diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/ssubo.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/ssubo.ll | 676 |
1 files changed, 650 insertions, 26 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/ssubo.ll b/llvm/test/CodeGen/AMDGPU/ssubo.ll index ed8b442..382d892 100644 --- a/llvm/test/CodeGen/AMDGPU/ssubo.ll +++ b/llvm/test/CodeGen/AMDGPU/ssubo.ll @@ -1,14 +1,116 @@ -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,SI,FUNC %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,VI,FUNC %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s - +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti | FileCheck %s --check-prefix=SI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga | FileCheck %s --check-prefix=VI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 | FileCheck %s --check-prefix=GFX9 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 | FileCheck %s --check-prefix=GFX10 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 | FileCheck %s --check-prefix=GFX11 declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone declare { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32>, <2 x i32>) nounwind readnone -; FUNC-LABEL: {{^}}ssubo_i64_zext: define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind { +; SI-LABEL: ssubo_i64_zext: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: v_mov_b32_e32 v0, s2 +; SI-NEXT: s_sub_u32 s10, s2, s8 +; SI-NEXT: s_subb_u32 s11, s3, s9 +; SI-NEXT: v_mov_b32_e32 v1, s3 +; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[10:11], v[0:1] +; SI-NEXT: v_cmp_gt_i64_e64 s[2:3], s[8:9], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: v_mov_b32_e32 v1, s11 +; SI-NEXT: v_add_i32_e32 v0, vcc, s10, v0 +; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: ssubo_i64_zext: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v1, s2 +; VI-NEXT: s_sub_u32 s6, s2, s4 +; VI-NEXT: v_mov_b32_e32 v2, s3 +; VI-NEXT: s_subb_u32 s7, s3, s5 +; VI-NEXT: v_cmp_gt_i64_e64 s[8:9], s[4:5], 0 +; VI-NEXT: v_cmp_lt_i64_e32 vcc, s[6:7], v[1:2] +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_xor_b64 s[0:1], s[8:9], vcc +; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, s[0:1] +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: v_add_u32_e32 v2, vcc, s6, v2 +; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: ssubo_i64_zext: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: s_sub_u32 s4, s2, s6 +; GFX9-NEXT: v_mov_b32_e32 v1, s3 +; GFX9-NEXT: s_subb_u32 s5, s3, s7 +; GFX9-NEXT: v_cmp_gt_i64_e64 s[8:9], s[6:7], 0 +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_xor_b64 s[2:3], s[8:9], vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[2:3] +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s4, v0 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: ssubo_i64_zext: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s2, s6 +; GFX10-NEXT: s_subb_u32 s5, s3, s7 +; GFX10-NEXT: v_cmp_gt_i64_e64 s6, s[6:7], 0 +; GFX10-NEXT: v_cmp_lt_i64_e64 s2, s[4:5], s[2:3] +; GFX10-NEXT: s_xor_b32 s2, s6, s2 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: ssubo_i64_zext: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s6, s2, s4 +; GFX11-NEXT: s_subb_u32 s7, s3, s5 +; GFX11-NEXT: v_cmp_gt_i64_e64 s4, s[4:5], 0 +; GFX11-NEXT: v_cmp_lt_i64_e64 s2, s[6:7], s[2:3] +; GFX11-NEXT: s_xor_b32 s2, s4, s2 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: v_add_co_u32 v0, s2, s6, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s7, 0, s2 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind %val = extractvalue { i64, i1 } %ssub, 0 %carry = extractvalue { i64, i1 } %ssub, 1 @@ -18,8 +120,102 @@ define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) ret void } -; FUNC-LABEL: {{^}}s_ssubo_i32: define amdgpu_kernel void @s_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i32 %a, i32 %b) nounwind { +; SI-LABEL: s_ssubo_i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_sub_i32 s12, s8, s9 +; SI-NEXT: s_cmp_gt_i32 s9, 0 +; SI-NEXT: s_cselect_b64 s[10:11], -1, 0 +; SI-NEXT: s_cmp_lt_i32 s12, s8 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_cselect_b64 s[8:9], -1, 0 +; SI-NEXT: v_mov_b32_e32 v0, s12 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_xor_b64 s[4:5], s[10:11], s[8:9] +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s6 +; SI-NEXT: s_mov_b32 s3, s7 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: s_ssubo_i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: s_sub_i32 s6, s4, s5 +; VI-NEXT: s_cmp_gt_i32 s5, 0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lt_i32 s6, s4 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: s_cselect_b64 s[2:3], -1, 0 +; VI-NEXT: v_mov_b32_e32 v4, s6 +; VI-NEXT: s_xor_b64 s[0:1], s[0:1], s[2:3] +; VI-NEXT: flat_store_dword v[0:1], v4 +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: s_ssubo_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mov_b32_e32 v1, s7 +; GFX9-NEXT: s_sub_i32 s4, s6, s7 +; GFX9-NEXT: v_sub_i32 v1, s6, v1 clamp +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, s4, v1 +; GFX9-NEXT: v_mov_b32_e32 v2, s4 +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: global_store_dword v0, v2, s[0:1] +; GFX9-NEXT: global_store_byte v0, v1, s[2:3] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_ssubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_i32 v0, s6, s7 clamp +; GFX10-NEXT: s_sub_i32 s4, s6, s7 +; GFX10-NEXT: v_mov_b32_e32 v2, s4 +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, s4, v0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dword v1, v2, s[0:1] +; GFX10-NEXT: global_store_byte v1, v0, s[2:3] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_ssubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_i32 v0, s6, s7 clamp +; GFX11-NEXT: s_sub_i32 s4, s6, s7 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s4 +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, s4, v0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v1, v2, s[0:1] +; GFX11-NEXT: global_store_b8 v1, v0, s[2:3] +; GFX11-NEXT: s_endpgm %ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind %val = extractvalue { i32, i1 } %ssub, 0 %carry = extractvalue { i32, i1 } %ssub, 1 @@ -28,8 +224,112 @@ define amdgpu_kernel void @s_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_i32: define amdgpu_kernel void @v_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s4, s2 +; SI-NEXT: s_mov_b32 s5, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v2, vcc, v0, v1 +; SI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v1 +; SI-NEXT: v_cmp_lt_i32_e64 s[0:1], v2, v0 +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_dword v2, off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dword v4, v[0:1] +; VI-NEXT: flat_load_dword v5, v[2:3] +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v6, vcc, v4, v5 +; VI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v5 +; VI-NEXT: v_cmp_lt_i32_e64 s[0:1], v6, v4 +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; VI-NEXT: flat_store_dword v[0:1], v6 +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dword v1, v0, s[12:13] +; GFX9-NEXT: global_load_dword v2, v0, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_i32 v3, v1, v2 clamp +; GFX9-NEXT: v_sub_u32_e32 v1, v1, v2 +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v1, v3 +; GFX9-NEXT: global_store_dword v0, v1, s[8:9] +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: global_store_byte v0, v1, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_nc_i32 v3, v1, v2 clamp +; GFX10-NEXT: v_sub_nc_u32_e32 v1, v1, v2 +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v1, v3 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_nc_i32 v3, v1, v2 clamp +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v1, v3 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %a = load i32, ptr addrspace(1) %aptr, align 4 %b = load i32, ptr addrspace(1) %bptr, align 4 %ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind @@ -40,10 +340,109 @@ define amdgpu_kernel void @v_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}s_ssubo_i64: -; GCN: s_sub_u32 -; GCN: s_subb_u32 define amdgpu_kernel void @s_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a, i64 %b) nounwind { +; SI-LABEL: s_ssubo_i64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_sub_u32 s12, s4, s6 +; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: s_subb_u32 s13, s5, s7 +; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[12:13], v[0:1] +; SI-NEXT: v_cmp_gt_i64_e64 s[4:5], s[6:7], 0 +; SI-NEXT: v_mov_b32_e32 v0, s12 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: v_mov_b32_e32 v1, s13 +; SI-NEXT: s_xor_b64 s[4:5], s[4:5], vcc +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s10 +; SI-NEXT: s_mov_b32 s3, s11 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: s_ssubo_i64: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: s_sub_u32 s0, s4, s6 +; VI-NEXT: v_mov_b32_e32 v4, s4 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_subb_u32 s1, s5, s7 +; VI-NEXT: v_mov_b32_e32 v5, s5 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[4:5] +; VI-NEXT: v_cmp_gt_i64_e64 s[2:3], s[6:7], 0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: s_ssubo_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s0, s12, s14 +; GFX9-NEXT: v_mov_b32_e32 v0, s12 +; GFX9-NEXT: v_mov_b32_e32 v1, s13 +; GFX9-NEXT: s_subb_u32 s1, s13, s15 +; GFX9-NEXT: v_cmp_gt_i64_e64 s[2:3], s[14:15], 0 +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v0, s0 +; GFX9-NEXT: v_mov_b32_e32 v1, s1 +; GFX9-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; GFX9-NEXT: global_store_byte v2, v0, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_ssubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s0, s12, s14 +; GFX10-NEXT: s_subb_u32 s1, s13, s15 +; GFX10-NEXT: v_cmp_gt_i64_e64 s2, s[14:15], 0 +; GFX10-NEXT: v_cmp_lt_i64_e64 s3, s[0:1], s[12:13] +; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: v_mov_b32_e32 v1, s1 +; GFX10-NEXT: s_xor_b32 s0, s2, s3 +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v2, v3, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_ssubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s8, s4, s6 +; GFX11-NEXT: s_subb_u32 s9, s5, s7 +; GFX11-NEXT: v_cmp_gt_i64_e64 s6, s[6:7], 0 +; GFX11-NEXT: v_cmp_lt_i64_e64 s4, s[8:9], s[4:5] +; GFX11-NEXT: v_mov_b32_e32 v0, s8 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s9 +; GFX11-NEXT: s_xor_b32 s4, s6, s4 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] +; GFX11-NEXT: s_endpgm %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind %val = extractvalue { i64, i1 } %ssub, 0 %carry = extractvalue { i64, i1 } %ssub, 1 @@ -52,16 +451,121 @@ define amdgpu_kernel void @s_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_i64: -; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc, -; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc, - -; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc, -; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc, - -; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc, -; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc, define amdgpu_kernel void @v_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_i64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s4, s2 +; SI-NEXT: s_mov_b32 s5, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v4, vcc, v0, v2 +; SI-NEXT: v_subb_u32_e32 v5, vcc, v1, v3, vcc +; SI-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; SI-NEXT: v_cmp_lt_i64_e64 s[0:1], v[4:5], v[0:1] +; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[8:11], 0 +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_i64: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dwordx2 v[0:1], v[0:1] +; VI-NEXT: flat_load_dwordx2 v[2:3], v[2:3] +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v6, s2 +; VI-NEXT: v_mov_b32_e32 v7, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v8, vcc, v0, v2 +; VI-NEXT: v_subb_u32_e32 v9, vcc, v1, v3, vcc +; VI-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; VI-NEXT: v_cmp_lt_i64_e64 s[0:1], v[8:9], v[0:1] +; VI-NEXT: flat_store_dwordx2 v[4:5], v[8:9] +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[6:7], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v6, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX9-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v4, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v5, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; GFX9-NEXT: v_cmp_lt_i64_e64 s[0:1], v[4:5], v[0:1] +; GFX9-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX9-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; GFX9-NEXT: global_store_byte v6, v0, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v6, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] +; GFX10-NEXT: v_cmp_lt_i64_e64 s0, v[4:5], v[0:1] +; GFX10-NEXT: s_xor_b32 s0, vcc_lo, s0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX10-NEXT: global_store_byte v6, v0, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[4:11], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v6, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v6, s[8:9] +; GFX11-NEXT: global_load_b64 v[2:3], v6, s[10:11] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v5, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] +; GFX11-NEXT: v_cmp_lt_i64_e64 s0, v[4:5], v[0:1] +; GFX11-NEXT: s_xor_b32 s0, vcc_lo, s0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v6, v[4:5], s[4:5] +; GFX11-NEXT: global_store_b8 v6, v0, s[6:7] +; GFX11-NEXT: s_endpgm %a = load i64, ptr addrspace(1) %aptr, align 4 %b = load i64, ptr addrspace(1) %bptr, align 4 %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind @@ -72,14 +576,134 @@ define amdgpu_kernel void @v_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_v2i32: -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_sub_{{[iu]}}32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_sub_{{[iu]}}32 define amdgpu_kernel void @v_ssubo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_v2i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v5, vcc, v1, v3 +; SI-NEXT: v_sub_i32_e32 v4, vcc, v0, v2 +; SI-NEXT: v_cmp_lt_i32_e64 s[0:1], 0, v3 +; SI-NEXT: v_cmp_lt_i32_e64 s[4:5], v5, v1 +; SI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v2 +; SI-NEXT: v_cmp_lt_i32_e64 s[2:3], v4, v0 +; SI-NEXT: s_xor_b64 s[0:1], s[0:1], s[4:5] +; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[2:3] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[8:11], 0 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_v2i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dwordx2 v[0:1], v[0:1] +; VI-NEXT: flat_load_dwordx2 v[2:3], v[2:3] +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v6, s2 +; VI-NEXT: v_mov_b32_e32 v7, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v9, vcc, v1, v3 +; VI-NEXT: v_sub_u32_e32 v8, vcc, v0, v2 +; VI-NEXT: v_cmp_lt_i32_e64 s[0:1], 0, v3 +; VI-NEXT: v_cmp_lt_i32_e64 s[4:5], v9, v1 +; VI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v2 +; VI-NEXT: v_cmp_lt_i32_e64 s[2:3], v8, v0 +; VI-NEXT: s_xor_b64 s[0:1], s[0:1], s[4:5] +; VI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[2:3] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_dwordx2 v[4:5], v[8:9] +; VI-NEXT: flat_store_dwordx2 v[6:7], v[0:1] +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_v2i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v6, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX9-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v5, v1, v3 +; GFX9-NEXT: v_sub_i32 v1, v1, v3 clamp +; GFX9-NEXT: v_sub_u32_e32 v4, v0, v2 +; GFX9-NEXT: v_sub_i32 v0, v0, v2 clamp +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v5, v1 +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v4, v0 +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX9-NEXT: global_store_dwordx2 v6, v[0:1], s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_v2i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v5, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v5, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v5, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_nc_u32_e32 v4, v1, v3 +; GFX10-NEXT: v_sub_nc_i32 v1, v1, v3 clamp +; GFX10-NEXT: v_sub_nc_u32_e32 v3, v0, v2 +; GFX10-NEXT: v_sub_nc_i32 v0, v0, v2 clamp +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v4, v1 +; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v3, v0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v5, v[3:4], s[8:9] +; GFX10-NEXT: global_store_dwordx2 v5, v[0:1], s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_v2i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v5, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v5, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v5, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v4, v1, v3 +; GFX11-NEXT: v_sub_nc_i32 v1, v1, v3 clamp +; GFX11-NEXT: v_sub_nc_u32_e32 v3, v0, v2 +; GFX11-NEXT: v_sub_nc_i32 v0, v0, v2 clamp +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3) +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v4, v1 +; GFX11-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v3, v0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v5, v[3:4], s[0:1] +; GFX11-NEXT: global_store_b64 v5, v[0:1], s[2:3] +; GFX11-NEXT: s_endpgm %a = load <2 x i32>, ptr addrspace(1) %aptr, align 4 %b = load <2 x i32>, ptr addrspace(1) %bptr, align 4 %sadd = call { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind |