AND %2, %3 ... --- diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir index b0199d3..e3c01c0 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir @@ -1,5 +1,107 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass="amdgpu-regbankselect,amdgpu-regbanklegalize" %s -o - | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s + +--- +name: sub_s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: sub_s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC]](s16) + ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[ANYEXT]], [[ANYEXT1]] + ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SUB]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[COPY2]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[TRUNC]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: sub_s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[TRUNC]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... --- name: sub_s32_ss @@ -14,9 +116,11 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -33,9 +137,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $vgpr0 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -52,9 +158,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $sgpr0 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -70,7 +178,376 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 +... + +--- +name: sub_v2s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: sub_v2s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr1 + ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY]](<2 x s16>) + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 + ; CHECK-NEXT: [[LSHR:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY1]](<2 x s16>) + ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[BITCAST]], [[BITCAST1]] + ; CHECK-NEXT: [[SUB1:%[0-9]+]]:sgpr(s32) = G_SUB [[LSHR]], [[LSHR1]] + ; CHECK-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[SUB]](s32), [[SUB1]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC]] + %0:_(<2 x s16>) = COPY $sgpr0 + %1:_(<2 x s16>) = COPY $sgpr1 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_v2s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY]](<2 x s16>) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $sgpr0 + %1:_(<2 x s16>) = COPY $vgpr0 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_v2s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY1]](<2 x s16>) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $vgpr0 + %1:_(<2 x s16>) = COPY $sgpr0 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: sub_v2s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr1 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $vgpr0 + %1:_(<2 x s16>) = COPY $vgpr1 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_s64_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $sgpr0_sgpr1 + ; CHECK-LABEL: name: sub_s64_ss + ; CHECK: liveins: $sgpr0_sgpr1, $sgpr0_sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s64) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s64) = G_AND [[SUB]], [[SUB]] + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: sub_s64_sv + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY]](s64) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $vgpr0_vgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: sub_s64_vs + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY1]](s64) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-LABEL: name: sub_s64_vv + ; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr2_vgpr3 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $vgpr2_vgpr3 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: usubo_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: usubo_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:sgpr(s32), [[USUBO1:%[0-9]+]]:sgpr(s32) = G_USUBO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1 + ; CHECK-LABEL: name: usubo_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1 + ; CHECK-LABEL: name: usubo_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: usubo_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usube_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:sgpr(s32), [[USUBE1:%[0-9]+]]:sgpr(s32) = G_USUBE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY3]], [[COPY1]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY]], [[COPY3]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-LABEL: name: usube_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[ICMP:%[0-9]+]]:vcc(s1) = G_ICMP intpred(ne), [[AND]](s32), [[C1]] + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY]], [[COPY1]], [[ICMP]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $vgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir index 088c20a3..d4baa5f 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir @@ -73,10 +73,14 @@ body: | ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s1) = G_ICMP intpred(eq), %0, %1 %3:_(s16) = G_ZEXT %2 + %4:_(s16) = G_CONSTANT i16 255 + %5:_(s16) = G_AND %3, %4 ... --- @@ -209,9 +213,13 @@ body: | ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]] %0:_(s32) = COPY $sgpr0 %1:_(s1) = G_TRUNC %0 %2:_(s16) = G_ZEXT %1 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(s16) = G_AND %2, %3 ... --- diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll new file mode 100644 index 0000000..8b5958d --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll @@ -0,0 +1,535 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=hawaii < %s | FileCheck -check-prefix=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck -check-prefix=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX11 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX12 %s + +define i16 @s_sub_i16(i16 inreg %a, i16 inreg %b) { +; GFX7-LABEL: s_sub_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_i32 s4, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_i32 s4, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_i32 s4, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_i32 s4, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i16 %a, %b + ret i16 %c +} + +define i16 @v_sub_i16(i16 %a, i16 %b) { +; GFX7-LABEL: v_sub_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u16_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u16_e32 v0, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u16 v0.l, v0.l, v1.l +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_nc_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i16 %a, %b + ret i16 %c +} + +define i32 @s_sub_i32(i32 inreg %a, i32 inreg %b) { +; GFX7-LABEL: s_sub_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_i32 s4, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_i32 s4, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_i32 s4, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_i32 s4, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i32 %a, %b + ret i32 %c +} + +define i32 @v_sub_i32(i32 %a, i32 %b) { +; GFX7-LABEL: v_sub_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i32 %a, %b + ret i32 %c +} + +; TODO: Add test for s_sub_v2i16. Instruction selector currently fails +; to handle G_UNMERGE_VALUES. + +define <2 x i16> @v_sub_v2i16(<2 x i16> %a, <2 x i16> %b) { +; GFX7-LABEL: v_sub_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_sub_i32_e32 v1, vcc, v1, v3 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u16_e32 v2, v0, v1 +; GFX8-NEXT: v_sub_u16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub <2 x i16> %a, %b + ret <2 x i16> %c +} + +define i64 @s_sub_i64(i64 inreg %a, i64 inreg %b) { +; GFX7-LABEL: s_sub_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_u32 s4, s16, s18 +; GFX7-NEXT: s_subb_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: v_mov_b32_e32 v1, s5 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s4, s16, s18 +; GFX9-NEXT: s_subb_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_u32 s4, s16, s18 +; GFX8-NEXT: s_subb_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: v_mov_b32_e32 v1, s5 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s16, s18 +; GFX10-NEXT: s_subb_u32 s5, s17, s19 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: v_mov_b32_e32 v1, s5 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s0, s0, s2 +; GFX11-NEXT: s_subb_u32 s1, s1, s3 +; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_nc_u64 s[0:1], s[0:1], s[2:3] +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i64 %a, %b + ret i64 %c +} + +define i64 @v_sub_i64(i64 %a, i64 %b) { +; GFX7-LABEL: v_sub_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_sub_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_sub_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i64 %a, %b + ret i64 %c +} + +define void @s_usubo_usube(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: s_usubo_usube: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_u32 s4, s16, s18 +; GFX7-NEXT: s_subb_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v4, s4 +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_cselect_b32 s8, 1, 0 +; GFX7-NEXT: v_mov_b32_e32 v5, s5 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: buffer_store_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 +; GFX7-NEXT: v_mov_b32_e32 v0, s8 +; GFX7-NEXT: buffer_store_dword v0, v[2:3], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_usubo_usube: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s4, s16, s18 +; GFX9-NEXT: s_subb_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v4, s4 +; GFX9-NEXT: s_cselect_b32 s6, 1, 0 +; GFX9-NEXT: v_mov_b32_e32 v5, s5 +; GFX9-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX9-NEXT: v_mov_b32_e32 v0, s6 +; GFX9-NEXT: global_store_dword v[2:3], v0, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_usubo_usube: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_u32 s4, s16, s18 +; GFX8-NEXT: s_subb_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v4, s4 +; GFX8-NEXT: s_cselect_b32 s6, 1, 0 +; GFX8-NEXT: v_mov_b32_e32 v5, s5 +; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; GFX8-NEXT: v_mov_b32_e32 v0, s6 +; GFX8-NEXT: flat_store_dword v[2:3], v0 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_usubo_usube: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s16, s18 +; GFX10-NEXT: s_subb_u32 s5, s17, s19 +; GFX10-NEXT: s_cselect_b32 s6, 1, 0 +; GFX10-NEXT: v_mov_b32_e32 v4, s4 +; GFX10-NEXT: v_mov_b32_e32 v5, s5 +; GFX10-NEXT: v_mov_b32_e32 v6, s6 +; GFX10-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX10-NEXT: global_store_dword v[2:3], v6, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_usubo_usube: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s0, s0, s2 +; GFX11-NEXT: s_subb_u32 s1, s1, s3 +; GFX11-NEXT: s_cselect_b32 s2, 1, 0 +; GFX11-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX11-NEXT: v_mov_b32_e32 v6, s2 +; GFX11-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX11-NEXT: global_store_b32 v[2:3], v6, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_usubo_usube: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_u32 s0, s0, s2 +; GFX12-NEXT: s_sub_co_ci_u32 s1, s1, s3 +; GFX12-NEXT: s_cselect_b32 s2, 1, 0 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX12-NEXT: v_mov_b32_e32 v6, s2 +; GFX12-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX12-NEXT: global_store_b32 v[2:3], v6, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %usubo = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b) + %sub = extractvalue {i64, i1} %usubo, 0 + %of = extractvalue {i64, i1} %usubo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %sub, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +define void @v_usubo_usube(i64 %a, i64 %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: v_usubo_usube: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX7-NEXT: buffer_store_dwordx2 v[0:1], v[4:5], s[4:7], 0 addr64 +; GFX7-NEXT: buffer_store_dword v2, v[6:7], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_usubo_usube: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX9-NEXT: global_store_dword v[6:7], v2, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_usubo_usube: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1] +; GFX8-NEXT: flat_store_dword v[6:7], v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_usubo_usube: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX10-NEXT: global_store_dword v[6:7], v2, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_usubo_usube: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: global_store_b32 v[6:7], v2, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_usubo_usube: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: global_store_b32 v[6:7], v2, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %usubo = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b) + %sub = extractvalue {i64, i1} %usubo, 0 + %of = extractvalue {i64, i1} %usubo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %sub, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll new file mode 100644 index 0000000..34d4c51 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll @@ -0,0 +1,173 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -o - %s | FileCheck %s +define amdgpu_kernel void @readfirstlane_with_readfirstlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readfirstlane_with_readfirstlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 5 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %v1 = call i32 @llvm.amdgcn.readfirstlane(i32 5) + %v2 = call i32 @llvm.amdgcn.readfirstlane(i32 %v1) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readfirstlane_with_readlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readfirstlane_with_readlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_bfe_u32 v1, v0, 10, 10 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v1 +; CHECK-NEXT: v_readlane_b32 s2, v0, s2 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %tidy = call i32 @llvm.amdgcn.workitem.id.y() + %v1 = call i32 @llvm.amdgcn.readlane(i32 %tidx, i32 %tidy) + %v2 = call i32 @llvm.amdgcn.readfirstlane(i32 %v1) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readlane_with_firstlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readlane_with_firstlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %v1 = call i32 @llvm.amdgcn.readfirstlane(i32 %tidx) + %v2 = call i32 @llvm.amdgcn.readlane(i32 %v1, i32 3) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readlane_readlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readlane_readlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_bfe_u32 v1, v0, 10, 10 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v1 +; CHECK-NEXT: v_readlane_b32 s2, v0, s2 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %tidy = call i32 @llvm.amdgcn.workitem.id.y() + %v1 = call i32 @llvm.amdgcn.readlane(i32 %tidx, i32 %tidy) + %v2 = call i32 @llvm.amdgcn.readlane(i32 %v1, i32 2) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @permlane64_uniform(ptr addrspace(1) %out, i32 %src) { +; CHECK-LABEL: permlane64_uniform: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_clause 0x1 +; CHECK-NEXT: s_load_b32 s2, s[4:5], 0x8 +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %v = call i32 @llvm.amdgcn.permlane64(i32 %src) + store i32 %v, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @permlane64_nonuniform(i32 addrspace(1)* %out) { +; CHECK-LABEL: permlane64_nonuniform: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_permlane64_b32 v1, v0 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tid = call i32 @llvm.amdgcn.workitem.id.x() + %v = call i32 @llvm.amdgcn.permlane64(i32 %tid) + %out_ptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid + store i32 %v, i32 addrspace(1)* %out_ptr + ret void +} + +define amdgpu_kernel void @permlane64_nonuniform_expression(i32 addrspace(1)* %out) { +; CHECK-LABEL: permlane64_nonuniform_expression: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) +; CHECK-NEXT: v_add_nc_u32_e32 v1, 1, v0 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 +; CHECK-NEXT: v_permlane64_b32 v1, v1 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tid = call i32 @llvm.amdgcn.workitem.id.x() + %tid2 = add i32 %tid, 1 + %v = call i32 @llvm.amdgcn.permlane64(i32 %tid2) + %out_ptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid + store i32 %v, i32 addrspace(1)* %out_ptr + ret void +} + +define protected amdgpu_kernel void @trivial_waterfall_eq_zero(ptr addrspace(1) %out) { +; CHECK-LABEL: trivial_waterfall_eq_zero: +; CHECK: ; %bb.0: ; %entry +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 5 +; CHECK-NEXT: s_mov_b32 s2, 0 +; CHECK-NEXT: s_branch .LBB7_2 +; CHECK-NEXT: .LBB7_1: ; %Flow +; CHECK-NEXT: ; in Loop: Header=BB7_2 Depth=1 +; CHECK-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s2 +; CHECK-NEXT: s_mov_b32 s2, -1 +; CHECK-NEXT: s_cbranch_vccz .LBB7_4 +; CHECK-NEXT: .LBB7_2: ; %while +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s2 +; CHECK-NEXT: s_mov_b32 s2, -1 +; CHECK-NEXT: s_cbranch_vccnz .LBB7_1 +; CHECK-NEXT: ; %bb.3: ; %if +; CHECK-NEXT: ; in Loop: Header=BB7_2 Depth=1 +; CHECK-NEXT: s_mov_b32 s2, 0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_branch .LBB7_1 +; CHECK-NEXT: .LBB7_4: ; %exit +; CHECK-NEXT: s_endpgm +entry: + br label %while + +while: + %done = phi i1 [ 0, %entry ], [ 1, %if ] + %not_done = xor i1 %done, true + %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done) + %is_done = icmp eq i64 %ballot, 0 ; in this case is_done = !not_done + br i1 %is_done, label %exit, label %if + +if: + store i32 5, ptr addrspace(1) %out + br label %while + +exit: + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll index 33ce278..c962c05 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll @@ -1,6 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-enable-uniform-intrinsic-combine=0 -O3 -S < %s | FileCheck %s -check-prefix=CURRENT-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O3 -S < %s | FileCheck %s -check-prefix=O3-CHECK define protected amdgpu_kernel void @trivial_waterfall_eq_zero(ptr addrspace(1) %out) { diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll index a3e42e5..a7e828c 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-enable-uniform-intrinsic-combine=0 -O3 -S < %s | FileCheck %s -check-prefix=CURRENT-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine,dce -S < %s | FileCheck %s -check-prefix=DCE-CHECK diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll index 2fde3e3..7929261 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine,instcombine,early-cse,simplifycfg -S < %s | FileCheck %s -check-prefix=COMB-CHECK ; This should not be optimized diff --git a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll index db32135..b8f084d 100644 --- a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll +++ b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll @@ -4,24 +4,14 @@ define amdgpu_gs i32 @main() { ; CHECK-LABEL: main: ; CHECK: ; %bb.0: ; %bb -; CHECK-NEXT: s_bitcmp1_b32 0, 0 ; CHECK-NEXT: s_mov_b32 s0, 0 -; CHECK-NEXT: s_cselect_b32 s1, -1, 0 -; CHECK-NEXT: s_or_saveexec_b32 s2, -1 -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s1 -; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) -; CHECK-NEXT: v_readfirstlane_b32 s1, v0 -; CHECK-NEXT: s_mov_b32 exec_lo, s2 -; CHECK-NEXT: s_or_b32 s0, s0, s1 -; CHECK-NEXT: s_wait_alu 0xfffe +; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) ; CHECK-NEXT: s_bitcmp1_b32 s0, 0 ; CHECK-NEXT: s_cselect_b32 s0, -1, 0 -; CHECK-NEXT: s_wait_alu 0xfffe ; CHECK-NEXT: s_xor_b32 s0, s0, -1 -; CHECK-NEXT: s_wait_alu 0xfffe -; CHECK-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 -; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) -; CHECK-NEXT: v_readfirstlane_b32 s0, v1 +; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; CHECK-NEXT: v_readfirstlane_b32 s0, v0 ; CHECK-NEXT: s_wait_alu 0xf1ff ; CHECK-NEXT: ; return to shader part epilog bb: diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll index 3aa3663..704ea37 100644 --- a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll +++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll @@ -9,11 +9,11 @@ ; RUN: | FileCheck -check-prefix=GCN-O3 %s -; GCN-O0: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O0>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(atomic-expand,verify,gc-lowering,lower-constant-intrinsics,unreachableblockelim,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,amdgpu-lower-kernel-arguments),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa,require<uniformity>,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,localstackalloc))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,phi-node-elimination,two-address-instruction,regallocfast,si-fix-vgpr-copies,remove-redundant-debug-values,fixup-statepoint-caller-saved,prolog-epilog,post-ra-pseudos,si-post-ra-bundler,fentry-insert,xray-instrumentation,patchable-function,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O0: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O0>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(atomic-expand,verify,gc-lowering,lower-constant-intrinsics,unreachableblockelim,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,amdgpu-lower-kernel-arguments),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa,require<uniformity>,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,localstackalloc))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,phi-node-elimination,two-address-instruction,regallocfast,si-fix-vgpr-copies,remove-redundant-debug-values,fixup-statepoint-caller-saved,prolog-epilog,post-ra-pseudos,si-post-ra-bundler,fentry-insert,xray-instrumentation,patchable-function,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) -; GCN-O2: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O2>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,early-cse<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,early-cse<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O2: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O2>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt,amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,early-cse<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,early-cse<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) -; GCN-O3: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O3>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,gvn<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,gvn<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O3: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O3>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt,amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,gvn<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,gvn<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) define void @empty() { ret void diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll index 6e52125..ee6caab 100644 --- a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll +++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll @@ -31,6 +31,11 @@ ; GCN-O0-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O0-NEXT: AMDGPU Printf lowering ; GCN-O0-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O0-NEXT: FunctionPass Manager +; GCN-O0-NEXT: Dominator Tree Construction +; GCN-O0-NEXT: Cycle Info Analysis +; GCN-O0-NEXT: Uniformity Analysis +; GCN-O0-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O0-NEXT: Expand variadic functions ; GCN-O0-NEXT: AMDGPU Inline All Functions ; GCN-O0-NEXT: Inliner for always_inline functions @@ -179,6 +184,11 @@ ; GCN-O1-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O1-NEXT: AMDGPU Printf lowering ; GCN-O1-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O1-NEXT: FunctionPass Manager +; GCN-O1-NEXT: Dominator Tree Construction +; GCN-O1-NEXT: Cycle Info Analysis +; GCN-O1-NEXT: Uniformity Analysis +; GCN-O1-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O1-NEXT: Expand variadic functions ; GCN-O1-NEXT: AMDGPU Inline All Functions ; GCN-O1-NEXT: Inliner for always_inline functions @@ -466,6 +476,11 @@ ; GCN-O1-OPTS-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O1-OPTS-NEXT: AMDGPU Printf lowering ; GCN-O1-OPTS-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O1-OPTS-NEXT: FunctionPass Manager +; GCN-O1-OPTS-NEXT: Dominator Tree Construction +; GCN-O1-OPTS-NEXT: Cycle Info Analysis +; GCN-O1-OPTS-NEXT: Uniformity Analysis +; GCN-O1-OPTS-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O1-OPTS-NEXT: Expand variadic functions ; GCN-O1-OPTS-NEXT: AMDGPU Inline All Functions ; GCN-O1-OPTS-NEXT: Inliner for always_inline functions @@ -783,6 +798,10 @@ ; GCN-O2-NEXT: Lower ctors and dtors for AMDGPU ; GCN-O2-NEXT: FunctionPass Manager ; GCN-O2-NEXT: AMDGPU Image Intrinsic Optimizer +; GCN-O2-NEXT: Dominator Tree Construction +; GCN-O2-NEXT: Cycle Info Analysis +; GCN-O2-NEXT: Uniformity Analysis +; GCN-O2-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O2-NEXT: Expand variadic functions ; GCN-O2-NEXT: AMDGPU Inline All Functions ; GCN-O2-NEXT: Inliner for always_inline functions @@ -1104,6 +1123,10 @@ ; GCN-O3-NEXT: Lower ctors and dtors for AMDGPU ; GCN-O3-NEXT: FunctionPass Manager ; GCN-O3-NEXT: AMDGPU Image Intrinsic Optimizer +; GCN-O3-NEXT: Dominator Tree Construction +; GCN-O3-NEXT: Cycle Info Analysis +; GCN-O3-NEXT: Uniformity Analysis +; GCN-O3-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O3-NEXT: Expand variadic functions ; GCN-O3-NEXT: AMDGPU Inline All Functions ; GCN-O3-NEXT: Inliner for always_inline functions diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll index e00e1f1..c1f3a12 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll @@ -110,9 +110,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc_lo, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB8_2 +; CHECK-NEXT: s_bitcmp0_b32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB8_3 @@ -156,15 +155,16 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc_lo, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_bitcmp1_b32 s0, 0 +; CHECK-NEXT: s_cselect_b32 s0, -1, 0 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -201,8 +201,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc_lo, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB12_2 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 +; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB12_3 @@ -245,14 +245,14 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc_lo, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -293,13 +293,13 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 ; CHECK-NEXT: s_cselect_b32 s0, -1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_lt_u32 s1, 35 ; CHECK-NEXT: s_cselect_b32 s1, -1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, s0, exec_lo -; CHECK-NEXT: s_cbranch_scc0 .LBB16_2 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB16_3 @@ -353,14 +353,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, -1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, s0, exec_lo -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 @@ -591,3 +591,24 @@ exit: store i32 %ballot, ptr addrspace(1) %out ret void } + +define amdgpu_cs i32 @compare_bfloats(bfloat %x, bfloat %y) { +; GFX10-LABEL: compare_bfloats: +; GFX10: ; %bb.0: +; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 +; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0 +; GFX10-NEXT: v_cmp_gt_f32_e64 s0, v0, v1 +; GFX10-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: compare_bfloats: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_mov_b16_e32 v2.l, 0 +; GFX11-NEXT: v_mov_b16_e32 v2.h, v1.l +; GFX11-NEXT: v_mov_b16_e32 v1.h, v0.l +; GFX11-NEXT: v_mov_b16_e32 v1.l, v2.l +; GFX11-NEXT: v_cmp_gt_f32_e64 s0, v1, v2 +; GFX11-NEXT: ; return to shader part epilog + %cmp = fcmp ogt bfloat %x, %y + %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %cmp) + ret i32 %ballot +} diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll index b4adf7f..827a01f 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll @@ -113,9 +113,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB8_2 +; CHECK-NEXT: s_bitcmp0_b32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB8_3 @@ -159,15 +158,16 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_bitcmp1_b32 s0, 0 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -204,8 +204,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB12_2 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 +; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB12_3 @@ -248,14 +248,14 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -296,13 +296,13 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 ; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_lt_u32 s1, 35 ; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 -; CHECK-NEXT: s_and_b64 s[0:1], s[2:3], s[0:1] -; CHECK-NEXT: s_and_b64 s[0:1], s[0:1], exec -; CHECK-NEXT: s_cbranch_scc0 .LBB16_2 +; CHECK-NEXT: s_or_b64 s[0:1], s[2:3], s[0:1] +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB16_3 @@ -356,14 +356,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 ; CHECK-NEXT: s_and_b64 s[0:1], s[2:3], s[0:1] -; CHECK-NEXT: s_and_b64 s[0:1], s[0:1], exec -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 @@ -557,3 +557,15 @@ exit: store i64 %ballot, ptr addrspace(1) %out ret void } + +define amdgpu_cs i64 @compare_bfloats(bfloat %x, bfloat %y) { +; CHECK-LABEL: compare_bfloats: +; CHECK: ; %bb.0: +; CHECK-NEXT: v_lshlrev_b32_e32 v1, 16, v1 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 16, v0 +; CHECK-NEXT: v_cmp_gt_f32_e64 s[0:1], v0, v1 +; CHECK-NEXT: ; return to shader part epilog + %cmp = fcmp ogt bfloat %x, %y + %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp) + ret i64 %ballot +} diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll index 6dd2258..39191d2 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll @@ -23,10 +23,8 @@ define amdgpu_kernel void @test_s_i32(ptr addrspace(1) %out, i32 %src0) { ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_i32: @@ -36,8 +34,6 @@ define amdgpu_kernel void @test_s_i32(ptr addrspace(1) %out, i32 %src0) { ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call i32 @llvm.amdgcn.permlane64.i32(i32 %src0) @@ -50,12 +46,9 @@ define amdgpu_kernel void @test_s_i64(ptr addrspace(1) %out, i64 %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_i64: @@ -64,9 +57,6 @@ define amdgpu_kernel void @test_s_i64(ptr addrspace(1) %out, i64 %src0) { ; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call i64 @llvm.amdgcn.permlane64.i64(i64 %src0) @@ -79,12 +69,9 @@ define amdgpu_kernel void @test_s_f64(ptr addrspace(1) %out, double %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_f64: @@ -93,9 +80,6 @@ define amdgpu_kernel void @test_s_f64(ptr addrspace(1) %out, double %src0) { ; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call double @llvm.amdgcn.permlane64.f64(double %src0) @@ -116,19 +100,15 @@ define amdgpu_kernel void @test_i_i32(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_i32: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0x63 :: v_dual_mov_b32 v1, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x63 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_i32: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, 0x63 :: v_dual_mov_b32 v1, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -141,19 +121,15 @@ define amdgpu_kernel void @test_i_f32(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_f32: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0x449a5000 :: v_dual_mov_b32 v1, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x449a5000 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_f32: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, 0x449a5000 :: v_dual_mov_b32 v1, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -166,23 +142,16 @@ define amdgpu_kernel void @test_i_i64(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_i64: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, 0x63 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v2 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, 0x63 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: global_store_b64 v1, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_i64: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_mov_b32_e32 v0, 0x63 -; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v2 +; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -195,22 +164,16 @@ define amdgpu_kernel void @test_i_f64(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_f64: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, 0x40934a00 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x40934a00 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: global_store_b64 v0, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_f64: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x40934a00 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v2 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 +; GFX11-GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0x40934a00 :: v_dual_mov_b32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll index b0149f7..672b658 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll @@ -6,12 +6,9 @@ define amdgpu_kernel void @test_p0(ptr addrspace(1) %out, ptr %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr @llvm.amdgcn.permlane64.p0(ptr %src0) store ptr %v, ptr addrspace(1) %out @@ -22,21 +19,14 @@ define amdgpu_kernel void @test_v3p0(ptr addrspace(1) %out, <3 x ptr> %src0) { ; GFX11-SDAG-LABEL: test_v3p0: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_clause 0x2 -; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x44 ; GFX11-SDAG-NEXT: s_load_b64 s[6:7], s[4:5], 0x54 +; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x44 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v1, s2 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s7 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v8, s6 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s3 :: v_dual_mov_b32 v7, s0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v1 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v4 -; GFX11-SDAG-NEXT: v_permlane64_b32 v5, v5 -; GFX11-SDAG-NEXT: v_permlane64_b32 v4, v8 -; GFX11-SDAG-NEXT: v_permlane64_b32 v3, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v7 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v5, s7 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v1, s1 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 ; GFX11-SDAG-NEXT: s_clause 0x1 ; GFX11-SDAG-NEXT: global_store_b64 v6, v[4:5], s[4:5] offset:16 ; GFX11-SDAG-NEXT: global_store_b128 v6, v[0:3], s[4:5] @@ -53,10 +43,8 @@ define amdgpu_kernel void @test_p3(ptr addrspace(1) %out, ptr addrspace(3) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(3) @llvm.amdgcn.permlane64.v3p0(ptr addrspace(3) %src0) store ptr addrspace(3) %v, ptr addrspace(1) %out @@ -70,14 +58,9 @@ define amdgpu_kernel void @test_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(3)> @llvm.amdgcn.permlane64.v3p3(<3 x ptr addrspace(3)> %src0) store <3 x ptr addrspace(3)> %v, ptr addrspace(1) %out @@ -91,10 +74,8 @@ define amdgpu_kernel void @test_p5(ptr addrspace(1) %out, ptr addrspace(5) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(5) @llvm.amdgcn.permlane64.p5(ptr addrspace(5) %src0) store ptr addrspace(5) %v, ptr addrspace(1) %out @@ -108,14 +89,9 @@ define amdgpu_kernel void @test_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(5)> @llvm.amdgcn.permlane64.v3p5(<3 x ptr addrspace(5)> %src0) store <3 x ptr addrspace(5)> %v, ptr addrspace(1) %out @@ -129,10 +105,8 @@ define amdgpu_kernel void @test_p6(ptr addrspace(1) %out, ptr addrspace(6) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(6) @llvm.amdgcn.permlane64.p6(ptr addrspace(6) %src0) store ptr addrspace(6) %v, ptr addrspace(1) %out @@ -146,14 +120,9 @@ define amdgpu_kernel void @test_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(6)> @llvm.amdgcn.permlane64.v3p6(<3 x ptr addrspace(6)> %src0) store <3 x ptr addrspace(6)> %v, ptr addrspace(1) %out diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll index d1ba892..02d2990 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll @@ -396,8 +396,7 @@ define amdgpu_kernel void @test_readfirstlane_imm_f64(ptr addrspace(1) %out) { ; ; CHECK-GISEL-LABEL: test_readfirstlane_imm_f64: ; CHECK-GISEL: ; %bb.0: -; CHECK-GISEL-NEXT: s_mov_b32 s0, 0 -; CHECK-GISEL-NEXT: s_mov_b32 s1, 0x40400000 +; CHECK-GISEL-NEXT: s_mov_b64 s[0:1], 0x4040000000000000 ; CHECK-GISEL-NEXT: ;;#ASMSTART ; CHECK-GISEL-NEXT: ; use s[0:1] ; CHECK-GISEL-NEXT: ;;#ASMEND @@ -456,14 +455,13 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_i64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_i64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b64 s[2:3], 32 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 32 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -490,15 +488,13 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_f64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_f64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b32 s2, 0 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: s_mov_b32 s3, 0x40400000 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 0 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0x40400000 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -588,17 +584,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_i64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_i64: @@ -628,17 +624,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_f64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_f64: diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll index 7ff5eb4..0795f40 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll @@ -9,7 +9,7 @@ declare double @llvm.amdgcn.readlane.f64(double, i32) #0 define amdgpu_kernel void @test_readlane_sreg_sreg_i32(i32 %src0, i32 %src1) #1 { ; CHECK-SDAG-LABEL: test_readlane_sreg_sreg_i32: ; CHECK-SDAG: ; %bb.0: -; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-SDAG-NEXT: s_load_dword s0, s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: ; use s0 @@ -18,7 +18,7 @@ define amdgpu_kernel void @test_readlane_sreg_sreg_i32(i32 %src0, i32 %src1) #1 ; ; CHECK-GISEL-LABEL: test_readlane_sreg_sreg_i32: ; CHECK-GISEL: ; %bb.0: -; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-GISEL-NEXT: s_load_dword s0, s[8:9], 0x0 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: ;;#ASMSTART ; CHECK-GISEL-NEXT: ; use s0 @@ -224,14 +224,13 @@ define amdgpu_kernel void @test_readlane_imm_sreg_i64(ptr addrspace(1) %out, i32 ; CHECK-GISEL-LABEL: test_readlane_imm_sreg_i64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b64 s[2:3], 32 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 32 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -258,15 +257,13 @@ define amdgpu_kernel void @test_readlane_imm_sreg_f64(ptr addrspace(1) %out, i32 ; CHECK-GISEL-LABEL: test_readlane_imm_sreg_f64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b32 s2, 0 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: s_mov_b32 s3, 0x40400000 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 0 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0x40400000 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -660,17 +657,17 @@ define amdgpu_kernel void @test_readlane_copy_from_sgpr_i64(ptr addrspace(1) %ou ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_i64: @@ -700,17 +697,17 @@ define amdgpu_kernel void @test_readlane_copy_from_sgpr_f64(ptr addrspace(1) %ou ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_f64: diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll new file mode 100644 index 0000000..05a0e39 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll @@ -0,0 +1,325 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 +; RUN: opt -S -mtriple=amdgcn-- -mcpu=gfx1100 -passes=amdgpu-promote-alloca < %s | FileCheck -check-prefix=OPT %s + +define amdgpu_kernel void @large_array_vectors_small_users(<16 x i8> %in, <16 x i8> %add, ptr addrspace(3) %out) #0 { +; OPT-LABEL: define amdgpu_kernel void @large_array_vectors_small_users( +; OPT-SAME: <16 x i8> [[IN:%.*]], <16 x i8> [[ADD:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] { +; OPT-NEXT: [[ENTRY:.*:]] +; OPT-NEXT: [[ALLOCA:%.*]] = freeze <128 x i8> poison +; OPT-NEXT: [[TMP0:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP1:%.*]] = insertelement <128 x i8> [[ALLOCA]], i8 [[TMP0]], i32 0 +; OPT-NEXT: [[TMP2:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP3:%.*]] = insertelement <128 x i8> [[TMP1]], i8 [[TMP2]], i32 1 +; OPT-NEXT: [[TMP4:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP5:%.*]] = insertelement <128 x i8> [[TMP3]], i8 [[TMP4]], i32 2 +; OPT-NEXT: [[TMP6:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP7:%.*]] = insertelement <128 x i8> [[TMP5]], i8 [[TMP6]], i32 3 +; OPT-NEXT: [[TMP8:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP9:%.*]] = insertelement <128 x i8> [[TMP7]], i8 [[TMP8]], i32 4 +; OPT-NEXT: [[TMP10:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP11:%.*]] = insertelement <128 x i8> [[TMP9]], i8 [[TMP10]], i32 5 +; OPT-NEXT: [[TMP12:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP13:%.*]] = insertelement <128 x i8> [[TMP11]], i8 [[TMP12]], i32 6 +; OPT-NEXT: [[TMP14:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP15:%.*]] = insertelement <128 x i8> [[TMP13]], i8 [[TMP14]], i32 7 +; OPT-NEXT: [[TMP16:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP17:%.*]] = insertelement <128 x i8> [[TMP15]], i8 [[TMP16]], i32 8 +; OPT-NEXT: [[TMP18:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP19:%.*]] = insertelement <128 x i8> [[TMP17]], i8 [[TMP18]], i32 9 +; OPT-NEXT: [[TMP20:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP21:%.*]] = insertelement <128 x i8> [[TMP19]], i8 [[TMP20]], i32 10 +; OPT-NEXT: [[TMP22:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP23:%.*]] = insertelement <128 x i8> [[TMP21]], i8 [[TMP22]], i32 11 +; OPT-NEXT: [[TMP24:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP25:%.*]] = insertelement <128 x i8> [[TMP23]], i8 [[TMP24]], i32 12 +; OPT-NEXT: [[TMP26:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP27:%.*]] = insertelement <128 x i8> [[TMP25]], i8 [[TMP26]], i32 13 +; OPT-NEXT: [[TMP28:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP29:%.*]] = insertelement <128 x i8> [[TMP27]], i8 [[TMP28]], i32 14 +; OPT-NEXT: [[TMP30:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP31:%.*]] = insertelement <128 x i8> [[TMP29]], i8 [[TMP30]], i32 15 +; OPT-NEXT: [[TMP32:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP33:%.*]] = insertelement <128 x i8> [[TMP31]], i8 [[TMP32]], i32 0 +; OPT-NEXT: [[TMP34:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP35:%.*]] = insertelement <128 x i8> [[TMP33]], i8 [[TMP34]], i32 1 +; OPT-NEXT: [[TMP36:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP37:%.*]] = insertelement <128 x i8> [[TMP35]], i8 [[TMP36]], i32 2 +; OPT-NEXT: [[TMP38:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP39:%.*]] = insertelement <128 x i8> [[TMP37]], i8 [[TMP38]], i32 3 +; OPT-NEXT: [[TMP40:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP41:%.*]] = insertelement <128 x i8> [[TMP39]], i8 [[TMP40]], i32 4 +; OPT-NEXT: [[TMP42:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP43:%.*]] = insertelement <128 x i8> [[TMP41]], i8 [[TMP42]], i32 5 +; OPT-NEXT: [[TMP44:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP45:%.*]] = insertelement <128 x i8> [[TMP43]], i8 [[TMP44]], i32 6 +; OPT-NEXT: [[TMP46:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP47:%.*]] = insertelement <128 x i8> [[TMP45]], i8 [[TMP46]], i32 7 +; OPT-NEXT: [[TMP48:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP49:%.*]] = insertelement <128 x i8> [[TMP47]], i8 [[TMP48]], i32 8 +; OPT-NEXT: [[TMP50:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP51:%.*]] = insertelement <128 x i8> [[TMP49]], i8 [[TMP50]], i32 9 +; OPT-NEXT: [[TMP52:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP53:%.*]] = insertelement <128 x i8> [[TMP51]], i8 [[TMP52]], i32 10 +; OPT-NEXT: [[TMP54:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP55:%.*]] = insertelement <128 x i8> [[TMP53]], i8 [[TMP54]], i32 11 +; OPT-NEXT: [[TMP56:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP57:%.*]] = insertelement <128 x i8> [[TMP55]], i8 [[TMP56]], i32 12 +; OPT-NEXT: [[TMP58:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP59:%.*]] = insertelement <128 x i8> [[TMP57]], i8 [[TMP58]], i32 13 +; OPT-NEXT: [[TMP60:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP61:%.*]] = insertelement <128 x i8> [[TMP59]], i8 [[TMP60]], i32 14 +; OPT-NEXT: [[TMP62:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP63:%.*]] = insertelement <128 x i8> [[TMP61]], i8 [[TMP62]], i32 15 +; OPT-NEXT: [[TMP64:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP65:%.*]] = insertelement <128 x i8> [[TMP63]], i8 [[TMP64]], i32 0 +; OPT-NEXT: [[TMP66:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP67:%.*]] = insertelement <128 x i8> [[TMP65]], i8 [[TMP66]], i32 1 +; OPT-NEXT: [[TMP68:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP69:%.*]] = insertelement <128 x i8> [[TMP67]], i8 [[TMP68]], i32 2 +; OPT-NEXT: [[TMP70:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP71:%.*]] = insertelement <128 x i8> [[TMP69]], i8 [[TMP70]], i32 3 +; OPT-NEXT: [[TMP72:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP73:%.*]] = insertelement <128 x i8> [[TMP71]], i8 [[TMP72]], i32 4 +; OPT-NEXT: [[TMP74:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP75:%.*]] = insertelement <128 x i8> [[TMP73]], i8 [[TMP74]], i32 5 +; OPT-NEXT: [[TMP76:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP77:%.*]] = insertelement <128 x i8> [[TMP75]], i8 [[TMP76]], i32 6 +; OPT-NEXT: [[TMP78:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP79:%.*]] = insertelement <128 x i8> [[TMP77]], i8 [[TMP78]], i32 7 +; OPT-NEXT: [[TMP80:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP81:%.*]] = insertelement <128 x i8> [[TMP79]], i8 [[TMP80]], i32 8 +; OPT-NEXT: [[TMP82:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP83:%.*]] = insertelement <128 x i8> [[TMP81]], i8 [[TMP82]], i32 9 +; OPT-NEXT: [[TMP84:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP85:%.*]] = insertelement <128 x i8> [[TMP83]], i8 [[TMP84]], i32 10 +; OPT-NEXT: [[TMP86:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP87:%.*]] = insertelement <128 x i8> [[TMP85]], i8 [[TMP86]], i32 11 +; OPT-NEXT: [[TMP88:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP89:%.*]] = insertelement <128 x i8> [[TMP87]], i8 [[TMP88]], i32 12 +; OPT-NEXT: [[TMP90:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP91:%.*]] = insertelement <128 x i8> [[TMP89]], i8 [[TMP90]], i32 13 +; OPT-NEXT: [[TMP92:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP93:%.*]] = insertelement <128 x i8> [[TMP91]], i8 [[TMP92]], i32 14 +; OPT-NEXT: [[TMP94:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP95:%.*]] = insertelement <128 x i8> [[TMP93]], i8 [[TMP94]], i32 15 +; OPT-NEXT: [[TMP96:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP97:%.*]] = insertelement <128 x i8> [[TMP95]], i8 [[TMP96]], i32 0 +; OPT-NEXT: [[TMP98:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP99:%.*]] = insertelement <128 x i8> [[TMP97]], i8 [[TMP98]], i32 1 +; OPT-NEXT: [[TMP100:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP101:%.*]] = insertelement <128 x i8> [[TMP99]], i8 [[TMP100]], i32 2 +; OPT-NEXT: [[TMP102:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP103:%.*]] = insertelement <128 x i8> [[TMP101]], i8 [[TMP102]], i32 3 +; OPT-NEXT: [[TMP104:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP105:%.*]] = insertelement <128 x i8> [[TMP103]], i8 [[TMP104]], i32 4 +; OPT-NEXT: [[TMP106:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP107:%.*]] = insertelement <128 x i8> [[TMP105]], i8 [[TMP106]], i32 5 +; OPT-NEXT: [[TMP108:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP109:%.*]] = insertelement <128 x i8> [[TMP107]], i8 [[TMP108]], i32 6 +; OPT-NEXT: [[TMP110:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP111:%.*]] = insertelement <128 x i8> [[TMP109]], i8 [[TMP110]], i32 7 +; OPT-NEXT: [[TMP112:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP113:%.*]] = insertelement <128 x i8> [[TMP111]], i8 [[TMP112]], i32 8 +; OPT-NEXT: [[TMP114:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP115:%.*]] = insertelement <128 x i8> [[TMP113]], i8 [[TMP114]], i32 9 +; OPT-NEXT: [[TMP116:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP117:%.*]] = insertelement <128 x i8> [[TMP115]], i8 [[TMP116]], i32 10 +; OPT-NEXT: [[TMP118:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP119:%.*]] = insertelement <128 x i8> [[TMP117]], i8 [[TMP118]], i32 11 +; OPT-NEXT: [[TMP120:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP121:%.*]] = insertelement <128 x i8> [[TMP119]], i8 [[TMP120]], i32 12 +; OPT-NEXT: [[TMP122:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP123:%.*]] = insertelement <128 x i8> [[TMP121]], i8 [[TMP122]], i32 13 +; OPT-NEXT: [[TMP124:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP125:%.*]] = insertelement <128 x i8> [[TMP123]], i8 [[TMP124]], i32 14 +; OPT-NEXT: [[TMP126:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP127:%.*]] = insertelement <128 x i8> [[TMP125]], i8 [[TMP126]], i32 15 +; OPT-NEXT: [[TMP128:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP129:%.*]] = insertelement <128 x i8> [[TMP127]], i8 [[TMP128]], i32 0 +; OPT-NEXT: [[TMP130:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP131:%.*]] = insertelement <128 x i8> [[TMP129]], i8 [[TMP130]], i32 1 +; OPT-NEXT: [[TMP132:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP133:%.*]] = insertelement <128 x i8> [[TMP131]], i8 [[TMP132]], i32 2 +; OPT-NEXT: [[TMP134:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP135:%.*]] = insertelement <128 x i8> [[TMP133]], i8 [[TMP134]], i32 3 +; OPT-NEXT: [[TMP136:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP137:%.*]] = insertelement <128 x i8> [[TMP135]], i8 [[TMP136]], i32 4 +; OPT-NEXT: [[TMP138:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP139:%.*]] = insertelement <128 x i8> [[TMP137]], i8 [[TMP138]], i32 5 +; OPT-NEXT: [[TMP140:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP141:%.*]] = insertelement <128 x i8> [[TMP139]], i8 [[TMP140]], i32 6 +; OPT-NEXT: [[TMP142:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP143:%.*]] = insertelement <128 x i8> [[TMP141]], i8 [[TMP142]], i32 7 +; OPT-NEXT: [[TMP144:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP145:%.*]] = insertelement <128 x i8> [[TMP143]], i8 [[TMP144]], i32 8 +; OPT-NEXT: [[TMP146:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP147:%.*]] = insertelement <128 x i8> [[TMP145]], i8 [[TMP146]], i32 9 +; OPT-NEXT: [[TMP148:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP149:%.*]] = insertelement <128 x i8> [[TMP147]], i8 [[TMP148]], i32 10 +; OPT-NEXT: [[TMP150:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP151:%.*]] = insertelement <128 x i8> [[TMP149]], i8 [[TMP150]], i32 11 +; OPT-NEXT: [[TMP152:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP153:%.*]] = insertelement <128 x i8> [[TMP151]], i8 [[TMP152]], i32 12 +; OPT-NEXT: [[TMP154:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP155:%.*]] = insertelement <128 x i8> [[TMP153]], i8 [[TMP154]], i32 13 +; OPT-NEXT: [[TMP156:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP157:%.*]] = insertelement <128 x i8> [[TMP155]], i8 [[TMP156]], i32 14 +; OPT-NEXT: [[TMP158:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP159:%.*]] = insertelement <128 x i8> [[TMP157]], i8 [[TMP158]], i32 15 +; OPT-NEXT: [[TMP160:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP161:%.*]] = insertelement <128 x i8> [[TMP159]], i8 [[TMP160]], i32 0 +; OPT-NEXT: [[TMP162:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP163:%.*]] = insertelement <128 x i8> [[TMP161]], i8 [[TMP162]], i32 1 +; OPT-NEXT: [[TMP164:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP165:%.*]] = insertelement <128 x i8> [[TMP163]], i8 [[TMP164]], i32 2 +; OPT-NEXT: [[TMP166:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP167:%.*]] = insertelement <128 x i8> [[TMP165]], i8 [[TMP166]], i32 3 +; OPT-NEXT: [[TMP168:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP169:%.*]] = insertelement <128 x i8> [[TMP167]], i8 [[TMP168]], i32 4 +; OPT-NEXT: [[TMP170:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP171:%.*]] = insertelement <128 x i8> [[TMP169]], i8 [[TMP170]], i32 5 +; OPT-NEXT: [[TMP172:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP173:%.*]] = insertelement <128 x i8> [[TMP171]], i8 [[TMP172]], i32 6 +; OPT-NEXT: [[TMP174:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP175:%.*]] = insertelement <128 x i8> [[TMP173]], i8 [[TMP174]], i32 7 +; OPT-NEXT: [[TMP176:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP177:%.*]] = insertelement <128 x i8> [[TMP175]], i8 [[TMP176]], i32 8 +; OPT-NEXT: [[TMP178:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP179:%.*]] = insertelement <128 x i8> [[TMP177]], i8 [[TMP178]], i32 9 +; OPT-NEXT: [[TMP180:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP181:%.*]] = insertelement <128 x i8> [[TMP179]], i8 [[TMP180]], i32 10 +; OPT-NEXT: [[TMP182:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP183:%.*]] = insertelement <128 x i8> [[TMP181]], i8 [[TMP182]], i32 11 +; OPT-NEXT: [[TMP184:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP185:%.*]] = insertelement <128 x i8> [[TMP183]], i8 [[TMP184]], i32 12 +; OPT-NEXT: [[TMP186:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP187:%.*]] = insertelement <128 x i8> [[TMP185]], i8 [[TMP186]], i32 13 +; OPT-NEXT: [[TMP188:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP189:%.*]] = insertelement <128 x i8> [[TMP187]], i8 [[TMP188]], i32 14 +; OPT-NEXT: [[TMP190:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP191:%.*]] = insertelement <128 x i8> [[TMP189]], i8 [[TMP190]], i32 15 +; OPT-NEXT: [[TMP192:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP193:%.*]] = insertelement <128 x i8> [[TMP191]], i8 [[TMP192]], i32 0 +; OPT-NEXT: [[TMP194:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP195:%.*]] = insertelement <128 x i8> [[TMP193]], i8 [[TMP194]], i32 1 +; OPT-NEXT: [[TMP196:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP197:%.*]] = insertelement <128 x i8> [[TMP195]], i8 [[TMP196]], i32 2 +; OPT-NEXT: [[TMP198:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP199:%.*]] = insertelement <128 x i8> [[TMP197]], i8 [[TMP198]], i32 3 +; OPT-NEXT: [[TMP200:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP201:%.*]] = insertelement <128 x i8> [[TMP199]], i8 [[TMP200]], i32 4 +; OPT-NEXT: [[TMP202:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP203:%.*]] = insertelement <128 x i8> [[TMP201]], i8 [[TMP202]], i32 5 +; OPT-NEXT: [[TMP204:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP205:%.*]] = insertelement <128 x i8> [[TMP203]], i8 [[TMP204]], i32 6 +; OPT-NEXT: [[TMP206:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP207:%.*]] = insertelement <128 x i8> [[TMP205]], i8 [[TMP206]], i32 7 +; OPT-NEXT: [[TMP208:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP209:%.*]] = insertelement <128 x i8> [[TMP207]], i8 [[TMP208]], i32 8 +; OPT-NEXT: [[TMP210:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP211:%.*]] = insertelement <128 x i8> [[TMP209]], i8 [[TMP210]], i32 9 +; OPT-NEXT: [[TMP212:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP213:%.*]] = insertelement <128 x i8> [[TMP211]], i8 [[TMP212]], i32 10 +; OPT-NEXT: [[TMP214:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP215:%.*]] = insertelement <128 x i8> [[TMP213]], i8 [[TMP214]], i32 11 +; OPT-NEXT: [[TMP216:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP217:%.*]] = insertelement <128 x i8> [[TMP215]], i8 [[TMP216]], i32 12 +; OPT-NEXT: [[TMP218:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP219:%.*]] = insertelement <128 x i8> [[TMP217]], i8 [[TMP218]], i32 13 +; OPT-NEXT: [[TMP220:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP221:%.*]] = insertelement <128 x i8> [[TMP219]], i8 [[TMP220]], i32 14 +; OPT-NEXT: [[TMP222:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP223:%.*]] = insertelement <128 x i8> [[TMP221]], i8 [[TMP222]], i32 15 +; OPT-NEXT: [[TMP224:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP225:%.*]] = insertelement <128 x i8> [[TMP223]], i8 [[TMP224]], i32 0 +; OPT-NEXT: [[TMP226:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP227:%.*]] = insertelement <128 x i8> [[TMP225]], i8 [[TMP226]], i32 1 +; OPT-NEXT: [[TMP228:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP229:%.*]] = insertelement <128 x i8> [[TMP227]], i8 [[TMP228]], i32 2 +; OPT-NEXT: [[TMP230:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP231:%.*]] = insertelement <128 x i8> [[TMP229]], i8 [[TMP230]], i32 3 +; OPT-NEXT: [[TMP232:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP233:%.*]] = insertelement <128 x i8> [[TMP231]], i8 [[TMP232]], i32 4 +; OPT-NEXT: [[TMP234:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP235:%.*]] = insertelement <128 x i8> [[TMP233]], i8 [[TMP234]], i32 5 +; OPT-NEXT: [[TMP236:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP237:%.*]] = insertelement <128 x i8> [[TMP235]], i8 [[TMP236]], i32 6 +; OPT-NEXT: [[TMP238:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP239:%.*]] = insertelement <128 x i8> [[TMP237]], i8 [[TMP238]], i32 7 +; OPT-NEXT: [[TMP240:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP241:%.*]] = insertelement <128 x i8> [[TMP239]], i8 [[TMP240]], i32 8 +; OPT-NEXT: [[TMP242:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP243:%.*]] = insertelement <128 x i8> [[TMP241]], i8 [[TMP242]], i32 9 +; OPT-NEXT: [[TMP244:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP245:%.*]] = insertelement <128 x i8> [[TMP243]], i8 [[TMP244]], i32 10 +; OPT-NEXT: [[TMP246:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP247:%.*]] = insertelement <128 x i8> [[TMP245]], i8 [[TMP246]], i32 11 +; OPT-NEXT: [[TMP248:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP249:%.*]] = insertelement <128 x i8> [[TMP247]], i8 [[TMP248]], i32 12 +; OPT-NEXT: [[TMP250:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP251:%.*]] = insertelement <128 x i8> [[TMP249]], i8 [[TMP250]], i32 13 +; OPT-NEXT: [[TMP252:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP253:%.*]] = insertelement <128 x i8> [[TMP251]], i8 [[TMP252]], i32 14 +; OPT-NEXT: [[TMP254:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP255:%.*]] = insertelement <128 x i8> [[TMP253]], i8 [[TMP254]], i32 15 +; OPT-NEXT: [[TMP256:%.*]] = extractelement <128 x i8> [[TMP255]], i32 80 +; OPT-NEXT: [[TMP257:%.*]] = insertelement <16 x i8> poison, i8 [[TMP256]], i64 0 +; OPT-NEXT: [[TMP258:%.*]] = extractelement <128 x i8> [[TMP255]], i32 81 +; OPT-NEXT: [[TMP259:%.*]] = insertelement <16 x i8> [[TMP257]], i8 [[TMP258]], i64 1 +; OPT-NEXT: [[TMP260:%.*]] = extractelement <128 x i8> [[TMP255]], i32 82 +; OPT-NEXT: [[TMP261:%.*]] = insertelement <16 x i8> [[TMP259]], i8 [[TMP260]], i64 2 +; OPT-NEXT: [[TMP262:%.*]] = extractelement <128 x i8> [[TMP255]], i32 83 +; OPT-NEXT: [[TMP263:%.*]] = insertelement <16 x i8> [[TMP261]], i8 [[TMP262]], i64 3 +; OPT-NEXT: [[TMP264:%.*]] = extractelement <128 x i8> [[TMP255]], i32 84 +; OPT-NEXT: [[TMP265:%.*]] = insertelement <16 x i8> [[TMP263]], i8 [[TMP264]], i64 4 +; OPT-NEXT: [[TMP266:%.*]] = extractelement <128 x i8> [[TMP255]], i32 85 +; OPT-NEXT: [[TMP267:%.*]] = insertelement <16 x i8> [[TMP265]], i8 [[TMP266]], i64 5 +; OPT-NEXT: [[TMP268:%.*]] = extractelement <128 x i8> [[TMP255]], i32 86 +; OPT-NEXT: [[TMP269:%.*]] = insertelement <16 x i8> [[TMP267]], i8 [[TMP268]], i64 6 +; OPT-NEXT: [[TMP270:%.*]] = extractelement <128 x i8> [[TMP255]], i32 87 +; OPT-NEXT: [[TMP271:%.*]] = insertelement <16 x i8> [[TMP269]], i8 [[TMP270]], i64 7 +; OPT-NEXT: [[TMP272:%.*]] = extractelement <128 x i8> [[TMP255]], i32 88 +; OPT-NEXT: [[TMP273:%.*]] = insertelement <16 x i8> [[TMP271]], i8 [[TMP272]], i64 8 +; OPT-NEXT: [[TMP274:%.*]] = extractelement <128 x i8> [[TMP255]], i32 89 +; OPT-NEXT: [[TMP275:%.*]] = insertelement <16 x i8> [[TMP273]], i8 [[TMP274]], i64 9 +; OPT-NEXT: [[TMP276:%.*]] = extractelement <128 x i8> [[TMP255]], i32 90 +; OPT-NEXT: [[TMP277:%.*]] = insertelement <16 x i8> [[TMP275]], i8 [[TMP276]], i64 10 +; OPT-NEXT: [[TMP278:%.*]] = extractelement <128 x i8> [[TMP255]], i32 91 +; OPT-NEXT: [[TMP279:%.*]] = insertelement <16 x i8> [[TMP277]], i8 [[TMP278]], i64 11 +; OPT-NEXT: [[TMP280:%.*]] = extractelement <128 x i8> [[TMP255]], i32 92 +; OPT-NEXT: [[TMP281:%.*]] = insertelement <16 x i8> [[TMP279]], i8 [[TMP280]], i64 12 +; OPT-NEXT: [[TMP282:%.*]] = extractelement <128 x i8> [[TMP255]], i32 93 +; OPT-NEXT: [[TMP283:%.*]] = insertelement <16 x i8> [[TMP281]], i8 [[TMP282]], i64 13 +; OPT-NEXT: [[TMP284:%.*]] = extractelement <128 x i8> [[TMP255]], i32 94 +; OPT-NEXT: [[TMP285:%.*]] = insertelement <16 x i8> [[TMP283]], i8 [[TMP284]], i64 14 +; OPT-NEXT: [[TMP286:%.*]] = extractelement <128 x i8> [[TMP255]], i32 95 +; OPT-NEXT: [[TMP287:%.*]] = insertelement <16 x i8> [[TMP285]], i8 [[TMP286]], i64 15 +; OPT-NEXT: [[SUM:%.*]] = add <16 x i8> [[TMP287]], [[ADD]] +; OPT-NEXT: store <16 x i8> [[SUM]], ptr addrspace(3) [[OUT]], align 16 +; OPT-NEXT: ret void +; +entry: + %alloca = alloca [8 x <16 x i8>], align 16, addrspace(5) + %gep0 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 0 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep1 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 1 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep2 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 2 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep3 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 3 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep4 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 4 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep5 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 5 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep6 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 6 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep7 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 7 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %load = load <16 x i8>, ptr addrspace(5) %gep5, align 16 + %sum = add <16 x i8> %load, %add + store <16 x i8> %sum, ptr addrspace(3) %out, align 16 + ret void +} + +attributes #0 = {"amdgpu-waves-per-eu"="2,2"} diff --git a/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll b/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll index f67cbe3..ddb522a8 100644 --- a/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll +++ b/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll @@ -1,17 +1,17 @@ ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s ; -global-isel=1 SI run line skipped since store not yet implemented. ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-SDAG -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-GISEL -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-GISEL -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s declare i64 @llvm.readcyclecounter() #0 diff --git a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll index c573253..48ed5c4 100644 --- a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll +++ b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll @@ -73,10 +73,10 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) % } ; CHECK-LABEL: {{^}}excess_soft_clause_reg_pressure: -; GFX908: NumSgprs: 64 -; GFX908-GCNTRACKERS: NumSgprs: 64 +; GFX908: NumSgprs: 56 +; GFX908-GCNTRACKERS: NumSgprs: 56 ; GFX908: NumVgprs: 43 -; GFX908-GCNTRACKERS: NumVgprs: 39 +; GFX908-GCNTRACKERS: NumVgprs: 40 ; GFX908: Occupancy: 5 ; GFX908-GCNTRACKERS: Occupancy: 6 diff --git a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll index 586579f..ef96944 100644 --- a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll +++ b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll @@ -20,38 +20,33 @@ define void @test() { ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 ; CHECK-NEXT: .LBB0_3: ; %bb.3 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: ; implicit-def: $sgpr4 -; CHECK-NEXT: v_mov_b32_e32 v0, s4 -; CHECK-NEXT: v_readfirstlane_b32 s6, v0 ; CHECK-NEXT: s_mov_b64 s[4:5], -1 -; CHECK-NEXT: s_mov_b32 s7, 0 -; CHECK-NEXT: s_cmp_eq_u32 s6, s7 ; CHECK-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_mov_b64 s[10:11], exec -; CHECK-NEXT: s_mov_b64 exec, -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 +; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: s_cbranch_scc1 .LBB0_5 ; CHECK-NEXT: ; %bb.4: ; %bb.4 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: s_mov_b64 s[4:5], 0 ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: .LBB0_5: ; %Flow ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: v_readlane_b32 s4, v1, 0 ; CHECK-NEXT: v_readlane_b32 s5, v1, 1 ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] diff --git a/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir b/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir index 0c694d9..6989583 100644 --- a/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir +++ b/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir @@ -1,5 +1,6 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=prologepilog -o - %s | FileCheck -check-prefix=EXPANDED %s +# RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1250 -mattr=+real-true16 -run-pass=prologepilog -o - %s | FileCheck -check-prefix=SRAMECC-EXPANDED %s --- name: spill_restore_vgpr16 @@ -31,6 +32,28 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16 + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 @@ -78,6 +101,29 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16_middle_of_block + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 @@ -124,6 +170,27 @@ body: | ; EXPANDED-NEXT: bb.2: ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16_end_of_block + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 diff --git a/llvm/test/CodeGen/AMDGPU/spillv16.ll b/llvm/test/CodeGen/AMDGPU/spillv16.ll index 0e45df2..2d54ac8 100644 --- a/llvm/test/CodeGen/AMDGPU/spillv16.ll +++ b/llvm/test/CodeGen/AMDGPU/spillv16.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GCN,GCN-TRUE16 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GCN,GCN-FAKE16 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 -mattr=+real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GFX1250,GFX1250-TRUE16 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 -mattr=-real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GFX1250,GFX1250-FAKE16 define void @spill_i16_alu() { ; GCN-TRUE16-LABEL: spill_i16_alu: @@ -32,6 +34,41 @@ define void @spill_i16_alu() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_i16_alu: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 ; 2-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_u16 v1, off, s32 offset:2 th:TH_LOAD_LU ; 2-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_i16_alu: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) @@ -88,6 +125,51 @@ define void @spill_i16_alu_two_vals() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_i16_alu_two_vals: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:6 ; 2-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v1, off, s32 offset:6 th:TH_LOAD_LU ; 2-byte Folded Reload +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l +; GFX1250-TRUE16-NEXT: scratch_store_d16_hi_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_i16_alu_two_vals: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v1, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v1, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) %alloca2 = alloca i16, i32 1, align 4, addrspace(5) @@ -140,6 +222,22 @@ define void @spill_i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) @@ -183,6 +281,22 @@ define void @spill_half() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_half: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca half, i32 1, align 4, addrspace(5) @@ -226,6 +340,22 @@ define void @spill_i16_from_v2i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_i16_from_v2i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -283,6 +413,54 @@ define void @spill_2xi16_from_v2i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_2xi16_from_v2i16: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: s_clause 0x1 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:12 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:12 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_2xi16_from_v2i16: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: s_clause 0x1 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:12 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:12 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -341,6 +519,47 @@ define void @spill_2xi16_from_v2i16_one_free_reg() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_2xi16_from_v2i16_one_free_reg: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v7, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.l, v7.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_2xi16_from_v2i16_one_free_reg: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v7, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v7, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -375,6 +594,22 @@ define void @spill_v2i16() { ; GCN-NEXT: scratch_store_b32 off, v0, s32 offset:4 dlc ; GCN-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_v2i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) diff --git a/llvm/test/CodeGen/AMDGPU/spillv16.mir b/llvm/test/CodeGen/AMDGPU/spillv16.mir index 05569bf..ba2d926 100644 --- a/llvm/test/CodeGen/AMDGPU/spillv16.mir +++ b/llvm/test/CodeGen/AMDGPU/spillv16.mir @@ -1,6 +1,7 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=regallocfast -o - %s | FileCheck -check-prefix=SPILLED %s # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=regallocfast,prologepilog -o - %s | FileCheck -check-prefix=EXPANDED %s +# RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1250 -mattr=+real-true16 -run-pass=regallocfast,prologepilog -o - %s | FileCheck -check-prefix=SRAMECC-EXPANDED %s --- name: spill_restore_vgpr16 @@ -46,6 +47,27 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16 + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: S_NOP 0, implicit-def %0:vgpr_16, implicit-def %1:vgpr_16 S_CBRANCH_SCC1 implicit undef $scc, %bb.1 diff --git a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll index 5aafb0f..364598f 100644 --- a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll +++ b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll @@ -31,8 +31,8 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[COPY13:%[0-9]+]]:sgpr_32 = COPY $sgpr10 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:sgpr_32 = COPY $sgpr8 ; CHECK-NEXT: undef [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM [[COPY]], 232, 0 :: (invariant load (s64) from %ir.39, addrspace 4) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %125:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: KILL undef %125:sgpr_128 + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %117:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: KILL undef %117:sgpr_128 ; CHECK-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_1:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_2:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 4, implicit-def dead $scc @@ -44,87 +44,85 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_SUB_I32_1:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 30, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.81, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.71, addrspace 4) ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM1:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM undef %74:sreg_64, 0, 0 :: (invariant load (s128) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 64, 0 :: (invariant load (s128) from %ir.88, addrspace 4) ; CHECK-NEXT: KILL undef %74:sreg_64 ; CHECK-NEXT: KILL [[S_ADD_U32_]].sub0, [[S_ADD_U32_]].sub1 ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_LOAD_DWORDX4_IMM]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec ; CHECK-NEXT: undef [[S_MOV_B32_:%[0-9]+]].sub1:sgpr_128 = S_MOV_B32 0 - ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %118:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %89:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %112:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %87:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL undef %89:sgpr_128 - ; CHECK-NEXT: KILL undef %118:sgpr_128 + ; CHECK-NEXT: KILL undef %112:sgpr_128 + ; CHECK-NEXT: KILL undef %87:sgpr_128 ; CHECK-NEXT: [[S_SUB_I32_2:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM1]], 31, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_1:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_1:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: undef [[S_ADD_U32_2:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_1]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_2:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.87, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.93, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1 + ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %148:sreg_32, 31, implicit-def dead $scc + ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %148:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.77, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.83, addrspace 4) ; CHECK-NEXT: KILL [[S_ADD_U32_2]].sub0, [[S_ADD_U32_2]].sub1 - ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %169:sreg_32, 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %169:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %169:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1 + ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %148:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_1:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_2]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %302:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %279:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_MOV_B32_]], 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %357:sgpr_128, undef %358:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %368:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.99, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 64, 0 :: (invariant load (s128) from %ir.107, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.117, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 0, 0 :: (invariant load (s128) from %ir.124, addrspace 4) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %352:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %363:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %334:sgpr_128, undef %335:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %345:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.95, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 0, 0 :: (invariant load (s128) from %ir.100, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.105, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %329:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %340:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_2:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM]], -98, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_3:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM1]], -114, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_4:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM2]], -130, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_5:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM2]], -178, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_3:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY12]], 4, implicit-def dead $scc - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_6:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_3]], 16, implicit-def dead $scc - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %384:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %361:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN5:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.129, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.145, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 224, 0 :: (invariant load (s128) from %ir.117, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.133, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 576, 0 :: (invariant load (s128) from %ir.138, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN6:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 224, 0 :: (invariant load (s128) from %ir.134, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.162, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 224, 0 :: (invariant load (s128) from %ir.140, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.122, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 224, 0 :: (invariant load (s128) from %ir.128, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN7:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN8:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_7:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM4]], -217, implicit-def dead $scc @@ -135,49 +133,49 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_12:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -329, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_13:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -345, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_14:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM6]], -441, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_4:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY13]], 4, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN9:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_4:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_4]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_5:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN10:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM12]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_5:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_5]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s32) from %ir.273, align 8, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 576, 0 :: (invariant load (s128) from %ir.157, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_17]], 168, 0 :: (invariant load (s32) from %ir.260, align 8, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.145, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN11:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM14]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN12:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM10]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN13:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM11]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub3:sgpr_128 = S_MOV_B32 553734060 ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub2:sgpr_128 = S_MOV_B32 -1 ; CHECK-NEXT: [[COPY15:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.170, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 0, 0 :: (invariant load (s128) from %ir.158, addrspace 4) ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_]].sub1 ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORD_IMM]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY15]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN14:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM15]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN15:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM13]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.178, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.183, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.166, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.171, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN16:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM16]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LSHL_B32_6:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_6:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_6]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_15:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM4]], -467, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.282, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s64) from %ir.269, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM17]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM18]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.205, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.211, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.193, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.199, addrspace 4) ; CHECK-NEXT: [[COPY16:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.216, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 0, 0 :: (invariant load (s128) from %ir.221, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.204, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.209, addrspace 4) ; CHECK-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM1]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM1]].sub0 ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_]] @@ -189,30 +187,30 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN20:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM22]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_7:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_7]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_16:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM5]], -468, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s64) from %ir.293, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.280, addrspace 4) ; CHECK-NEXT: [[COPY17:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM2]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM2]].sub0 ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_1]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY17]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.256, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %470:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_16]].sub0, [[S_ADD_U32_16]].sub1 - ; CHECK-NEXT: KILL undef %470:sreg_64 + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 160, 0 :: (invariant load (s128) from %ir.244, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %443:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: KILL [[S_ADD_U32_15]].sub0, [[S_ADD_U32_15]].sub1 ; CHECK-NEXT: KILL [[COPY17]].sub0_sub1_sub2, [[COPY17]].sub3 + ; CHECK-NEXT: KILL undef %443:sreg_64 ; CHECK-NEXT: [[S_LSHL_B32_8:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY14]], 3, implicit-def dead $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_17]], 160, 0 :: (invariant load (s128) from %ir.265, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.252, addrspace 4) ; CHECK-NEXT: [[S_ASHR_I32_8:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_8]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_17:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM6]], -469, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_21]], 168, 0 :: (invariant load (s32) from %ir.305, align 8, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s32) from %ir.291, align 8, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN21:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM23]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN22:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM24]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM24]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM23]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM24]] ; CHECK-NEXT: [[S_AND_B32_2:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORD_IMM1]], 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY18:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[COPY18:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_2]] @@ -224,22 +222,22 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_21:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -507, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_22:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -539, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_23:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM7]], -473, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.323, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.329, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_24:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_24:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_24]], 96, 0 :: (invariant load (s128) from %ir.335, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_21]], 96, 0 :: (invariant load (s128) from %ir.309, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.315, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.321, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN23:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM25]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN24:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM26]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN25:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM27]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM25]] - ; CHECK-NEXT: KILL [[V_MOV_B32_e32_]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM26]] + ; CHECK-NEXT: KILL [[V_MOV_B32_e32_]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] ; CHECK-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -2, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -1, [[BUFFER_LOAD_FORMAT_X_IDXEN1]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -3, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec @@ -351,13 +349,13 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[V_OR_B32_e64_64:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_63]], [[V_ADD_U32_e64_28]], implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -593, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_OR_B32_e64_65:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_64]], [[V_ADD_U32_e64_29]], implicit $exec - ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %543:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %516:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) ; CHECK-NEXT: [[V_OR_B32_e64_66:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_65]], [[V_ADD_U32_e64_30]], implicit $exec ; CHECK-NEXT: [[S_ADD_I32_24:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM8]], -594, implicit-def dead $scc ; CHECK-NEXT: [[V_OR_B32_e64_67:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[S_ADD_I32_24]], [[V_OR_B32_e64_66]], implicit $exec ; CHECK-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 0, [[V_OR_B32_e64_67]], implicit $exec ; CHECK-NEXT: undef [[V_CNDMASK_B32_e64_:%[0-9]+]].sub3:vreg_128 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec - ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %557:vgpr_32, undef %559:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) + ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %530:vgpr_32, undef %532:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) ; CHECK-NEXT: S_ENDPGM 0 .expVert: %0 = extractelement <31 x i32> %userData, i64 2 diff --git a/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll new file mode 100644 index 0000000..22e4a24 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll @@ -0,0 +1,236 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 < %s | FileCheck -check-prefix=GFX11 %s + +define i16 @v_underflow_compare_fold_i16(i16 %a, i16 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u16_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u16_e32 v0, v1, v0 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u16 v0.h, v0.l, v1.l +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u16 v0.l, v0.h, v0.l +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i16 %a, %b + %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) + ret i16 %cond +} + +define i32 @v_underflow_compare_fold_i32(i32 %a, i32 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define i32 @v_underflow_compare_fold_i32_commute(i32 %a, i32 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32_commute: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32_commute: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %a, i32 %sub) + ret i32 %cond +} + +define i32 @v_underflow_compare_fold_i32_multi_use(i32 %a, i32 %b, ptr addrspace(1) %ptr) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32_multi_use: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX9-NEXT: global_store_dword v[2:3], v1, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32_multi_use: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX11-NEXT: global_store_b32 v[2:3], v1, off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + store i32 %sub, ptr addrspace(1) %ptr + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define i64 @v_underflow_compare_fold_i64(i64 %a, i64 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +define i64 @v_underflow_compare_fold_i64_commute(i64 %a, i64 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64_commute: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[0:1], v[2:3] +; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64_commute: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[2:3] +; GFX11-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v3, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %a, i64 %sub) + ret i64 %cond +} + +define i64 @v_underflow_compare_fold_i64_multi_use(i64 %a, i64 %b, ptr addrspace(1) %ptr) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64_multi_use: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[2:3], off +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64_multi_use: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX11-NEXT: global_store_b64 v[4:5], v[2:3], off +; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + store i64 %sub, ptr addrspace(1) %ptr + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +define amdgpu_ps i16 @s_underflow_compare_fold_i16(i16 inreg %a, i16 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_i32 s1, s0, s1 +; GFX9-NEXT: s_and_b32 s0, 0xffff, s0 +; GFX9-NEXT: s_and_b32 s1, s1, 0xffff +; GFX9-NEXT: s_min_u32 s0, s1, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_i32 s1, s0, s1 +; GFX11-NEXT: s_and_b32 s0, 0xffff, s0 +; GFX11-NEXT: s_and_b32 s1, s1, 0xffff +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: s_min_u32 s0, s1, s0 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i16 %a, %b + %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) + ret i16 %cond +} + +define amdgpu_ps i32 @s_underflow_compare_fold_i32(i32 inreg %a, i32 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_i32 s1, s0, s1 +; GFX9-NEXT: s_min_u32 s0, s1, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_i32 s1, s0, s1 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: s_min_u32 s0, s1, s0 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define amdgpu_ps i64 @s_underflow_compare_fold_i64(i64 inreg %a, i64 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_u32 s2, s0, s2 +; GFX9-NEXT: v_mov_b32_e32 v0, s0 +; GFX9-NEXT: s_subb_u32 s3, s1, s3 +; GFX9-NEXT: v_mov_b32_e32 v1, s1 +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1] +; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec +; GFX9-NEXT: s_cselect_b32 s1, s3, s1 +; GFX9-NEXT: s_cselect_b32 s0, s2, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_u32 s2, s0, s2 +; GFX11-NEXT: s_subb_u32 s3, s1, s3 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[2:3], s[0:1] +; GFX11-NEXT: s_and_b32 s4, s4, exec_lo +; GFX11-NEXT: s_cselect_b32 s0, s2, s0 +; GFX11-NEXT: s_cselect_b32 s1, s3, s1 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +attributes #0 = { nounwind } diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir index 8a70a8a..32cc398 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir +++ b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir @@ -36,7 +36,7 @@ body: | ; GCN-NEXT: v_add_f16_e64 v128.l /*v384.l*/, v129.l /*v385.l*/, v130.l /*v386.l*/ $vgpr384_lo16 = V_ADD_F16_t16_e64 0, undef $vgpr385_lo16, 0, undef $vgpr386_lo16, 0, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x8a + ; GCN-NEXT: s_set_vgpr_msb 0x458a ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=0 ; GCN-NEXT: v_add_f16_e64 v0.h /*v512.h*/, v1.h /*v513.h*/, v2.h /*v514.h*/ $vgpr512_hi16 = V_ADD_F16_t16_e64 0, undef $vgpr513_hi16, 0, undef $vgpr514_hi16, 0, 0, 0, implicit $exec, implicit $mode @@ -50,7 +50,7 @@ body: | ; GCN-NEXT: v_add_f16_e64 v128.l /*v640.l*/, v129.l /*v641.l*/, v130.l /*v642.l*/ $vgpr640_lo16 = V_ADD_F16_t16_e64 0, undef $vgpr641_lo16, 0, undef $vgpr642_lo16, 0, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xcf + ; GCN-NEXT: s_set_vgpr_msb 0x8acf ; ASM-SAME: ; msbs: dst=3 src0=3 src1=3 src2=0 ; GCN-NEXT: v_add_f16_e64 v0.h /*v768.h*/, v1.h /*v769.h*/, v2.h /*v770.h*/ $vgpr768_hi16 = V_ADD_F16_t16_e64 0, undef $vgpr769_hi16, 0, undef $vgpr770_hi16, 0, 0, 0, implicit $exec, implicit $mode diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir index f508df2..7e1c28f 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir +++ b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir @@ -22,13 +22,13 @@ body: | $vgpr257 = V_MOV_B32_e32 undef $vgpr510, implicit $exec ; Single bit change - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4101 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=0 ; GCN-NEXT: v_rcp_f32_e64 v255, v2 /*v258*/ $vgpr255 = V_RCP_F32_e64 0, undef $vgpr258, 0, 0, implicit $exec, implicit $mode ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: v_rcp_f32_e64 v255, v1 $vgpr255 = V_RCP_F32_e64 0, undef $vgpr1, 0, 0, implicit $exec, implicit $mode @@ -40,7 +40,7 @@ body: | ; GCN-NEXT: v_add_nc_u32_e32 v0, v253 /*v509*/, v252 /*v508*/ $vgpr0 = V_ADD_U32_e32 undef $vgpr509, undef $vgpr508, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GCN-NEXT: v_add_f32_e64 v2 /*v258*/, v0, v251 /*v507*/ @@ -48,7 +48,7 @@ body: | ; VOP3 - ; GCN-NEXT: s_set_vgpr_msb 0x55 + ; GCN-NEXT: s_set_vgpr_msb 0x4455 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=1 ; GCN-NEXT: v_fma_f32 v3 /*v259*/, v4 /*v260*/, v5 /*v261*/, v6 /*v262*/ $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode @@ -58,32 +58,32 @@ body: | $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode ; Tuple crossing the 256 boundary - ; GCN-NEXT: s_set_vgpr_msb 17 + ; GCN-NEXT: s_set_vgpr_msb 0x5511 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=1 ; GCN-NEXT: v_mqsad_u32_u8 v[254:257], v[2:3] /*v[258:259]*/, v0, v[244:247] /*v[500:503]*/ $vgpr254_vgpr255_vgpr256_vgpr257 = V_MQSAD_U32_U8_e64 $vgpr258_vgpr259, $vgpr0, undef $vgpr500_vgpr501_vgpr502_vgpr503, 0, implicit $exec ; DPP/tied operand - ; GCN-NEXT: s_set_vgpr_msb 0x45 + ; GCN-NEXT: s_set_vgpr_msb 0x1145 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=0 ; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GCN-NEXT: v_add_nc_u16_e64_dpp v0 /*v256*/, v1 /*v257*/, v2 /*v258*/ quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1 $vgpr256 = V_ADD_NC_U16_fake16_e64_dpp $vgpr256, 0, $vgpr257, 0, undef $vgpr258, 0, 0, 1, 15, 15, 1, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 17 + ; GCN-NEXT: s_set_vgpr_msb 0x4511 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=1 ; GCN-NEXT: v_add3_u32_e64_dpp v0, v1 /*v257*/, v0, v2 /*v258*/ quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1 $vgpr0 = V_ADD3_U32_e64_dpp $vgpr0, $vgpr257, $vgpr0, undef $vgpr258, 1, 15, 15, 1, implicit $exec ; DS (addr, data0, and data1 operands) - ; GCN-NEXT: s_set_vgpr_msb 20 + ; GCN-NEXT: s_set_vgpr_msb 0x1114 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=1 src2=1 ; GCN-NEXT: ds_store_2addr_b32 v0, v248 /*v504*/, v249 /*v505*/ offset1:1 DS_WRITE2_B32_gfx9 $vgpr0, undef $vgpr504, undef $vgpr505, 0, 1, 0, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x1400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: ds_store_2addr_b32 v0, v248, v249 offset1:1 DS_WRITE2_B32_gfx9 $vgpr0, undef $vgpr248, undef $vgpr249, 0, 1, 0, implicit $exec @@ -93,13 +93,13 @@ body: | ; GCN-NEXT: ds_load_b32 v0, v255 /*v511*/ $vgpr0 = DS_READ_B32_gfx9 $vgpr511, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x144 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: ds_add_rtn_u32 v255 /*v511*/, v0, v248 /*v504*/ $vgpr511 = DS_ADD_RTN_U32_gfx9 $vgpr0, undef $vgpr504, 0, 0, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: ds_add_rtn_u32 v0, v0, v0 $vgpr0 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr0, 0, 0, implicit $exec @@ -111,17 +111,17 @@ body: | ; GCN-NEXT: global_load_b32 v2, v[2:3] /*v[258:259]*/, off $vgpr2 = GLOBAL_LOAD_DWORD undef $vgpr258_vgpr259, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: s_set_vgpr_msb 0x140 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=0 src2=0 ; GCN-NEXT: global_load_b32 v255 /*v511*/, v0, s[0:1] $vgpr511 = GLOBAL_LOAD_DWORD_SADDR undef $sgpr0_sgpr1, $vgpr0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4001 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=0 ; GCN-NEXT: scratch_load_u8 v0, v255 /*v511*/, s0 $vgpr0 = SCRATCH_LOAD_UBYTE_SVS $vgpr511, undef $sgpr0, 0, 0, implicit $exec, implicit $flat_scr - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: global_store_b32 v[0:1], v2, off GLOBAL_STORE_DWORD $vgpr0_vgpr1, $vgpr2, 0, 0, implicit $exec @@ -135,13 +135,13 @@ body: | ; GCN-NEXT: global_store_b96 v[0:1] /*v[256:257]*/, v[244:246] /*v[500:502]*/, off GLOBAL_STORE_DWORDX3 $vgpr256_vgpr257, $vgpr500_vgpr501_vgpr502, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: flat_atomic_add_u32 v254 /*v510*/, v[0:1], v255 /*v511*/ th:TH_ATOMIC_RETURN $vgpr510 = FLAT_ATOMIC_ADD_RTN $vgpr0_vgpr1, $vgpr511, 0, 1, implicit $exec, implicit $flat_scr ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: flat_atomic_add_u32 v0, v[0:1], v255 th:TH_ATOMIC_RETURN $vgpr0 = FLAT_ATOMIC_ADD_RTN $vgpr0_vgpr1, $vgpr255, 0, 1, implicit $exec, implicit $flat_scr @@ -156,12 +156,12 @@ body: | ; GCN-NEXT: buffer_load_b32 v1 /*v257*/, v0, s[8:11], s3 offen $vgpr257 = BUFFER_LOAD_DWORD_VBUFFER_OFFEN $vgpr0, undef $sgpr8_sgpr9_sgpr10_sgpr11, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x4041 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=0 src2=0 ; GCN-NEXT: buffer_load_b32 v1 /*v257*/, v0 /*v256*/, s[8:11], s3 offen $vgpr257 = BUFFER_LOAD_DWORD_VBUFFER_OFFEN $vgpr256, undef $sgpr8_sgpr9_sgpr10_sgpr11, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: buffer_store_b32 v0, v1, s[0:3], s3 offen BUFFER_STORE_DWORD_VBUFFER_OFFEN $vgpr0, $vgpr1, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, 0, implicit $exec @@ -171,7 +171,7 @@ body: | ; GCN-NEXT: buffer_store_b32 v0 /*v256*/, v1 /*v257*/, s[0:3], s3 offen BUFFER_STORE_DWORD_VBUFFER_OFFEN $vgpr256, $vgpr257, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: buffer_atomic_add_f32 v0, v1, s[0:3], s3 offen BUFFER_ATOMIC_ADD_F32_VBUFFER_OFFEN $vgpr0, $vgpr1, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, implicit $exec @@ -183,44 +183,44 @@ body: | ; VGPRs above 512 - ; GCN-NEXT: s_set_vgpr_msb 0xaa + ; GCN-NEXT: s_set_vgpr_msb 0x41aa ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v514*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xab + ; GCN-NEXT: s_set_vgpr_msb 0xaaab ; ASM-SAME: ; msbs: dst=2 src0=3 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v0 /*v768*/, v2 /*v514*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr768, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xae + ; GCN-NEXT: s_set_vgpr_msb 0xabae ; ASM-SAME: ; msbs: dst=2 src0=2 src1=3 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v770*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr770, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xba + ; GCN-NEXT: s_set_vgpr_msb 0xaeba ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=3 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v514*/, v3 /*v771*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr771, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xea + ; GCN-NEXT: s_set_vgpr_msb 0xbaea ; ASM-SAME: ; msbs: dst=3 src0=2 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v255 /*v1023*/, v1 /*v513*/, v2 /*v514*/, v3 /*v515*/ $vgpr1023 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xff + ; GCN-NEXT: s_set_vgpr_msb 0xeaff ; ASM-SAME: ; msbs: dst=3 src0=3 src1=3 src2=3 ; GCN-NEXT: v_fma_f32 v0 /*v768*/, v1 /*v769*/, v2 /*v770*/, v3 /*v771*/ $vgpr768 = V_FMA_F32_e64 0, undef $vgpr769, 0, undef $vgpr770, 0, undef $vgpr771, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x42 + ; GCN-NEXT: s_set_vgpr_msb 0xff42 ; ASM-SAME: ; msbs: dst=1 src0=2 src1=0 src2=0 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v0 /*v512*/ $vgpr256 = V_MOV_B32_e32 undef $vgpr512, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4200 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: v_fma_f32 v0, v1, v2, v3 $vgpr0 = V_FMA_F32_e64 0, undef $vgpr1, 0, undef $vgpr2, 0, undef $vgpr3, 0, 0, implicit $exec, implicit $mode @@ -232,12 +232,12 @@ body: | ; GCN-NEXT: global_store_b96 v[0:1] /*v[512:513]*/, v[0:2] /*v[512:514]*/, off GLOBAL_STORE_DWORDX3 $vgpr512_vgpr513, $vgpr512_vgpr513_vgpr514, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 11 + ; GCN-NEXT: s_set_vgpr_msb 0xa0b ; ASM-SAME: ; msbs: dst=0 src0=3 src1=2 src2=0 ; GCN-NEXT: global_store_b64 v[254:255] /*v[1022:1023]*/, v[254:255] /*v[766:767]*/, off GLOBAL_STORE_DWORDX2 $vgpr1022_vgpr1023, $vgpr766_vgpr767, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x55 + ; GCN-NEXT: s_set_vgpr_msb 0xb55 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=1 ; GCN-NEXT: v_wmma_f32_16x16x32_bf16 v[14:21] /*v[270:277]*/, v[26:33] /*v[282:289]*/, v[34:41] /*v[290:297]*/, v[14:21] /*v[270:277]*/ early-clobber $vgpr270_vgpr271_vgpr272_vgpr273_vgpr274_vgpr275_vgpr276_vgpr277 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, undef $vgpr282_vgpr283_vgpr284_vgpr285_vgpr286_vgpr287_vgpr288_vgpr289, 8, undef $vgpr290_vgpr291_vgpr292_vgpr293_vgpr294_vgpr295_vgpr296_vgpr297, 8, killed undef $vgpr270_vgpr271_vgpr272_vgpr273_vgpr274_vgpr275_vgpr276_vgpr277, 0, 0, 0, 0, implicit $exec @@ -247,6 +247,7 @@ body: | ... # ASM-LABEL: {{^}}vopd: + # DIS-LABEL: <vopd>: --- name: vopd @@ -262,35 +263,35 @@ body: | ; GCN-NEXT: v_dual_sub_f32 v244 /*v500*/, v1, v2 :: v_dual_mul_f32 v0 /*v256*/, v3, v4 $vgpr500, $vgpr256 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $vgpr1, undef $vgpr2, undef $vgpr3, undef $vgpr4, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x4041 ; GCN-NEXT: v_dual_sub_f32 v244 /*v500*/, s1, v2 :: v_dual_mul_f32 v0 /*v256*/, v44 /*v300*/, v4 $vgpr500, $vgpr256 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $sgpr1, undef $vgpr2, undef $vgpr300, undef $vgpr4, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x4104 ; GCN-NEXT: v_dual_sub_f32 v255, v1, v44 /*v300*/ :: v_dual_mul_f32 v6, v0, v1 /*v257*/ $vgpr255, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $vgpr1, undef $vgpr300, undef $vgpr0, $vgpr257, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_dual_sub_f32 v255, 0, v1 :: v_dual_mul_f32 v6, v44 /*v300*/, v3 $vgpr255, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 0, undef $vgpr1, undef $vgpr300, undef $vgpr3, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: s_set_vgpr_msb 0x140 ; GCN-NEXT: v_dual_fmamk_f32 v243 /*v499*/, v0, 0xa, v3 :: v_dual_fmac_f32 v0 /*v256*/, v1, v1 $vgpr499, $vgpr256 = V_DUAL_FMAMK_F32_X_FMAC_F32_e32_gfx1250 undef $vgpr0, 10, undef $vgpr3, undef $vgpr1, undef $vgpr1, $vgpr256, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4005 ; GCN-NEXT: v_dual_mov_b32 v2, v3 /*v259*/ :: v_dual_add_f32 v3, v1 /*v257*/, v2 /*v258*/ $vgpr2, $vgpr3 = V_DUAL_MOV_B32_e32_X_ADD_F32_e32_gfx1250 undef $vgpr259, undef $vgpr257, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; GCN-NEXT: v_dual_fmamk_f32 v244 /*v500*/, v0, 0xa, v44 /*v300*/ :: v_dual_fmac_f32 v3 /*v259*/, v1, v1 /*v257*/ $vgpr500, $vgpr259 = V_DUAL_FMAMK_F32_X_FMAC_F32_e32_gfx1250 undef $vgpr0, 10, undef $vgpr300, undef $vgpr1, undef $vgpr257, $vgpr259, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 16 + ; GCN-NEXT: s_set_vgpr_msb 0x4410 ; GCN-NEXT: v_dual_fma_f32 v0, v6, v6, v44 /*v300*/ :: v_dual_fma_f32 v1, v4, v5, v45 /*v301*/ $vgpr0, $vgpr1 = V_DUAL_FMA_F32_e64_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr6, 0, undef $vgpr300, 0, undef $vgpr4, 0, undef $vgpr5, 0, undef $vgpr301, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x1000 ; GCN-NEXT: v_dual_fmac_f32 v2, v6, v6 :: v_dual_fma_f32 v3, v4, v5, v3 $vgpr2, $vgpr3 = V_DUAL_FMAC_F32_e32_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr6, undef $vgpr2, 0, undef $vgpr4, 0, undef $vgpr5, 0, $vgpr3, implicit $mode, implicit $exec @@ -298,7 +299,7 @@ body: | ; GCN-NEXT: v_dual_fma_f32 v244 /*v500*/, v6, v7, v8 :: v_dual_add_f32 v3 /*v259*/, v4, v5 $vgpr500, $vgpr259 = V_DUAL_FMA_F32_e64_X_ADD_F32_e32_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr7, 0, undef $vgpr8, 0, undef $vgpr4, 0, undef $vgpr5, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0xae + ; GCN-NEXT: s_set_vgpr_msb 0x40ae ; GCN-NEXT: v_dual_fmac_f32 v2 /*v514*/, v6 /*v518*/, v8 /*v776*/ :: v_dual_fma_f32 v3 /*v515*/, v4 /*v516*/, v7 /*v775*/, v3 /*v515*/ $vgpr514, $vgpr515 = V_DUAL_FMAC_F32_e32_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr518, 0, undef $vgpr776, undef $vgpr514, 0, undef $vgpr516, 0, undef $vgpr775, 0, $vgpr515, implicit $mode, implicit $exec @@ -319,31 +320,31 @@ body: | ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1 /*v257*/, v2 /*v258*/, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr257, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4505 ; GCN-NEXT: v_fmaak_f32 v0, v1 /*v257*/, v2 /*v258*/, 0x1 $vgpr0 = V_FMAAK_F32 undef $vgpr257, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x541 ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1 /*v257*/, v2, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr257, undef $vgpr2, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x4144 ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1, v2 /*v258*/, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr1, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x45 + ; GCN-NEXT: s_set_vgpr_msb 0x4445 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1 /*v257*/, 0x1, v2 /*v258*/ $vgpr256 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4505 ; GCN-NEXT: v_fmamk_f32 v0, v1 /*v257*/, 0x1, v2 /*v258*/ $vgpr0 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x541 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1 /*v257*/, 0x1, v2 $vgpr256 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr2, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x4144 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1, 0x1, v2 /*v258*/ $vgpr256 = V_FMAMK_F32 undef $vgpr1, 1, undef $vgpr258, implicit $exec, implicit $mode @@ -389,15 +390,15 @@ body: | ; GCN-NEXT: v_lshlrev_b32_e64 v0, v0 /*v256*/, v2 $vgpr0 = V_LSHLREV_B32_e64 undef $vgpr256, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_lshlrev_b32_e64 v0, v1, v0 /*v256*/ $vgpr0 = V_LSHLREV_B32_e64 undef $vgpr1, undef $vgpr256, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_subrev_nc_u32_e32 v0, v0 /*v256*/, v2 $vgpr0 = V_SUBREV_U32_e32 undef $vgpr256, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_subrev_nc_u32_e32 v0, v1, v0 /*v256*/ $vgpr0 = V_SUBREV_U32_e32 undef $vgpr1, undef $vgpr256, implicit $exec @@ -417,7 +418,7 @@ body: | ; GCN-NEXT: v_fma_f32 v3 /*v259*/, v4 /*v260*/, v5 /*v261*/, v6 /*v262*/ $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x5500 ; GCN-NEXT: v_add_nc_u32_e32 v0, v1, v2 $vgpr0 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr2, implicit $exec @@ -431,7 +432,7 @@ body: | ; GCN-NEXT: v_add_nc_u32_e32 v0 /*v256*/, v1, v2 $vgpr256 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_fma_f32 v3, v4, v5, s2 $vgpr3 = V_FMA_F32_e64 0, undef $vgpr4, 0, undef $vgpr5, 0, undef $sgpr2, 0, 0, implicit $exec, implicit $mode @@ -439,17 +440,17 @@ body: | ; GCN-NEXT: v_fma_f32 v3, v4 /*v260*/, v5, 1 $vgpr3 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr5, 0, 1, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_mov_b32_e32 v0, v1 $vgpr0 = V_MOV_B32_e32 undef $vgpr1, implicit $exec ; GCN-NEXT: v_add_nc_u32_e32 v2, v1, v3 /*v259*/ $vgpr2 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr259, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_mov_b32_e32 v0, v0 /*v256*/ ; GCN-NEXT: v_add_nc_u32_e32 v1, v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x105 ; GCN-NEXT: v_add_nc_u32_e32 v2, v2 /*v258*/, v2 /*v258*/ $vgpr0 = V_MOV_B32_e32 undef $vgpr256, implicit $exec $vgpr1 = V_ADD_U32_e32 undef $vgpr257, undef $vgpr1, implicit $exec @@ -478,16 +479,18 @@ body: | ; ASM: .LBB{{.*_1}}: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec - ; No mode switch on fall through + ; Reset on fallthrough block end bb.2: ; ASM-NEXT: %bb.2: - ; GCN-NEXT: s_nop 0 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_branch - S_NOP 0 + $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_BRANCH %bb.3 ; Reset mode on terminator @@ -496,7 +499,7 @@ body: | ; ASM: .LBB{{.*_3}}: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_swap_pc_i64 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec $exec = S_SWAPPC_B64 undef $sgpr0_sgpr1 @@ -518,7 +521,7 @@ body: | ; GCN-NEXT: v_mov_b32_e32 v0, v1 ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_set_pc_i64 $vgpr0 = V_MOV_B32_e32 undef $vgpr1, implicit $exec $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec @@ -538,7 +541,7 @@ body: | ; ASM-NEXT: %bb.7: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; ASM-NEXT: ; return to shader part epilog $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec SI_RETURN_TO_EPILOG undef $vgpr0, implicit-def $exec @@ -556,7 +559,7 @@ body: | ; ASM-NEXT: %bb.9: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_set_pc_i64 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_SETPC_B64_return undef $sgpr0_sgpr1, implicit-def $exec @@ -574,13 +577,14 @@ body: | ; ASM: %bb.0: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 $vgpr256 = V_MOV_B32_e32 undef $vgpr0, implicit $exec bb.1: ; ASM: .LBB{{[0-9]+}}_1: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_cbranch_scc0 $vgpr257 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_CBRANCH_SCC0 %bb.1, undef implicit $scc @@ -604,7 +608,7 @@ body: | ; ASM: %bb.0: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; ASM: def v0 ; GCN-NOT: s_set_vgpr_msb ; ASM: use v0 @@ -638,7 +642,7 @@ body: | ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 ; GCN-NEXT: s_nop 0 - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4001 ; GCN-NEXT: v_mov_b32_e32 v1, v0 /*v256*/ BUNDLE implicit-def $vgpr256 { $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec @@ -680,7 +684,7 @@ body: | ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 ; GCN-NEXT: v_mov_b32_e32 v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_mov_b32_e32 v2, v1 ; GCN-NEXT: v_mov_b32_e32 v3, v1 BUNDLE implicit-def $vgpr256, implicit-def $vgpr257, implicit-def $vgpr2, implicit-def $vgpr3, implicit undef $vgpr1 { @@ -709,7 +713,7 @@ body: | ; GCN-NEXT: s_clause 0x3e ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_mov_b32_e32 v1, v1 ; GCN-NEXT: v_mov_b32_e32 v2, v1 ; GCN-COUNT-60: v_mov_b32_e32 v1, v1 @@ -823,7 +827,7 @@ body: | ; GCN-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[210:217], v[244:259] /*v[500:515]*/, v[244:259] /*v[500:515]*/, v[10:17], v1, v2 $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr1, undef $vgpr2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x500 ; GCN-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[210:217], v[100:115], v[100:115], v[10:17], v1, v2 $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr1, undef $vgpr2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec @@ -835,11 +839,11 @@ body: | ; GCN-NEXT: v_wmma_ld_scale16_paired_b64 v[0:1], v[2:3] V_WMMA_LD_SCALE16_PAIRED_B64 undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x105 ; GCN-NEXT: v_wmma_scale16_f32_16x16x128_f8f6f4 v[210:217], v[244:259] /*v[500:515]*/, v[244:259] /*v[500:515]*/, v[10:17], v[0:1], v[2:3] $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE16_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x500 ; GCN-NEXT: v_wmma_scale16_f32_16x16x128_f8f6f4 v[210:217], v[100:115], v[100:115], v[10:17], v[0:1], v[2:3] $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE16_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll b/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll index db49339..9c16b3c 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll +++ b/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll @@ -22,8 +22,6 @@ ; GFX9-DAG: s_mov_b32 s[[DESC3:[0-9]+]], 0xe00000 ; OFFREG is offset system SGPR -; GCN: buffer_store_dword {{v[0-9]+}}, off, s[[[DESC0]]:[[DESC3]]], 0 offset:{{[0-9]+}} ; 4-byte Folded Spill -; GCN: buffer_load_dword v{{[0-9]+}}, off, s[[[DESC0]]:[[DESC3]]], 0 offset:{{[0-9]+}} ; 4-byte Folded Reload ; GCN: NumVgprs: 256 ; GCN: ScratchSize: 640 diff --git a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll index a42c8ac7..7581710 100644 --- a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll +++ b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll @@ -3182,7 +3182,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s33 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s33 offset:1596 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s33 offset:1600 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s33 offset:1608 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s33 offset:1612 @@ -3443,7 +3443,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s33 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s33 offset:2620 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s33 offset:2624 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s33 offset:2632 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s33 offset:2636 @@ -3706,7 +3706,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v1023*/, s33 offset:3648 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: v_writelane_b32 v40, s0, 3 @@ -4135,7 +4135,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s33 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s33 offset:1596 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s33 offset:1600 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s33 offset:1608 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s33 offset:1612 @@ -4396,7 +4396,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s33 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s33 offset:2620 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s33 offset:2624 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s33 offset:2632 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s33 offset:2636 @@ -4661,7 +4661,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s4 ; GFX1250-DAGISEL-NEXT: s_mov_b32 s33, s0 ; GFX1250-DAGISEL-NEXT: s_wait_loadcnt 0x0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[30:31] %ret = call amdgpu_gfx <2 x half>(<2 x half>, <2 x half>) @gfx_callee(<2 x half> %y, <2 x half> %x) convergent ret <2 x half> %ret @@ -6346,7 +6346,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s32 offset:1588 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s32 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s32 offset:1596 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s32 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s32 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s32 offset:1608 @@ -6607,7 +6607,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s32 offset:2612 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s32 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s32 offset:2620 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s32 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s32 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s32 offset:2632 @@ -6872,7 +6872,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 ; GFX1250-DAGISEL-NEXT: v_mov_b32_e32 v2, v0 ; GFX1250-DAGISEL-NEXT: s_mov_b64 s[36:37], gfx_callee@abs64 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: v_swap_b32 v0, v1 ; GFX1250-DAGISEL-NEXT: s_xor_b32 exec_lo, s0, -1 ; GFX1250-DAGISEL-NEXT: s_clause 0x3e @@ -7283,7 +7283,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s32 offset:1588 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s32 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s32 offset:1596 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s32 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s32 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s32 offset:1608 @@ -7544,7 +7544,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s32 offset:2612 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s32 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s32 offset:2620 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s32 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s32 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s32 offset:2632 @@ -7807,7 +7807,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v1023*/, off, s32 offset:3644 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[36:37] %ret = tail call amdgpu_gfx <2 x half>(<2 x half>, <2 x half>) @gfx_callee(<2 x half> %y, <2 x half> %x) convergent ret <2 x half> %ret @@ -9657,7 +9657,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s33 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s33 offset:1608 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s33 offset:1612 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s33 offset:1616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s33 offset:1620 @@ -9918,7 +9918,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s33 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s33 offset:2632 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s33 offset:2636 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s33 offset:2640 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s33 offset:2644 @@ -10181,7 +10181,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v1023*/, s33 offset:3656 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_clause 0x2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v42, s33 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v40, s33 offset:164 @@ -10616,7 +10616,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s33 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s33 offset:1608 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s33 offset:1612 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s33 offset:1616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s33 offset:1620 @@ -10877,7 +10877,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s33 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s33 offset:2632 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s33 offset:2636 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s33 offset:2640 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s33 offset:2644 @@ -11142,7 +11142,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s4 ; GFX1250-DAGISEL-NEXT: s_mov_b32 s33, s0 ; GFX1250-DAGISEL-NEXT: s_wait_loadcnt_dscnt 0x0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[30:31] %ret = call float(ptr, ...) @llvm.amdgcn.call.whole.wave(ptr @callee, <8 x float> %x) convergent store float %ret, ptr %p diff --git a/llvm/test/CodeGen/AMDGPU/wqm.ll b/llvm/test/CodeGen/AMDGPU/wqm.ll index ad8dcd3..21f0c00 100644 --- a/llvm/test/CodeGen/AMDGPU/wqm.ll +++ b/llvm/test/CodeGen/AMDGPU/wqm.ll @@ -3477,13 +3477,10 @@ define amdgpu_gs void @wqm_init_exec_wwm() { ; GFX9-W64-NEXT: s_mov_b64 exec, 0 ; GFX9-W64-NEXT: s_mov_b32 s1, 0 ; GFX9-W64-NEXT: s_mov_b32 s0, s1 -; GFX9-W64-NEXT: s_cmp_lg_u64 exec, 0 -; GFX9-W64-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX9-W64-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-W64-NEXT: s_cmp_eq_u64 s[0:1], 0 ; GFX9-W64-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX9-W64-NEXT: s_xor_b64 s[0:1], s[2:3], s[0:1] -; GFX9-W64-NEXT: v_cndmask_b32_e64 v0, 0, 1.0, s[0:1] -; GFX9-W64-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-W64-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-W64-NEXT: v_cndmask_b32_e64 v1, 0, 1.0, s[0:1] ; GFX9-W64-NEXT: exp mrt0 off, off, off, off ; GFX9-W64-NEXT: s_endpgm ; @@ -3491,14 +3488,11 @@ define amdgpu_gs void @wqm_init_exec_wwm() { ; GFX10-W32: ; %bb.0: ; GFX10-W32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-W32-NEXT: s_mov_b32 s1, 0 -; GFX10-W32-NEXT: s_cmp_lg_u64 exec, 0 +; GFX10-W32-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-W32-NEXT: s_mov_b32 s0, s1 -; GFX10-W32-NEXT: s_cselect_b32 s2, -1, 0 -; GFX10-W32-NEXT: s_cmp_lg_u64 s[0:1], 0 -; GFX10-W32-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-W32-NEXT: s_cmp_eq_u64 s[0:1], 0 ; GFX10-W32-NEXT: s_cselect_b32 s0, -1, 0 -; GFX10-W32-NEXT: s_xor_b32 s0, s2, s0 -; GFX10-W32-NEXT: v_cndmask_b32_e64 v0, 0, 1.0, s0 +; GFX10-W32-NEXT: v_cndmask_b32_e64 v1, 0, 1.0, s0 ; GFX10-W32-NEXT: exp mrt0 off, off, off, off ; GFX10-W32-NEXT: s_endpgm call void @llvm.amdgcn.init.exec(i64 0) |