[llvm] a1df10d - AMDGPU/NFC: Add predicate for supporting ds_add_f64 (#80379)

[via llvm-commits]

Author: Konstantin Zhuravlyov
Date: 2024-02-02T08:29:25-05:00
New Revision: a1df10da59e8eb0c1d90df81bf9000d2f7869328

URL: https://github.com/llvm/llvm-project/commit/a1df10da59e8eb0c1d90df81bf9000d2f7869328
DIFF: https://github.com/llvm/llvm-project/commit/a1df10da59e8eb0c1d90df81bf9000d2f7869328.diff

LOG: AMDGPU/NFC: Add predicate for supporting ds_add_f64 (#80379)

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/AMDGPU.td
    llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
    llvm/lib/Target/AMDGPU/DSInstructions.td
    llvm/lib/Target/AMDGPU/GCNSubtarget.h

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 18c3efc7b9d46..55dbc1a803e13 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1803,6 +1803,9 @@ def HasFlatAddressSpace : Predicate<"Subtarget->hasFlatAddressSpace()">,
 def HasBufferFlatGlobalAtomicsF64 :
   Predicate<"Subtarget->hasBufferFlatGlobalAtomicsF64()">,
   AssemblerPredicate<(any_of FeatureGFX90AInsts)>;
+def HasLdsAtomicAddF64 :
+  Predicate<"Subtarget->hasLdsAtomicAddF64()">,
+  AssemblerPredicate<(any_of FeatureGFX90AInsts)>;
 
 def HasFlatGlobalInsts : Predicate<"Subtarget->hasFlatGlobalInsts()">,
   AssemblerPredicate<(all_of FeatureFlatGlobalInsts)>;

diff  --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 17ffb7ec988f0..97952de3e6a37 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1593,7 +1593,7 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
   auto &Atomic = getActionDefinitionsBuilder(G_ATOMICRMW_FADD);
   if (ST.hasLDSFPAtomicAdd()) {
     Atomic.legalFor({{S32, LocalPtr}, {S32, RegionPtr}});
-    if (ST.hasGFX90AInsts())
+    if (ST.hasLdsAtomicAddF64())
       Atomic.legalFor({{S64, LocalPtr}});
     if (ST.hasAtomicDsPkAdd16Insts())
       Atomic.legalFor({{V2S16, LocalPtr}});

diff  --git a/llvm/lib/Target/AMDGPU/DSInstructions.td b/llvm/lib/Target/AMDGPU/DSInstructions.td
index 0888fb84a22fa..9ff74d9958304 100644
--- a/llvm/lib/Target/AMDGPU/DSInstructions.td
+++ b/llvm/lib/Target/AMDGPU/DSInstructions.td
@@ -486,10 +486,10 @@ def DS_WRITE_ADDTID_B32 : DS_0A1D_NORET<"ds_write_addtid_b32">;
 
 } // End mayLoad = 0
 
-let SubtargetPredicate = isGFX90APlus in {
+let SubtargetPredicate = HasLdsAtomicAddF64 in {
   defm DS_ADD_F64     : DS_1A1D_NORET_mc_gfx9<"ds_add_f64", VReg_64>;
   defm DS_ADD_RTN_F64 : DS_1A1D_RET_mc_gfx9<"ds_add_rtn_f64", VReg_64, "ds_add_f64">;
-} // End SubtargetPredicate = isGFX90APlus
+} // End SubtargetPredicate = HasLdsAtomicAddF64
 
 let SubtargetPredicate = HasAtomicDsPkAdd16Insts in {
   defm DS_PK_ADD_F16      : DS_1A1D_NORET_mc<"ds_pk_add_f16">;
@@ -1128,7 +1128,7 @@ let SubtargetPredicate = isGFX11Plus in {
 defm : DSAtomicCmpXChg_mc<DS_CMPSTORE_RTN_B64, DS_CMPSTORE_B64, i64, "atomic_cmp_swap">;
 } // End SubtargetPredicate = isGFX11Plus
 
-let SubtargetPredicate = isGFX90APlus in {
+let SubtargetPredicate = HasLdsAtomicAddF64 in {
 def : DSAtomicRetPat<DS_ADD_RTN_F64, f64, atomic_load_fadd_local_64>;
 let AddedComplexity = 1 in
 def : DSAtomicRetPat<DS_ADD_F64, f64, atomic_load_fadd_local_noret_64>;

diff  --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index cbc5ffa8c5123..4f8eeaaf500b4 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -641,6 +641,9 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   // BUFFER/FLAT/GLOBAL_ATOMIC_ADD/MIN/MAX_F64
   bool hasBufferFlatGlobalAtomicsF64() const { return hasGFX90AInsts(); }
 
+  // DS_ADD_F64/DS_ADD_RTN_F64
+  bool hasLdsAtomicAddF64() const { return hasGFX90AInsts(); }
+
   bool hasMultiDwordFlatScratchAddressing() const {
     return getGeneration() >= GFX9;
   }