diff options
| author | Mingming Liu <mingmingl@google.com> | 2025-09-10 15:25:31 -0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-09-10 15:25:31 -0700 |
| commit | 1417dafa1db9cb1b2b09438aa9f53ea5ab6e36e2 (patch) | |
| tree | 57f4b1f313c8cf74eed8819870f39c36ea263c68 /llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll | |
| parent | 898b813bc8a6d0276bf0f4769f5f2f64b34e632d (diff) | |
| parent | b8cefcb601ddaa18482555c4ff363c01a270c2fe (diff) | |
Merge branch 'main' into users/mingmingl-llvm/samplefdo-profile-formatusers/mingmingl-llvm/samplefdo-profile-format
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll index 78be949baaba..c1508c1675fe 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll @@ -5058,10 +5058,10 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0 ; GFX90A-NEXT: v_mov_b32_e32 v1, 1.0 ; GFX90A-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 ; GFX90A-NEXT: v_accvgpr_write_b32 a0, 1.0 -; GFX90A-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX90A-NEXT: v_accvgpr_write_b32 a1, 2.0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a2, a0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a3, a0 +; GFX90A-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX90A-NEXT: v_mov_b32_e32 v0, 0 ; GFX90A-NEXT: s_nop 0 ; GFX90A-NEXT: v_mfma_f32_4x4x1f32 a[0:3], v1, v2, a[0:3] @@ -5075,10 +5075,10 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0 ; GFX942-NEXT: v_mov_b32_e32 v1, 1.0 ; GFX942-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 ; GFX942-NEXT: v_accvgpr_write_b32 a0, 1.0 -; GFX942-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX942-NEXT: v_accvgpr_write_b32 a1, 2.0 ; GFX942-NEXT: v_accvgpr_mov_b32 a2, a0 ; GFX942-NEXT: v_accvgpr_mov_b32 a3, a0 +; GFX942-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX942-NEXT: v_mov_b32_e32 v0, 0 ; GFX942-NEXT: s_nop 0 ; GFX942-NEXT: v_mfma_f32_4x4x1_16b_f32 a[0:3], v1, v2, a[0:3] @@ -5222,7 +5222,6 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) # ; GFX90A: ; %bb.0: ; %bb ; GFX90A-NEXT: v_mov_b32_e32 v1, 1.0 ; GFX90A-NEXT: v_accvgpr_write_b32 a0, 1.0 -; GFX90A-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX90A-NEXT: v_accvgpr_write_b32 a15, 2.0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a1, a0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a2, a0 @@ -5238,6 +5237,7 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) # ; GFX90A-NEXT: v_accvgpr_mov_b32 a12, a0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a13, a0 ; GFX90A-NEXT: v_accvgpr_mov_b32 a14, a0 +; GFX90A-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX90A-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 ; GFX90A-NEXT: v_mov_b32_e32 v0, 0 ; GFX90A-NEXT: v_mfma_f32_16x16x1f32 a[0:15], v1, v2, a[0:15] @@ -5254,7 +5254,6 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) # ; GFX942: ; %bb.0: ; %bb ; GFX942-NEXT: v_mov_b32_e32 v1, 1.0 ; GFX942-NEXT: v_accvgpr_write_b32 a0, 1.0 -; GFX942-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX942-NEXT: v_accvgpr_write_b32 a15, 2.0 ; GFX942-NEXT: v_accvgpr_mov_b32 a1, a0 ; GFX942-NEXT: v_accvgpr_mov_b32 a2, a0 @@ -5270,6 +5269,7 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) # ; GFX942-NEXT: v_accvgpr_mov_b32 a12, a0 ; GFX942-NEXT: v_accvgpr_mov_b32 a13, a0 ; GFX942-NEXT: v_accvgpr_mov_b32 a14, a0 +; GFX942-NEXT: v_mov_b32_e32 v2, 2.0 ; GFX942-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 ; GFX942-NEXT: v_mov_b32_e32 v0, 0 ; GFX942-NEXT: v_mfma_f32_16x16x1_4b_f32 a[0:15], v1, v2, a[0:15] |