[llvm-branch-commits] [llvm] AMDGPU: Add subtarget feature for global atomic fadd denormal support (PR #96443)

[Matt Arsenault via llvm-branch-commits]

https://github.com/arsenm updated https://github.com/llvm/llvm-project/pull/96443

>From f99d34b66486a17e2fe70d372d67fbabde82d5fb Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Sun, 23 Jun 2024 16:44:08 +0200
Subject: [PATCH 1/2] AMDGPU: Add subtarget feature for global atomic fadd
 denormal support

Not sure what the behavior for gfx90a is. The SPG says it always flushes.
The instruction documentation says it does not.
---
 llvm/lib/Target/AMDGPU/AMDGPU.td      | 13 +++++++++++--
 llvm/lib/Target/AMDGPU/GCNSubtarget.h |  7 +++++++
 2 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 7ff861f5b144d..5f798b4391704 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -788,6 +788,13 @@ def FeatureFlatAtomicFaddF32Inst
   "Has flat_atomic_add_f32 instruction"
 >;
 
+def FeatureMemoryAtomicFaddF32DenormalSupport
+  : SubtargetFeature<"memory-atomic-fadd-f32-denormal-support",
+  "HasAtomicMemoryAtomicFaddF32DenormalSupport",
+  "true",
+  "global/flat/buffer atomic fadd for float supports denormal handling"
+>;
+
 def FeatureAgentScopeFineGrainedRemoteMemoryAtomics
   : SubtargetFeature<"agent-scope-fine-grained-remote-memory-atomics",
   "HasAgentScopeFineGrainedRemoteMemoryAtomics",
@@ -1425,7 +1432,8 @@ def FeatureISAVersion9_4_Common : FeatureSet<
    FeatureKernargPreload,
    FeatureAtomicFMinFMaxF64GlobalInsts,
    FeatureAtomicFMinFMaxF64FlatInsts,
-   FeatureAgentScopeFineGrainedRemoteMemoryAtomics
+   FeatureAgentScopeFineGrainedRemoteMemoryAtomics,
+   FeatureMemoryAtomicFaddF32DenormalSupport
    ]>;
 
 def FeatureISAVersion9_4_0 : FeatureSet<
@@ -1628,7 +1636,8 @@ def FeatureISAVersion12 : FeatureSet<
    FeatureVGPRSingleUseHintInsts,
    FeatureScalarDwordx3Loads,
    FeatureDPPSrc1SGPR,
-   FeatureMaxHardClauseLength32]>;
+   FeatureMaxHardClauseLength32,
+   FeatureMemoryAtomicFaddF32DenormalSupport]>;
 
 def FeatureISAVersion12_Generic: FeatureSet<
   !listconcat(FeatureISAVersion12.Features,
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index c40efbdcf7f0b..674d84422538f 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -167,6 +167,7 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   bool HasAtomicFlatPkAdd16Insts = false;
   bool HasAtomicFaddRtnInsts = false;
   bool HasAtomicFaddNoRtnInsts = false;
+  bool HasAtomicMemoryAtomicFaddF32DenormalSupport = false;
   bool HasAtomicBufferGlobalPkAddF16NoRtnInsts = false;
   bool HasAtomicBufferGlobalPkAddF16Insts = false;
   bool HasAtomicCSubNoRtnInsts = false;
@@ -872,6 +873,12 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
 
   bool hasFlatAtomicFaddF32Inst() const { return HasFlatAtomicFaddF32Inst; }
 
+  /// \return true if the target's flat, global, and buffer atomic fadd for
+  /// float supports denormal handling.
+  bool hasMemoryAtomicFaddF32DenormalSupport() const {
+    return HasAtomicMemoryAtomicFaddF32DenormalSupport;
+  }
+
   /// \return true if atomic operations targeting fine-grained memory work
   /// correctly at device scope, in allocations in host or peer PCIe device
   /// memory.

>From 45941357740eb44a7c5828bf7c534aef65448226 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Mon, 24 Jun 2024 12:10:37 +0200
Subject: [PATCH 2/2] Add to gfx11.

RDNA 3 manual says "Floating-point addition handles NAN/INF/denorm"
thought I'm not sure I trust it.
---
 llvm/lib/Target/AMDGPU/AMDGPU.td | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 5f798b4391704..0ec65f759bc35 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1545,7 +1545,8 @@ def FeatureISAVersion11_Common : FeatureSet<
    FeatureFlatAtomicFaddF32Inst,
    FeatureImageInsts,
    FeaturePackedTID,
-   FeatureVcmpxPermlaneHazard]>;
+   FeatureVcmpxPermlaneHazard,
+   FeatureMemoryAtomicFaddF32DenormalSupport]>;
 
 // There are few workarounds that need to be
 // added to all targets. This pessimizes codegen