[llvm] AMDGPU: Create InstrMapping from VGPR MFMA to equivalent AGPR instruction (PR #117102)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm updated https://github.com/llvm/llvm-project/pull/117102

>From a488c17156af20f48ae15e6b9225f52d5a1e5960 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Wed, 20 Nov 2024 18:07:04 -0800
Subject: [PATCH] AMDGPU: Create InstrMapping from VGPR MFMA to equivalent AGPR
 instruction

This provides infrastructure for a future optimization.
---
 llvm/lib/Target/AMDGPU/SIInstrInfo.h        |  5 +++++
 llvm/lib/Target/AMDGPU/SIInstrInfo.td       |  9 +++++++++
 llvm/lib/Target/AMDGPU/VOP3PInstructions.td | 12 +++++++++---
 3 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.h b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
index e55418326a4bd0..960fbb7ea15ce7 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
@@ -1568,6 +1568,11 @@ namespace AMDGPU {
   LLVM_READONLY
   int getMFMAEarlyClobberOp(uint16_t Opcode);
 
+  /// \returns Version of an MFMA instruction which uses AGPRs for srcC and
+  /// vdst, given an \p Opcode of an MFMA which uses VGPRs for srcC/vdst.
+  LLVM_READONLY
+  int getMFMASrcCVDstAGPROp(uint16_t Opcode);
+
   /// \returns v_cmpx version of a v_cmp instruction.
   LLVM_READONLY
   int getVCMPXOpFromVCMP(uint16_t Opcode);
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index afb0b2cede045c..db2e2386648315 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -3167,6 +3167,15 @@ def getMFMAEarlyClobberOp : InstrMapping {
   let ValueCols = [["0"]];
 }
 
+// Map from an mfma using VGPRs to one using AGPRs.
+def getMFMASrcCVDstAGPROp : InstrMapping {
+  let FilterClass = "MFMATable";
+  let RowFields = ["AGPROp"];
+  let ColFields = ["MFMAKind"];
+  let KeyCol = ["VGPR"];
+  let ValueCols = [["AGPR"]];
+}
+
 // Maps an v_cmp instruction to its v_cmpx equivalent.
 def getVCMPXOpFromVCMP : InstrMapping {
   let FilterClass = "VCMPVCMPXTable";
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index a487c311fe13a3..4c3f9e57ef388e 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -741,9 +741,14 @@ def VOPProfileMAI_I32_V4I32_X16     : VOPProfileMAI<VOP_V16I32_V4I32_V4I32_V16I3
 def VOPProfileMAI_I32_V4I32_X16_VCD : VOPProfileMAI<VOP_V16I32_V4I32_V4I32_V16I32,    VISrc_512_b32,  VDst_512, AVSrc_128>;
 
 
-class MFMATable <bit is_mac, string Name> {
+class MFMATable <bit is_mac, string Kind, string Name,
+                 string AGPROpName = NAME> {
   bit IsMac = is_mac;
   string FMAOp = Name;
+  string AGPROp = AGPROpName;
+
+  // Does this MFMA use "AGPR" or "VGPR" for srcC/vdst
+  string MFMAKind = Kind;
 }
 
 class MFMA_F8F6F4_WithSizeTable<int A, int B, Instruction ThisVariant, Instruction F8F8Variant> {
@@ -849,12 +854,12 @@ multiclass MAIInst<string OpName, string P, SDPatternOperator node = null_frag,
     let Constraints = !if(NoDstOverlap, "@earlyclobber $vdst", "") in {
       def _e64 : MAIInst<OpName, !cast<VOPProfileMAI>("VOPProfileMAI_" # P),
                          !if(!or(NoDstOverlap, !eq(node, null_frag)), null_frag, AgprMAIFrag<node, HasAbid, Scaled>), Scaled>,
-                 MFMATable<0, NAME # "_e64">;
+                 MFMATable<0, "AGPR", NAME # "_e64">;
 
       let OtherPredicates = [isGFX90APlus], Mnemonic = OpName in
       def _vgprcd_e64 : MAIInst<OpName # "_vgprcd", !cast<VOPProfileMAI>("VOPProfileMAI_" # P # "_VCD"),
                                 !if(!or(NoDstOverlap, !eq(node, null_frag)), null_frag, VgprMAIFrag<node, HasAbid, Scaled>), Scaled>,
-                        MFMATable<0, NAME # "_vgprcd_e64">;
+                        MFMATable<0, "VGPR", NAME # "_vgprcd_e64", NAME # "_e64">;
     }
 
     if NoDstOverlap then {
@@ -863,6 +868,7 @@ multiclass MAIInst<string OpName, string P, SDPatternOperator node = null_frag,
           Mnemonic = OpName in {
         def "_mac_e64" : MAIInst<OpName # "_mac", !cast<VOPProfileMAI>("VOPProfileMAI_" # P),
                                  !if(!eq(node, null_frag), null_frag, AgprMAIFrag<node, HasAbid, Scaled>), Scaled>,
+                         MFMATable<1, "AGPR", NAME # "_e64", NAME # "_mac_e64">;
                          MFMATable<1, NAME # "_e64">;
 
         let OtherPredicates = [isGFX90APlus] in