diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/fma-combine.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/fma-combine.ll | 2013 |
1 files changed, 1199 insertions, 814 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/fma-combine.ll b/llvm/test/CodeGen/AMDGPU/fma-combine.ll index d7cf411..8fc6904 100644 --- a/llvm/test/CodeGen/AMDGPU/fma-combine.ll +++ b/llvm/test/CodeGen/AMDGPU/fma-combine.ll @@ -1,9 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -denormal-fp-math-f32=preserve-sign < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-NOFMA %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=verde -denormal-fp-math-f32=preserve-sign < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-NOFMA %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -denormal-fp-math-f32=ieee -enable-no-infs-fp-math < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-FMA %s +; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -denormal-fp-math-f32=preserve-sign < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-NOFMA,TAHITI %s +; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=verde -denormal-fp-math-f32=preserve-sign < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-NOFMA,VERDE %s +; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -denormal-fp-math-f32=ieee < %s | FileCheck -enable-var-scope -check-prefixes=SI,SI-FMA %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -denormal-fp-math-f32=preserve-sign < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-NOFMA %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -denormal-fp-math-f32=ieee -enable-no-infs-fp-math < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-FMA %s +; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -denormal-fp-math-f32=ieee < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-FMA %s ; Note: The SI-FMA conversions of type x * (y + 1) --> x * y + x would be ; beneficial even without fp32 denormals, but they do require no-infs-fp-math @@ -648,53 +648,53 @@ define amdgpu_kernel void @combine_to_fma_fsub_2_f64_2uses_mul(ptr addrspace(1) ; fold (fsub (fma x, y, (fmul u, v)), z) -> (fma x, y (fma u, v, (fneg z))) define amdgpu_kernel void @aggressive_combine_to_fma_fsub_0_f64(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) #1 { -; SI-FMA-LABEL: aggressive_combine_to_fma_fsub_0_f64: -; SI-FMA: ; %bb.0: -; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-FMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-FMA-NEXT: s_mov_b32 s6, 0 -; SI-FMA-NEXT: v_lshlrev_b32_e32 v0, 3, v0 -; SI-FMA-NEXT: v_mov_b32_e32 v1, 0 -; SI-FMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-FMA-NEXT: s_mov_b64 s[4:5], s[2:3] -; SI-FMA-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: s_mov_b64 s[2:3], s[6:7] -; SI-FMA-NEXT: v_fma_f64 v[6:7], v[8:9], v[10:11], -v[6:7] -; SI-FMA-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[6:7] -; SI-FMA-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 -; SI-FMA-NEXT: s_endpgm +; SI-LABEL: aggressive_combine_to_fma_fsub_0_f64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, 0 +; SI-NEXT: v_lshlrev_b32_e32 v0, 3, v0 +; SI-NEXT: v_mov_b32_e32 v1, 0 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b64 s[4:5], s[2:3] +; SI-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b64 s[2:3], s[6:7] +; SI-NEXT: v_fma_f64 v[6:7], v[8:9], v[10:11], -v[6:7] +; SI-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[6:7] +; SI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: aggressive_combine_to_fma_fsub_0_f64: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: v_and_b32_e32 v0, 0x3ff, v0 -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_lshlrev_b32_e32 v10, 3, v0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f64 v[4:5], v[6:7], v[8:9], -v[4:5] -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5] -; GFX11-FMA-NEXT: global_store_b64 v10, v[0:1], s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: aggressive_combine_to_fma_fsub_0_f64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_lshlrev_b32_e32 v10, 3, v0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f64 v[4:5], v[6:7], v[8:9], -v[4:5] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5] +; GFX11-NEXT: global_store_b64 v10, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %gep.0 = getelementptr double, ptr addrspace(1) %in, i32 %tid %gep.1 = getelementptr double, ptr addrspace(1) %gep.0, i32 1 @@ -718,55 +718,55 @@ define amdgpu_kernel void @aggressive_combine_to_fma_fsub_0_f64(ptr addrspace(1) } define amdgpu_kernel void @no_aggressive_combine_to_fma_fsub_0_f64(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) #1 { -; SI-NOFMA-LABEL: no_aggressive_combine_to_fma_fsub_0_f64: -; SI-NOFMA: ; %bb.0: -; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NOFMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-NOFMA-NEXT: s_mov_b32 s6, 0 -; SI-NOFMA-NEXT: v_lshlrev_b32_e32 v0, 3, v0 -; SI-NOFMA-NEXT: v_mov_b32_e32 v1, 0 -; SI-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-NOFMA-NEXT: s_mov_b64 s[4:5], s[2:3] -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: s_mov_b64 s[2:3], s[6:7] -; SI-NOFMA-NEXT: v_mul_f64 v[8:9], v[8:9], v[10:11] -; SI-NOFMA-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[8:9] -; SI-NOFMA-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7] -; SI-NOFMA-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 -; SI-NOFMA-NEXT: s_endpgm +; SI-LABEL: no_aggressive_combine_to_fma_fsub_0_f64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, 0 +; SI-NEXT: v_lshlrev_b32_e32 v0, 3, v0 +; SI-NEXT: v_mov_b32_e32 v1, 0 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b64 s[4:5], s[2:3] +; SI-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b64 s[2:3], s[6:7] +; SI-NEXT: v_mul_f64 v[8:9], v[8:9], v[10:11] +; SI-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[8:9] +; SI-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7] +; SI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 +; SI-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: no_aggressive_combine_to_fma_fsub_0_f64: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: v_and_b32_e32 v0, 0x3ff, v0 -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_lshlrev_b32_e32 v10, 3, v0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_mul_f64 v[6:7], v[6:7], v[8:9] -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[6:7] -; GFX11-NOFMA-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5] -; GFX11-NOFMA-NEXT: global_store_b64 v10, v[0:1], s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: no_aggressive_combine_to_fma_fsub_0_f64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_lshlrev_b32_e32 v10, 3, v0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_mul_f64 v[6:7], v[6:7], v[8:9] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[6:7] +; GFX11-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5] +; GFX11-NEXT: global_store_b64 v10, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %gep.0 = getelementptr double, ptr addrspace(1) %in, i32 %tid %gep.1 = getelementptr double, ptr addrspace(1) %gep.0, i32 1 @@ -792,53 +792,53 @@ define amdgpu_kernel void @no_aggressive_combine_to_fma_fsub_0_f64(ptr addrspace ; fold (fsub x, (fma y, z, (fmul u, v))) ; -> (fma (fneg y), z, (fma (fneg u), v, x)) define amdgpu_kernel void @aggressive_combine_to_fma_fsub_1_f64(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) #1 { -; SI-FMA-LABEL: aggressive_combine_to_fma_fsub_1_f64: -; SI-FMA: ; %bb.0: -; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-FMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-FMA-NEXT: s_mov_b32 s6, 0 -; SI-FMA-NEXT: v_lshlrev_b32_e32 v0, 3, v0 -; SI-FMA-NEXT: v_mov_b32_e32 v1, 0 -; SI-FMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-FMA-NEXT: s_mov_b64 s[4:5], s[2:3] -; SI-FMA-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: s_mov_b64 s[2:3], s[6:7] -; SI-FMA-NEXT: v_fma_f64 v[2:3], -v[8:9], v[10:11], v[2:3] -; SI-FMA-NEXT: v_fma_f64 v[2:3], -v[4:5], v[6:7], v[2:3] -; SI-FMA-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 -; SI-FMA-NEXT: s_endpgm +; SI-LABEL: aggressive_combine_to_fma_fsub_1_f64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, 0 +; SI-NEXT: v_lshlrev_b32_e32 v0, 3, v0 +; SI-NEXT: v_mov_b32_e32 v1, 0 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b64 s[4:5], s[2:3] +; SI-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b64 s[2:3], s[6:7] +; SI-NEXT: v_fma_f64 v[2:3], -v[8:9], v[10:11], v[2:3] +; SI-NEXT: v_fma_f64 v[2:3], -v[4:5], v[6:7], v[2:3] +; SI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: aggressive_combine_to_fma_fsub_1_f64: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: v_and_b32_e32 v0, 0x3ff, v0 -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_lshlrev_b32_e32 v10, 3, v0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f64 v[0:1], -v[6:7], v[8:9], v[0:1] -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_fma_f64 v[0:1], -v[2:3], v[4:5], v[0:1] -; GFX11-FMA-NEXT: global_store_b64 v10, v[0:1], s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: aggressive_combine_to_fma_fsub_1_f64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_lshlrev_b32_e32 v10, 3, v0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f64 v[0:1], -v[6:7], v[8:9], v[0:1] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_fma_f64 v[0:1], -v[2:3], v[4:5], v[0:1] +; GFX11-NEXT: global_store_b64 v10, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %gep.0 = getelementptr double, ptr addrspace(1) %in, i32 %tid %gep.1 = getelementptr double, ptr addrspace(1) %gep.0, i32 1 @@ -862,55 +862,55 @@ define amdgpu_kernel void @aggressive_combine_to_fma_fsub_1_f64(ptr addrspace(1) ret void } define amdgpu_kernel void @no_aggressive_combine_to_fma_fsub_1_f64(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) #1 { -; SI-NOFMA-LABEL: no_aggressive_combine_to_fma_fsub_1_f64: -; SI-NOFMA: ; %bb.0: -; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NOFMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-NOFMA-NEXT: s_mov_b32 s6, 0 -; SI-NOFMA-NEXT: v_lshlrev_b32_e32 v0, 3, v0 -; SI-NOFMA-NEXT: v_mov_b32_e32 v1, 0 -; SI-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-NOFMA-NEXT: s_mov_b64 s[4:5], s[2:3] -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: s_mov_b64 s[2:3], s[6:7] -; SI-NOFMA-NEXT: v_mul_f64 v[8:9], v[8:9], v[10:11] -; SI-NOFMA-NEXT: v_fma_f64 v[4:5], v[4:5], v[6:7], v[8:9] -; SI-NOFMA-NEXT: v_add_f64 v[2:3], v[2:3], -v[4:5] -; SI-NOFMA-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 -; SI-NOFMA-NEXT: s_endpgm +; SI-LABEL: no_aggressive_combine_to_fma_fsub_1_f64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, 0 +; SI-NEXT: v_lshlrev_b32_e32 v0, 3, v0 +; SI-NEXT: v_mov_b32_e32 v1, 0 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b64 s[4:5], s[2:3] +; SI-NEXT: buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:8 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:16 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[8:9], v[0:1], s[4:7], 0 addr64 offset:24 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dwordx2 v[10:11], v[0:1], s[4:7], 0 addr64 offset:32 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b64 s[2:3], s[6:7] +; SI-NEXT: v_mul_f64 v[8:9], v[8:9], v[10:11] +; SI-NEXT: v_fma_f64 v[4:5], v[4:5], v[6:7], v[8:9] +; SI-NEXT: v_add_f64 v[2:3], v[2:3], -v[4:5] +; SI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64 +; SI-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: no_aggressive_combine_to_fma_fsub_1_f64: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: v_and_b32_e32 v0, 0x3ff, v0 -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_lshlrev_b32_e32 v10, 3, v0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_mul_f64 v[6:7], v[6:7], v[8:9] -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[6:7] -; GFX11-NOFMA-NEXT: v_add_f64 v[0:1], v[0:1], -v[2:3] -; GFX11-NOFMA-NEXT: global_store_b64 v10, v[0:1], s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: no_aggressive_combine_to_fma_fsub_1_f64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_lshlrev_b32_e32 v10, 3, v0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v10, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[2:3], v10, s[2:3] offset:8 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[4:5], v10, s[2:3] offset:16 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[6:7], v10, s[2:3] offset:24 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b64 v[8:9], v10, s[2:3] offset:32 glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_mul_f64 v[6:7], v[6:7], v[8:9] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], v[6:7] +; GFX11-NEXT: v_add_f64 v[0:1], v[0:1], -v[2:3] +; GFX11-NEXT: global_store_b64 v10, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %gep.0 = getelementptr double, ptr addrspace(1) %in, i32 %tid %gep.1 = getelementptr double, ptr addrspace(1) %gep.0, i32 1 @@ -939,7 +939,58 @@ define amdgpu_kernel void @no_aggressive_combine_to_fma_fsub_1_f64(ptr addrspace ; define amdgpu_kernel void @test_f32_mul_add_x_one_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_add_x_one_y: +; SI-LABEL: test_f32_mul_add_x_one_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_add_x_one_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load volatile float, ptr addrspace(1) %in1 + %y = load volatile float, ptr addrspace(1) %in2 + %a = fadd contract float %x, 1.0 + %m = fmul contract float %a, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_add_x_one_y_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_add_x_one_y_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -958,12 +1009,11 @@ define amdgpu_kernel void @test_f32_mul_add_x_one_y(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: v_add_f32_e32 v0, 1.0, v0 -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NOFMA-NEXT: v_mac_f32_e32 v1, v0, v1 +; SI-NOFMA-NEXT: buffer_store_dword v1, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_add_x_one_y: +; SI-FMA-LABEL: test_f32_mul_add_x_one_y_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -986,49 +1036,83 @@ define amdgpu_kernel void @test_f32_mul_add_x_one_y(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_add_x_one_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_add_x_one_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fmac_f32_e32 v2, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v2, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load volatile float, ptr addrspace(1) %in1 + %y = load volatile float, ptr addrspace(1) %in2 + %a = fadd contract ninf float %x, 1.0 + %m = fmul contract ninf float %a, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_add_x_one(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_y_add_x_one: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 glc +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_add_x_one_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fmac_f32_e32 v2, v1, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v2, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_add_x_one: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load volatile float, ptr addrspace(1) %in1 %y = load volatile float, ptr addrspace(1) %in2 %a = fadd contract float %x, 1.0 - %m = fmul contract float %a, %y + %m = fmul contract float %y, %a store float %m, ptr addrspace(1) %out ret void } -define amdgpu_kernel void @test_f32_mul_y_add_x_one(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_add_x_one: +define amdgpu_kernel void @test_f32_mul_y_add_x_one_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_add_x_one_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1047,12 +1131,11 @@ define amdgpu_kernel void @test_f32_mul_y_add_x_one(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: v_add_f32_e32 v0, 1.0, v0 -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NOFMA-NEXT: v_mac_f32_e32 v1, v0, v1 +; SI-NOFMA-NEXT: buffer_store_dword v1, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_y_add_x_one: +; SI-FMA-LABEL: test_f32_mul_y_add_x_one_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1075,126 +1158,72 @@ define amdgpu_kernel void @test_f32_mul_y_add_x_one(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_add_x_one: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_mul_y_add_x_one: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fmac_f32_e32 v2, v1, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v2, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_add_x_one_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] glc dlc +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fmac_f32_e32 v2, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v2, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load volatile float, ptr addrspace(1) %in1 %y = load volatile float, ptr addrspace(1) %in2 - %a = fadd contract float %x, 1.0 - %m = fmul contract float %y, %a + %a = fadd contract ninf float %x, 1.0 + %m = fmul contract ninf float %y, %a store float %m, ptr addrspace(1) %out ret void } define amdgpu_kernel void @test_f32_mul_add_x_negone_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_add_x_negone_y: -; SI-NOFMA: ; %bb.0: -; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd -; SI-NOFMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-NOFMA-NEXT: s_mov_b32 s6, -1 -; SI-NOFMA-NEXT: s_mov_b32 s14, s6 -; SI-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-NOFMA-NEXT: s_mov_b32 s12, s2 -; SI-NOFMA-NEXT: s_mov_b32 s13, s3 -; SI-NOFMA-NEXT: s_mov_b32 s15, s7 -; SI-NOFMA-NEXT: s_mov_b32 s10, s6 -; SI-NOFMA-NEXT: s_mov_b32 s11, s7 -; SI-NOFMA-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 -; SI-NOFMA-NEXT: s_mov_b32 s4, s0 -; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, -1.0, v0 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-NOFMA-NEXT: s_endpgm -; -; SI-FMA-LABEL: test_f32_mul_add_x_negone_y: -; SI-FMA: ; %bb.0: -; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd -; SI-FMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-FMA-NEXT: s_mov_b32 s6, -1 -; SI-FMA-NEXT: s_mov_b32 s14, s6 -; SI-FMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-FMA-NEXT: s_mov_b32 s12, s2 -; SI-FMA-NEXT: s_mov_b32 s13, s3 -; SI-FMA-NEXT: s_mov_b32 s15, s7 -; SI-FMA-NEXT: s_mov_b32 s10, s6 -; SI-FMA-NEXT: s_mov_b32 s11, s7 -; SI-FMA-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-FMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 -; SI-FMA-NEXT: s_mov_b32 s4, s0 -; SI-FMA-NEXT: s_mov_b32 s5, s1 -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f32 v0, v0, v1, -v1 -; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-FMA-NEXT: s_endpgm -; -; GFX11-NOFMA-LABEL: test_f32_mul_add_x_negone_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; SI-LABEL: test_f32_mul_add_x_negone_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_add_x_negone_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_add_x_negone_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 @@ -1205,8 +1234,8 @@ define amdgpu_kernel void @test_f32_mul_add_x_negone_y(ptr addrspace(1) %out, ret void } -define amdgpu_kernel void @test_f32_mul_y_add_x_negone(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_add_x_negone: +define amdgpu_kernel void @test_f32_mul_add_x_negone_y_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_add_x_negone_y_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1223,14 +1252,12 @@ define amdgpu_kernel void @test_f32_mul_y_add_x_negone(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, -1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NOFMA-NEXT: v_mad_f32 v0, v0, v1, -v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_y_add_x_negone: +; SI-FMA-LABEL: test_f32_mul_add_x_negone_y_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1252,38 +1279,72 @@ define amdgpu_kernel void @test_f32_mul_y_add_x_negone(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_add_x_negone: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_add_x_negone_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, v1, v2, -v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %a = fadd contract ninf float %x, -1.0 + %m = fmul contract ninf float %a, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_add_x_negone(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_y_add_x_negone: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_y_add_x_negone: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_add_x_negone: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 @@ -1294,8 +1355,8 @@ define amdgpu_kernel void @test_f32_mul_y_add_x_negone(ptr addrspace(1) %out, ret void } -define amdgpu_kernel void @test_f32_mul_sub_one_x_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_sub_one_x_y: +define amdgpu_kernel void @test_f32_mul_y_add_x_negone_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_add_x_negone_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1312,14 +1373,12 @@ define amdgpu_kernel void @test_f32_mul_sub_one_x_y(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_sub_f32_e32 v0, 1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NOFMA-NEXT: v_mad_f32 v0, v0, v1, -v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_sub_one_x_y: +; SI-FMA-LABEL: test_f32_mul_y_add_x_negone_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1337,42 +1396,76 @@ define amdgpu_kernel void @test_f32_mul_sub_one_x_y(ptr addrspace(1) %out, ; SI-FMA-NEXT: s_mov_b32 s4, s0 ; SI-FMA-NEXT: s_mov_b32 s5, s1 ; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f32 v0, -v0, v1, v1 +; SI-FMA-NEXT: v_fma_f32 v0, v0, v1, -v1 ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_sub_one_x_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_sub_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_add_x_negone_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, v1, v2, -v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %a = fadd contract ninf float %x, -1.0 + %m = fmul contract ninf float %y, %a + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_sub_one_x_y(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_sub_one_x_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, 1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_sub_one_x_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, -v1, v2, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_sub_one_x_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 @@ -1383,85 +1476,100 @@ define amdgpu_kernel void @test_f32_mul_sub_one_x_y(ptr addrspace(1) %out, ret void } -define amdgpu_kernel void @test_f32_mul_y_sub_one_x(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_sub_one_x: -; SI-NOFMA: ; %bb.0: -; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd -; SI-NOFMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-NOFMA-NEXT: s_mov_b32 s6, -1 -; SI-NOFMA-NEXT: s_mov_b32 s14, s6 -; SI-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-NOFMA-NEXT: s_mov_b32 s12, s2 -; SI-NOFMA-NEXT: s_mov_b32 s13, s3 -; SI-NOFMA-NEXT: s_mov_b32 s15, s7 -; SI-NOFMA-NEXT: s_mov_b32 s10, s6 -; SI-NOFMA-NEXT: s_mov_b32 s11, s7 -; SI-NOFMA-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 -; SI-NOFMA-NEXT: s_mov_b32 s4, s0 -; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_sub_f32_e32 v0, 1.0, v0 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-NOFMA-NEXT: s_endpgm -; -; SI-FMA-LABEL: test_f32_mul_y_sub_one_x: -; SI-FMA: ; %bb.0: -; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd -; SI-FMA-NEXT: s_mov_b32 s7, 0xf000 -; SI-FMA-NEXT: s_mov_b32 s6, -1 -; SI-FMA-NEXT: s_mov_b32 s14, s6 -; SI-FMA-NEXT: s_waitcnt lgkmcnt(0) -; SI-FMA-NEXT: s_mov_b32 s12, s2 -; SI-FMA-NEXT: s_mov_b32 s13, s3 -; SI-FMA-NEXT: s_mov_b32 s15, s7 -; SI-FMA-NEXT: s_mov_b32 s10, s6 -; SI-FMA-NEXT: s_mov_b32 s11, s7 -; SI-FMA-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-FMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 -; SI-FMA-NEXT: s_mov_b32 s4, s0 -; SI-FMA-NEXT: s_mov_b32 s5, s1 -; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f32 v0, -v0, v1, v1 -; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-FMA-NEXT: s_endpgm +define amdgpu_kernel void @test_f32_mul_sub_one_x_y_ninf(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_sub_one_x_y_ninf: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, 1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_sub_one_x: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_sub_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_sub_one_x_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract ninf float 1.0, %x + %m = fmul contract ninf float %s, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_sub_one_x(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_y_sub_one_x: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, 1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_y_sub_one_x: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, -v1, v2, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_one_x: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 @@ -1472,8 +1580,8 @@ define amdgpu_kernel void @test_f32_mul_y_sub_one_x(ptr addrspace(1) %out, ret void } -define amdgpu_kernel void @test_f32_mul_sub_negone_x_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_sub_negone_x_y: +define amdgpu_kernel void @test_f32_mul_y_sub_one_x_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_sub_one_x_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1490,14 +1598,12 @@ define amdgpu_kernel void @test_f32_mul_sub_negone_x_y(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_sub_f32_e32 v0, -1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NOFMA-NEXT: v_mad_f32 v0, -v0, v1, v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_sub_negone_x_y: +; SI-FMA-LABEL: test_f32_mul_y_sub_one_x_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1515,42 +1621,76 @@ define amdgpu_kernel void @test_f32_mul_sub_negone_x_y(ptr addrspace(1) %out, ; SI-FMA-NEXT: s_mov_b32 s4, s0 ; SI-FMA-NEXT: s_mov_b32 s5, s1 ; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f32 v0, -v0, v1, -v1 +; SI-FMA-NEXT: v_fma_f32 v0, -v0, v1, v1 ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_sub_negone_x_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_sub_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_one_x_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, -v1, v2, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract ninf float 1.0, %x + %m = fmul contract ninf float %y, %s + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_sub_negone_x_y(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_sub_negone_x_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_sub_negone_x_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, -v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_sub_negone_x_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 @@ -1561,8 +1701,112 @@ define amdgpu_kernel void @test_f32_mul_sub_negone_x_y(ptr addrspace(1) %out, ret void } +define amdgpu_kernel void @test_f32_mul_sub_negone_x_y_ninf(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_sub_negone_x_y_ninf: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_sub_negone_x_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract ninf float -1.0, %x + %m = fmul contract ninf float %s, %y + store float %m, ptr addrspace(1) %out + ret void +} + define amdgpu_kernel void @test_f32_mul_y_sub_negone_x(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_sub_negone_x: +; SI-LABEL: test_f32_mul_y_sub_negone_x: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_sub_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_y_sub_negone_x: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_sub_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract float -1.0, %x + %m = fmul contract float %y, %s + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_sub_negone_x_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_sub_negone_x_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1579,14 +1823,12 @@ define amdgpu_kernel void @test_f32_mul_y_sub_negone_x(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_sub_f32_e32 v0, -1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NOFMA-NEXT: v_mad_f32 v0, -v0, v1, -v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_y_sub_negone_x: +; SI-FMA-LABEL: test_f32_mul_y_sub_negone_x_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1608,50 +1850,84 @@ define amdgpu_kernel void @test_f32_mul_y_sub_negone_x(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_sub_negone_x: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_sub_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_mul_y_sub_negone_x: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, -v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_negone_x_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, -v1, v2, -v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 %y = load float, ptr addrspace(1) %in2 - %s = fsub contract float -1.0, %x - %m = fmul contract float %y, %s + %s = fsub contract ninf float -1.0, %x + %m = fmul contract ninf float %y, %s store float %m, ptr addrspace(1) %out ret void } define amdgpu_kernel void @test_f32_mul_sub_x_one_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_sub_x_one_y: +; SI-LABEL: test_f32_mul_sub_x_one_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_sub_x_one_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract float %x, 1.0 + %m = fmul contract float %s, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_sub_x_one_y_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_sub_x_one_y_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1668,14 +1944,12 @@ define amdgpu_kernel void @test_f32_mul_sub_x_one_y(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, -1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NOFMA-NEXT: v_mad_f32 v0, v0, v1, -v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_sub_x_one_y: +; SI-FMA-LABEL: test_f32_mul_sub_x_one_y_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1697,50 +1971,84 @@ define amdgpu_kernel void @test_f32_mul_sub_x_one_y(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_sub_x_one_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_mul_sub_x_one_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_sub_x_one_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, v1, v2, -v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 %y = load float, ptr addrspace(1) %in2 - %s = fsub contract float %x, 1.0 - %m = fmul contract float %s, %y + %s = fsub contract ninf float %x, 1.0 + %m = fmul contract ninf float %s, %y store float %m, ptr addrspace(1) %out ret void } define amdgpu_kernel void @test_f32_mul_y_sub_x_one(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_sub_x_one: +; SI-LABEL: test_f32_mul_y_sub_x_one: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, -1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_y_sub_x_one: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, -1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract float %x, 1.0 + %m = fmul contract float %y, %s + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_sub_x_one_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_sub_x_one_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1757,14 +2065,12 @@ define amdgpu_kernel void @test_f32_mul_y_sub_x_one(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, -1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NOFMA-NEXT: v_mad_f32 v0, v0, v1, -v1 ; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_y_sub_x_one: +; SI-FMA-LABEL: test_f32_mul_y_sub_x_one_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1786,50 +2092,84 @@ define amdgpu_kernel void @test_f32_mul_y_sub_x_one(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_sub_x_one: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, -1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_mul_y_sub_x_one: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fma_f32 v1, v1, v2, -v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_x_one_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f32 v1, v1, v2, -v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 %y = load float, ptr addrspace(1) %in2 - %s = fsub contract float %x, 1.0 - %m = fmul contract float %y, %s + %s = fsub contract ninf float %x, 1.0 + %m = fmul contract ninf float %y, %s store float %m, ptr addrspace(1) %out ret void } define amdgpu_kernel void @test_f32_mul_sub_x_negone_y(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_sub_x_negone_y: +; SI-LABEL: test_f32_mul_sub_x_negone_y: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v0, v1 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; GFX11-LABEL: test_f32_mul_sub_x_negone_y: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract float %x, -1.0 + %m = fmul contract float %s, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_sub_x_negone_y_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_sub_x_negone_y_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1846,14 +2186,12 @@ define amdgpu_kernel void @test_f32_mul_sub_x_negone_y(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, 1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v0, v1 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NOFMA-NEXT: v_mac_f32_e32 v1, v0, v1 +; SI-NOFMA-NEXT: buffer_store_dword v1, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_sub_x_negone_y: +; SI-FMA-LABEL: test_f32_mul_sub_x_negone_y_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1875,50 +2213,84 @@ define amdgpu_kernel void @test_f32_mul_sub_x_negone_y(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_sub_x_negone_y: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v1, v2 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_sub_x_negone_y_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fmac_f32_e32 v2, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v2, s[0:1] +; GFX11-NEXT: s_endpgm + ptr addrspace(1) %in1, + ptr addrspace(1) %in2) { + %x = load float, ptr addrspace(1) %in1 + %y = load float, ptr addrspace(1) %in2 + %s = fsub contract ninf float %x, -1.0 + %m = fmul contract ninf float %s, %y + store float %m, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @test_f32_mul_y_sub_x_negone(ptr addrspace(1) %out, +; SI-LABEL: test_f32_mul_y_sub_x_negone: +; 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_mov_b32 s14, s6 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_mov_b32 s15, s7 +; SI-NEXT: s_mov_b32 s10, s6 +; SI-NEXT: s_mov_b32 s11, s7 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_waitcnt vmcnt(1) +; SI-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_mul_f32_e32 v0, v1, v0 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm ; -; GFX11-FMA-LABEL: test_f32_mul_sub_x_negone_y: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fmac_f32_e32 v2, v1, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v2, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_x_negone: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: v_add_f32_e32 v1, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v1, v2, v1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 %y = load float, ptr addrspace(1) %in2 %s = fsub contract float %x, -1.0 - %m = fmul contract float %s, %y + %m = fmul contract float %y, %s store float %m, ptr addrspace(1) %out ret void } -define amdgpu_kernel void @test_f32_mul_y_sub_x_negone(ptr addrspace(1) %out, -; SI-NOFMA-LABEL: test_f32_mul_y_sub_x_negone: +define amdgpu_kernel void @test_f32_mul_y_sub_x_negone_ninf(ptr addrspace(1) %out, +; SI-NOFMA-LABEL: test_f32_mul_y_sub_x_negone_ninf: ; SI-NOFMA: ; %bb.0: ; SI-NOFMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-NOFMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1935,14 +2307,12 @@ define amdgpu_kernel void @test_f32_mul_y_sub_x_negone(ptr addrspace(1) %out, ; SI-NOFMA-NEXT: buffer_load_dword v1, off, s[8:11], 0 ; SI-NOFMA-NEXT: s_mov_b32 s4, s0 ; SI-NOFMA-NEXT: s_mov_b32 s5, s1 -; SI-NOFMA-NEXT: s_waitcnt vmcnt(1) -; SI-NOFMA-NEXT: v_add_f32_e32 v0, 1.0, v0 ; SI-NOFMA-NEXT: s_waitcnt vmcnt(0) -; SI-NOFMA-NEXT: v_mul_f32_e32 v0, v1, v0 -; SI-NOFMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NOFMA-NEXT: v_mac_f32_e32 v1, v0, v1 +; SI-NOFMA-NEXT: buffer_store_dword v1, off, s[4:7], 0 ; SI-NOFMA-NEXT: s_endpgm ; -; SI-FMA-LABEL: test_f32_mul_y_sub_x_negone: +; SI-FMA-LABEL: test_f32_mul_y_sub_x_negone_ninf: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; SI-FMA-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd @@ -1964,44 +2334,26 @@ define amdgpu_kernel void @test_f32_mul_y_sub_x_negone(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_store_dword v0, off, s[4:7], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_mul_y_sub_x_negone: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x1 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: v_add_f32_e32 v1, 1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v1, v2, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_mul_y_sub_x_negone: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX11-FMA-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x1 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[2:3] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: v_fmac_f32_e32 v2, v1, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v2, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_mul_y_sub_x_negone_ninf: +; 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 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[2:3] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fmac_f32_e32 v2, v1, v2 +; GFX11-NEXT: global_store_b32 v0, v2, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2) { %x = load float, ptr addrspace(1) %in1 %y = load float, ptr addrspace(1) %in2 - %s = fsub contract float %x, -1.0 - %m = fmul contract float %y, %s + %s = fsub contract ninf float %x, -1.0 + %m = fmul contract ninf float %y, %s store float %m, ptr addrspace(1) %out ret void } @@ -2048,66 +2400,51 @@ define amdgpu_kernel void @test_f32_interp(ptr addrspace(1) %out, ; SI-FMA-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 ; SI-FMA-NEXT: s_mov_b32 s11, 0xf000 ; SI-FMA-NEXT: s_mov_b32 s10, -1 -; SI-FMA-NEXT: s_mov_b32 s18, s10 -; SI-FMA-NEXT: s_mov_b32 s19, s11 +; SI-FMA-NEXT: s_mov_b32 s14, s10 +; SI-FMA-NEXT: s_mov_b32 s15, s11 ; SI-FMA-NEXT: s_waitcnt lgkmcnt(0) ; SI-FMA-NEXT: s_mov_b32 s16, s4 ; SI-FMA-NEXT: s_mov_b32 s17, s5 -; SI-FMA-NEXT: s_mov_b32 s14, s10 -; SI-FMA-NEXT: s_mov_b32 s12, s2 -; SI-FMA-NEXT: s_mov_b32 s13, s3 -; SI-FMA-NEXT: s_mov_b32 s15, s11 ; SI-FMA-NEXT: s_mov_b32 s4, s6 ; SI-FMA-NEXT: s_mov_b32 s5, s7 ; SI-FMA-NEXT: s_mov_b32 s6, s10 ; SI-FMA-NEXT: s_mov_b32 s7, s11 -; SI-FMA-NEXT: buffer_load_dword v0, off, s[16:19], 0 -; SI-FMA-NEXT: buffer_load_dword v1, off, s[4:7], 0 +; SI-FMA-NEXT: s_mov_b32 s12, s2 +; SI-FMA-NEXT: s_mov_b32 s13, s3 +; SI-FMA-NEXT: s_mov_b32 s18, s10 +; SI-FMA-NEXT: s_mov_b32 s19, s11 +; SI-FMA-NEXT: buffer_load_dword v0, off, s[4:7], 0 +; SI-FMA-NEXT: buffer_load_dword v1, off, s[16:19], 0 ; SI-FMA-NEXT: buffer_load_dword v2, off, s[12:15], 0 ; SI-FMA-NEXT: s_mov_b32 s8, s0 ; SI-FMA-NEXT: s_mov_b32 s9, s1 +; SI-FMA-NEXT: s_waitcnt vmcnt(2) +; SI-FMA-NEXT: v_sub_f32_e32 v3, 1.0, v0 ; SI-FMA-NEXT: s_waitcnt vmcnt(1) -; SI-FMA-NEXT: v_fma_f32 v0, -v1, v0, v0 +; SI-FMA-NEXT: v_mul_f32_e32 v1, v1, v3 ; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f32 v0, v2, v1, v0 +; SI-FMA-NEXT: v_fma_f32 v0, v2, v0, v1 ; SI-FMA-NEXT: buffer_store_dword v0, off, s[8:11], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f32_interp: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x2 -; GFX11-NOFMA-NEXT: global_load_b32 v1, v0, s[6:7] -; GFX11-NOFMA-NEXT: global_load_b32 v2, v0, s[4:5] -; GFX11-NOFMA-NEXT: global_load_b32 v3, v0, s[2:3] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(2) -; GFX11-NOFMA-NEXT: v_sub_f32_e32 v4, 1.0, v1 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f32_e32 v2, v2, v4 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_fmac_f32_e32 v2, v3, v1 -; GFX11-NOFMA-NEXT: global_store_b32 v0, v2, s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f32_interp: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX11-FMA-NEXT: v_mov_b32_e32 v0, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x2 -; GFX11-FMA-NEXT: global_load_b32 v1, v0, s[4:5] -; GFX11-FMA-NEXT: global_load_b32 v2, v0, s[6:7] -; GFX11-FMA-NEXT: global_load_b32 v3, v0, s[2:3] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-FMA-NEXT: v_fma_f32 v1, -v2, v1, v1 -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_fmac_f32_e32 v1, v3, v2 -; GFX11-FMA-NEXT: global_store_b32 v0, v1, s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f32_interp: +; 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 0x2 +; GFX11-NEXT: global_load_b32 v1, v0, s[6:7] +; GFX11-NEXT: global_load_b32 v2, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v3, v0, s[2:3] +; GFX11-NEXT: s_waitcnt vmcnt(2) +; GFX11-NEXT: v_sub_f32_e32 v4, 1.0, v1 +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_mul_f32_e32 v2, v2, v4 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fmac_f32_e32 v2, v3, v1 +; GFX11-NEXT: global_store_b32 v0, v2, s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2, ptr addrspace(1) %in3) { @@ -2123,6 +2460,66 @@ define amdgpu_kernel void @test_f32_interp(ptr addrspace(1) %out, } define amdgpu_kernel void @test_f64_interp(ptr addrspace(1) %out, +; TAHITI-LABEL: test_f64_interp: +; TAHITI: ; %bb.0: +; TAHITI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; TAHITI-NEXT: s_mov_b32 s11, 0xf000 +; TAHITI-NEXT: s_mov_b32 s10, -1 +; TAHITI-NEXT: s_mov_b32 s18, s10 +; TAHITI-NEXT: s_mov_b32 s19, s11 +; TAHITI-NEXT: s_waitcnt lgkmcnt(0) +; TAHITI-NEXT: s_mov_b32 s16, s4 +; TAHITI-NEXT: s_mov_b32 s17, s5 +; TAHITI-NEXT: s_mov_b32 s4, s6 +; TAHITI-NEXT: s_mov_b32 s5, s7 +; TAHITI-NEXT: s_mov_b32 s6, s10 +; TAHITI-NEXT: s_mov_b32 s7, s11 +; TAHITI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 +; TAHITI-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 +; TAHITI-NEXT: s_mov_b32 s14, s10 +; TAHITI-NEXT: s_mov_b32 s12, s2 +; TAHITI-NEXT: s_mov_b32 s13, s3 +; TAHITI-NEXT: s_mov_b32 s15, s11 +; TAHITI-NEXT: buffer_load_dwordx2 v[4:5], off, s[12:15], 0 +; TAHITI-NEXT: s_mov_b32 s8, s0 +; TAHITI-NEXT: s_mov_b32 s9, s1 +; TAHITI-NEXT: s_waitcnt vmcnt(2) +; TAHITI-NEXT: v_add_f64 v[6:7], -v[0:1], 1.0 +; TAHITI-NEXT: s_waitcnt vmcnt(1) +; TAHITI-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] +; TAHITI-NEXT: s_waitcnt vmcnt(0) +; TAHITI-NEXT: v_fma_f64 v[0:1], v[4:5], v[0:1], v[2:3] +; TAHITI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; TAHITI-NEXT: s_endpgm +; +; VERDE-LABEL: test_f64_interp: +; VERDE: ; %bb.0: +; VERDE-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; VERDE-NEXT: s_mov_b32 s11, 0xf000 +; VERDE-NEXT: s_mov_b32 s10, -1 +; VERDE-NEXT: s_mov_b32 s14, s10 +; VERDE-NEXT: s_mov_b32 s15, s11 +; VERDE-NEXT: s_waitcnt lgkmcnt(0) +; VERDE-NEXT: s_mov_b32 s12, s6 +; VERDE-NEXT: s_mov_b32 s13, s7 +; VERDE-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; VERDE-NEXT: s_mov_b32 s6, s10 +; VERDE-NEXT: s_mov_b32 s7, s11 +; VERDE-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; VERDE-NEXT: s_mov_b32 s4, s2 +; VERDE-NEXT: s_mov_b32 s5, s3 +; VERDE-NEXT: buffer_load_dwordx2 v[4:5], off, s[4:7], 0 +; VERDE-NEXT: s_mov_b32 s8, s0 +; VERDE-NEXT: s_mov_b32 s9, s1 +; VERDE-NEXT: s_waitcnt vmcnt(2) +; VERDE-NEXT: v_add_f64 v[6:7], -v[0:1], 1.0 +; VERDE-NEXT: s_waitcnt vmcnt(1) +; VERDE-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] +; VERDE-NEXT: s_waitcnt vmcnt(0) +; VERDE-NEXT: v_fma_f64 v[0:1], v[4:5], v[0:1], v[2:3] +; VERDE-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; VERDE-NEXT: s_endpgm +; ; SI-FMA-LABEL: test_f64_interp: ; SI-FMA: ; %bb.0: ; SI-FMA-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 @@ -2137,8 +2534,8 @@ define amdgpu_kernel void @test_f64_interp(ptr addrspace(1) %out, ; SI-FMA-NEXT: s_mov_b32 s5, s7 ; SI-FMA-NEXT: s_mov_b32 s6, s10 ; SI-FMA-NEXT: s_mov_b32 s7, s11 -; SI-FMA-NEXT: buffer_load_dwordx2 v[0:1], off, s[16:19], 0 -; SI-FMA-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; SI-FMA-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 +; SI-FMA-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 ; SI-FMA-NEXT: s_mov_b32 s14, s10 ; SI-FMA-NEXT: s_mov_b32 s12, s2 ; SI-FMA-NEXT: s_mov_b32 s13, s3 @@ -2146,48 +2543,33 @@ define amdgpu_kernel void @test_f64_interp(ptr addrspace(1) %out, ; SI-FMA-NEXT: buffer_load_dwordx2 v[4:5], off, s[12:15], 0 ; SI-FMA-NEXT: s_mov_b32 s8, s0 ; SI-FMA-NEXT: s_mov_b32 s9, s1 +; SI-FMA-NEXT: s_waitcnt vmcnt(2) +; SI-FMA-NEXT: v_add_f64 v[6:7], -v[0:1], 1.0 ; SI-FMA-NEXT: s_waitcnt vmcnt(1) -; SI-FMA-NEXT: v_fma_f64 v[0:1], -v[2:3], v[0:1], v[0:1] +; SI-FMA-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] ; SI-FMA-NEXT: s_waitcnt vmcnt(0) -; SI-FMA-NEXT: v_fma_f64 v[0:1], v[4:5], v[2:3], v[0:1] +; SI-FMA-NEXT: v_fma_f64 v[0:1], v[4:5], v[0:1], v[2:3] ; SI-FMA-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 ; SI-FMA-NEXT: s_endpgm ; -; GFX11-NOFMA-LABEL: test_f64_interp: -; GFX11-NOFMA: ; %bb.0: -; GFX11-NOFMA-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX11-NOFMA-NEXT: v_mov_b32_e32 v8, 0 -; GFX11-NOFMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NOFMA-NEXT: s_clause 0x2 -; GFX11-NOFMA-NEXT: global_load_b64 v[0:1], v8, s[6:7] -; GFX11-NOFMA-NEXT: global_load_b64 v[2:3], v8, s[4:5] -; GFX11-NOFMA-NEXT: global_load_b64 v[4:5], v8, s[2:3] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(2) -; GFX11-NOFMA-NEXT: v_add_f64 v[6:7], -v[0:1], 1.0 -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-NOFMA-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) -; GFX11-NOFMA-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] -; GFX11-NOFMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-NOFMA-NEXT: v_fma_f64 v[0:1], v[4:5], v[0:1], v[2:3] -; GFX11-NOFMA-NEXT: global_store_b64 v8, v[0:1], s[0:1] -; GFX11-NOFMA-NEXT: s_endpgm -; -; GFX11-FMA-LABEL: test_f64_interp: -; GFX11-FMA: ; %bb.0: -; GFX11-FMA-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX11-FMA-NEXT: v_mov_b32_e32 v6, 0 -; GFX11-FMA-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-FMA-NEXT: s_clause 0x2 -; GFX11-FMA-NEXT: global_load_b64 v[0:1], v6, s[4:5] -; GFX11-FMA-NEXT: global_load_b64 v[2:3], v6, s[6:7] -; GFX11-FMA-NEXT: global_load_b64 v[4:5], v6, s[2:3] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(1) -; GFX11-FMA-NEXT: v_fma_f64 v[0:1], -v[2:3], v[0:1], v[0:1] -; GFX11-FMA-NEXT: s_waitcnt vmcnt(0) -; GFX11-FMA-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-FMA-NEXT: v_fma_f64 v[0:1], v[4:5], v[2:3], v[0:1] -; GFX11-FMA-NEXT: global_store_b64 v6, v[0:1], s[0:1] -; GFX11-FMA-NEXT: s_endpgm +; GFX11-LABEL: test_f64_interp: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v8, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: global_load_b64 v[0:1], v8, s[6:7] +; GFX11-NEXT: global_load_b64 v[2:3], v8, s[4:5] +; GFX11-NEXT: global_load_b64 v[4:5], v8, s[2:3] +; GFX11-NEXT: s_waitcnt vmcnt(2) +; GFX11-NEXT: v_add_f64 v[6:7], -v[0:1], 1.0 +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_fma_f64 v[0:1], v[4:5], v[0:1], v[2:3] +; GFX11-NEXT: global_store_b64 v8, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm ptr addrspace(1) %in1, ptr addrspace(1) %in2, ptr addrspace(1) %in3) { @@ -2356,3 +2738,6 @@ define amdgpu_kernel void @fma_neg_b_c_v4f32(ptr addrspace(1) %out, ptr addrspac attributes #0 = { nounwind readnone } attributes #1 = { nounwind } attributes #2 = { nounwind "no-signed-zeros-fp-math"="true" } +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: +; GFX11-FMA: {{.*}} +; GFX11-NOFMA: {{.*}} |