[PATCH] D28615: AMDGPU: Fold free fneg into sin

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added a subscriber: llvm-commits.
Herald added a reviewer: tstellarAMD.
Herald added subscribers: tony-tye, yaxunl, nhaehnle, wdng, kzhuravl.

https://reviews.llvm.org/D28615

Files:
  lib/Target/AMDGPU/AMDGPUISelLowering.cpp
  test/CodeGen/AMDGPU/fneg-combines.ll


Index: test/CodeGen/AMDGPU/fneg-combines.ll
===================================================================
--- test/CodeGen/AMDGPU/fneg-combines.ll
+++ test/CodeGen/AMDGPU/fneg-combines.ll
@@ -1229,9 +1229,50 @@
   ret void
 }
 
+; --------------------------------------------------------------------------------
+; sin tests
+; --------------------------------------------------------------------------------
+
+; GCN-LABEL: {{^}}v_fneg_sin_f32:
+; GCN: {{buffer|flat}}_load_dword [[A:v[0-9]+]]
+; GCN: v_mul_f32_e32 [[MUL:v[0-9]+]], 0x3e22f983, [[A]]
+; GCN: v_fract_f32_e32 [[FRACT:v[0-9]+]], [[MUL]]
+; GCN: v_sin_f32_e64 [[RESULT:V[0-9]+]], -[[FRACT]]
+; GCN: buffer_store_dword [[RESULT]]
+define void @v_fneg_sin_f32(float addrspace(1)* %out, float addrspace(1)* %a.ptr) #0 {
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %tid.ext = sext i32 %tid to i64
+  %a.gep = getelementptr inbounds float, float addrspace(1)* %a.ptr, i64 %tid.ext
+  %out.gep = getelementptr inbounds float, float addrspace(1)* %out, i64 %tid.ext
+  %a = load volatile float, float addrspace(1)* %a.gep
+  %sin = call float @llvm.sin.f32(float %a)
+  %fneg = fsub float -0.000000e+00, %sin
+  store float %fneg, float addrspace(1)* %out.gep
+  ret void
+}
+
+; GCN-LABEL: {{^}}v_fneg_amdgcn_sin_f32:
+; GCN: {{buffer|flat}}_load_dword [[A:v[0-9]+]]
+; GCN: v_sin_f32_e64 [[RESULT:V[0-9]+]], -[[A]]
+; GCN: buffer_store_dword [[RESULT]]
+define void @v_fneg_amdgcn_sin_f32(float addrspace(1)* %out, float addrspace(1)* %a.ptr) #0 {
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %tid.ext = sext i32 %tid to i64
+  %a.gep = getelementptr inbounds float, float addrspace(1)* %a.ptr, i64 %tid.ext
+  %out.gep = getelementptr inbounds float, float addrspace(1)* %out, i64 %tid.ext
+  %a = load volatile float, float addrspace(1)* %a.gep
+  %sin = call float @llvm.amdgcn.sin.f32(float %a)
+  %fneg = fsub float -0.000000e+00, %sin
+  store float %fneg, float addrspace(1)* %out.gep
+  ret void
+}
+
 declare i32 @llvm.amdgcn.workitem.id.x() #1
 declare float @llvm.fma.f32(float, float, float) #1
 declare float @llvm.fmuladd.f32(float, float, float) #1
+declare float @llvm.sin.f32(float) #1
+
+declare float @llvm.amdgcn.sin.f32(float) #1
 declare float @llvm.amdgcn.rcp.f32(float) #1
 declare float @llvm.amdgcn.rcp.legacy(float) #1
 declare float @llvm.amdgcn.fmul.legacy(float, float) #1
Index: lib/Target/AMDGPU/AMDGPUISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -2813,8 +2813,10 @@
   case ISD::FMUL:
   case ISD::FMA:
   case ISD::FMAD:
+  case ISD::FSIN:
   case AMDGPUISD::RCP:
   case AMDGPUISD::RCP_LEGACY:
+  case AMDGPUISD::SIN_HW:
   case AMDGPUISD::FMUL_LEGACY:
     return true;
   default:
@@ -2905,7 +2907,9 @@
   }
   case ISD::FP_EXTEND:
   case AMDGPUISD::RCP:
-  case AMDGPUISD::RCP_LEGACY: {
+  case AMDGPUISD::RCP_LEGACY:
+  case ISD::FSIN:
+  case AMDGPUISD::SIN_HW: {
     SDValue CvtSrc = N0.getOperand(0);
     if (CvtSrc.getOpcode() == ISD::FNEG) {
       // (fneg (fp_extend (fneg x))) -> (fp_extend x)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D28615.84138.patch
Type: text/x-patch
Size: 3156 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170112/309ba2d4/attachment-0001.bin>