[PATCH] D110156: [AMDGPU] Convert mac/fmac to mad/fma when folding output modifiers

[Jay Foad via Phabricator via llvm-commits]

This revision was automatically updated to reflect the committed changes.
Closed by commit rG0205806d0fe5: [AMDGPU] Convert mac/fmac to mad/fma when folding output modifiers (authored by foad).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D110156/new/

https://reviews.llvm.org/D110156

Files:
  llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
  llvm/test/CodeGen/AMDGPU/mad-mix-lo.ll
  llvm/test/CodeGen/AMDGPU/mad-mix.ll


Index: llvm/test/CodeGen/AMDGPU/mad-mix.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/mad-mix.ll
+++ llvm/test/CodeGen/AMDGPU/mad-mix.ll
@@ -328,8 +328,7 @@
 ; GCN-LABEL: {{^}}v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
 ; GFX900: v_mad_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1] clamp ; encoding
 ; GFX906: v_fma_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1] clamp ; encoding
-; VI: v_mac_f32_e64 v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}} clamp{{$}}
-; CI: v_mad_f32 v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}} clamp{{$}}
+; CIVI: v_mad_f32 v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}} clamp{{$}}
 define float @v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt(<2 x half> %src0, <2 x half> %src1, <2 x half> %src2) #0 {
   %src0.hi = extractelement <2 x half> %src0, i32 1
   %src1.hi = extractelement <2 x half> %src1, i32 1
Index: llvm/test/CodeGen/AMDGPU/mad-mix-lo.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/mad-mix-lo.ll
+++ llvm/test/CodeGen/AMDGPU/mad-mix-lo.ll
@@ -70,7 +70,7 @@
 ; GFX9-NEXT: v_cvt_f16_f32_e32 v0, v0
 ; GFX9-NEXT: s_setpc_b64
 
-; CIVI: v_mac_f32_e64 v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}} clamp{{$}}
+; CIVI: v_mad_f32 v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}}, v{{[0-9]}} clamp{{$}}
 define half @v_mad_mixlo_f16_f16lo_f16lo_f32_clamp_pre_cvt(half %src0, half %src1, float %src2) #0 {
   %src0.ext = fpext half %src0 to float
   %src1.ext = fpext half %src1 to float
Index: llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
+++ llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
@@ -1388,6 +1388,14 @@
   DefClamp->setImm(1);
   MRI->replaceRegWith(MI.getOperand(0).getReg(), Def->getOperand(0).getReg());
   MI.eraseFromParent();
+
+  // Use of output modifiers forces VOP3 encoding for a VOP2 mac/fmac
+  // instruction, so we might as well convert it to the more flexible VOP3-only
+  // mad/fma form.
+  MachineFunction::iterator MBBI = Def->getParent()->getIterator();
+  if (MachineInstr *NewMI = TII->convertToThreeAddress(MBBI, *Def, nullptr))
+    Def->eraseFromParent();
+
   return true;
 }
 
@@ -1526,6 +1534,14 @@
   DefOMod->setImm(OMod);
   MRI->replaceRegWith(MI.getOperand(0).getReg(), Def->getOperand(0).getReg());
   MI.eraseFromParent();
+
+  // Use of output modifiers forces VOP3 encoding for a VOP2 mac/fmac
+  // instruction, so we might as well convert it to the more flexible VOP3-only
+  // mad/fma form.
+  MachineFunction::iterator MBBI = Def->getParent()->getIterator();
+  if (MachineInstr *NewMI = TII->convertToThreeAddress(MBBI, *Def, nullptr))
+    Def->eraseFromParent();
+
   return true;
 }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D110156.374153.patch
Type: text/x-patch
Size: 2770 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210922/37452d3e/attachment.bin>