[AMDGPU] Handle v_fmac_legacy_f32 in SIFoldOperands

Convert it to v_fma_legacy_f32 if it is profitable to do so, just like
other mac instructions that are converted to their mad equivalents.

Differential Revision: https://reviews.llvm.org/D94010

diff --git a/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp b/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
index 6dc01c3..892dc1f 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
+++ b/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
@@ -140,6 +140,8 @@ static unsigned macToMad(unsigned Opc) {
     return AMDGPU::V_FMA_F32;
   case AMDGPU::V_FMAC_F16_e64:
     return AMDGPU::V_FMA_F16_gfx9;
+  case AMDGPU::V_FMAC_LEGACY_F32_e64:
+    return AMDGPU::V_FMA_LEGACY_F32;
   }
   return AMDGPU::INSTRUCTION_LIST_END;
 }

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fma.legacy.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fma.legacy.ll
index 8bfb81d..e641d12 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fma.legacy.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fma.legacy.ll
@@ -70,16 +70,10 @@ define amdgpu_ps <{ i32, i32, i32, i32, i32, float, float, float, float, float,
 ; SDAG-NEXT:    v_interp_p2_f32_e32 v3, v1, attr0.y
 ; SDAG-NEXT:    s_and_b32 exec_lo, exec_lo, s16
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    image_sample v[7:10], v[2:3], s[8:15], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_2D
-; SDAG-NEXT:    v_mov_b32_e32 v4, -1.0
-; SDAG-NEXT:    v_mov_b32_e32 v5, -1.0
+; SDAG-NEXT:    image_sample v[0:3], v[2:3], s[8:15], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_fmac_legacy_f32_e64 v4, v7, 2.0
-; SDAG-NEXT:    v_fmac_legacy_f32_e64 v5, v8, 2.0
-; SDAG-NEXT:    v_mov_b32_e32 v2, v9
-; SDAG-NEXT:    v_mov_b32_e32 v3, v10
-; SDAG-NEXT:    v_mov_b32_e32 v0, v4
-; SDAG-NEXT:    v_mov_b32_e32 v1, v5
+; SDAG-NEXT:    v_fma_legacy_f32 v0, v0, 2.0, -1.0
+; SDAG-NEXT:    v_fma_legacy_f32 v1, v1, 2.0, -1.0
 ; SDAG-NEXT:    ; return to shader part epilog
 ;
 ; GISEL-LABEL: main:
@@ -100,16 +94,10 @@ define amdgpu_ps <{ i32, i32, i32, i32, i32, float, float, float, float, float,
 ; GISEL-NEXT:    v_interp_p2_f32_e32 v2, v1, attr0.x
 ; GISEL-NEXT:    s_and_b32 exec_lo, exec_lo, s16
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    image_sample v[7:10], v[2:3], s[8:15], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_2D
-; GISEL-NEXT:    v_mov_b32_e32 v4, -1.0
-; GISEL-NEXT:    v_mov_b32_e32 v5, -1.0
+; GISEL-NEXT:    image_sample v[0:3], v[2:3], s[8:15], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GISEL-NEXT:    v_fmac_legacy_f32_e64 v4, v7, 2.0
-; GISEL-NEXT:    v_fmac_legacy_f32_e64 v5, v8, 2.0
-; GISEL-NEXT:    v_mov_b32_e32 v2, v9
-; GISEL-NEXT:    v_mov_b32_e32 v3, v10
-; GISEL-NEXT:    v_mov_b32_e32 v0, v4
-; GISEL-NEXT:    v_mov_b32_e32 v1, v5
+; GISEL-NEXT:    v_fma_legacy_f32 v0, v0, 2.0, -1.0
+; GISEL-NEXT:    v_fma_legacy_f32 v1, v1, 2.0, -1.0
 ; GISEL-NEXT:    ; return to shader part epilog
   %i = bitcast <2 x i32> %arg7 to <2 x float>
   %i22 = extractelement <2 x float> %i, i32 0