[AMDGPU] Introduce a new generic target `gfx9-4-generic` #115190

shiltian · 2024-11-06T18:18:55Z

This patch introduces a new generic target, gfx9-4-generic. Since it doesn’t support FP8 and XF32-related instructions, the patch includes several code reorganizations to accommodate these changes.

shiltian · 2024-11-06T18:19:15Z

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @shiltian and the rest of your teammates on Graphite

llvmbot · 2024-11-06T18:19:28Z

@llvm/pr-subscribers-mc
@llvm/pr-subscribers-clang
@llvm/pr-subscribers-objectyaml
@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-backend-amdgpu

Author: Shilei Tian (shiltian)

Changes

Patch is 145.20 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/115190.diff

33 Files Affected:

(modified) clang/include/clang/Basic/Cuda.h (+1)
(modified) clang/lib/Basic/Cuda.cpp (+1)
(modified) clang/lib/Basic/Targets/NVPTX.cpp (+1)
(modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+1)
(modified) clang/test/CodeGenOpenCL/amdgpu-features.cl (+3)
(modified) clang/test/Driver/amdgpu-macros.cl (+1)
(modified) clang/test/Driver/amdgpu-mcpu.cl (+2)
(modified) clang/test/Misc/target-invalid-cpu-note/amdgcn.c (+1)
(modified) clang/test/Misc/target-invalid-cpu-note/nvptx.c (+1)
(modified) llvm/docs/AMDGPUUsage.rst (+5)
(modified) llvm/include/llvm/BinaryFormat/ELF.h (+2-1)
(modified) llvm/include/llvm/TargetParser/TargetParser.h (+2-1)
(modified) llvm/lib/Object/ELFObjectFile.cpp (+2)
(modified) llvm/lib/ObjectYAML/ELFYAML.cpp (+1)
(modified) llvm/lib/Target/AMDGPU/AMDGPU.td (+4)
(modified) llvm/lib/Target/AMDGPU/GCNProcessors.td (+5)
(modified) llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp (+5)
(modified) llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h (+1)
(modified) llvm/lib/TargetParser/TargetParser.cpp (+5)
(modified) llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll (+443)
(modified) llvm/test/CodeGen/AMDGPU/dst-sel-hazard.mir (+1)
(modified) llvm/test/CodeGen/AMDGPU/elf-header-flags-mach.ll (+2)
(modified) llvm/test/CodeGen/AMDGPU/generic-targets-require-v6.ll (+3)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.large.mir (+1144)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.small.mir (+496)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx940.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.update.dpp.gfx90a.ll (+1)
(modified) llvm/test/CodeGen/AMDGPU/no-corresponding-integer-type.ll (+1)
(modified) llvm/test/Object/AMDGPU/elf-header-flags-mach.yaml (+7)
(modified) llvm/test/tools/llvm-objdump/ELF/AMDGPU/subtarget.ll (+6)
(modified) llvm/test/tools/llvm-readobj/ELF/AMDGPU/elf-headers.test (+3)
(modified) llvm/tools/llvm-readobj/ELFDumper.cpp (+1)

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index 7b4f435dc39f29..721e8981af6ffc 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -103,6 +103,7 @@ enum class OffloadArch {
   GFX909,
   GFX90a,
   GFX90c,
+  GFX9_4_GENERIC,
   GFX940,
   GFX941,
   GFX942,
diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index d765baef913e2f..59c932468cd891 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -121,6 +121,7 @@ static const OffloadArchToStringMap arch_names[] = {
     GFX(909),  // gfx909
     GFX(90a),  // gfx90a
     GFX(90c),  // gfx90c
+    {OffloadArch::GFX9_4_GENERIC, "gfx9-4-generic", "compute_amdgcn"},
     GFX(940),  // gfx940
     GFX(941),  // gfx941
     GFX(942),  // gfx942
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index e0bd0b096324d8..0897032c4b8546 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -205,6 +205,7 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 598b946ad88dbb..43dc0e62284602 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -2301,6 +2301,7 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/test/CodeGenOpenCL/amdgpu-features.cl b/clang/test/CodeGenOpenCL/amdgpu-features.cl
index f3473346baae5a..692f5103724342 100644
--- a/clang/test/CodeGenOpenCL/amdgpu-features.cl
+++ b/clang/test/CodeGenOpenCL/amdgpu-features.cl
@@ -56,6 +56,8 @@
 
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize64 -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103-W64 %s
 
+// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx9-4-generic -emit-llvm -o - %s | FileCheck --check-prefix=GFX9_4_Generic %s
+
 // NOCPU-NOT: "target-features"
 // NOCPU-WAVE32: "target-features"="+wavefrontsize32"
 // NOCPU-WAVE64: "target-features"="+wavefrontsize64"
@@ -85,6 +87,7 @@
 // GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX942: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX9_4_Generic: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
diff --git a/clang/test/Driver/amdgpu-macros.cl b/clang/test/Driver/amdgpu-macros.cl
index dd5a4483e4d607..9352c59ef28b7f 100644
--- a/clang/test/Driver/amdgpu-macros.cl
+++ b/clang/test/Driver/amdgpu-macros.cl
@@ -133,6 +133,7 @@
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1201 -DFAMILY=GFX12
 
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_1_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_3_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx11_generic -DFAMILY=GFX11
diff --git a/clang/test/Driver/amdgpu-mcpu.cl b/clang/test/Driver/amdgpu-mcpu.cl
index 42ce33db6eec07..3c090b0254e782 100644
--- a/clang/test/Driver/amdgpu-mcpu.cl
+++ b/clang/test/Driver/amdgpu-mcpu.cl
@@ -118,6 +118,7 @@
 // RUN: %clang -### -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefix=GFX1201 %s
 
 // RUN: %clang -### -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefix=GFX9_GENERIC %s
+// RUN: %clang -### -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefix=GFX9_4_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefix=GFX10_1_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefix=GFX10_3_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefix=GFX11_GENERIC %s
@@ -172,6 +173,7 @@
 // GFX1201:   "-target-cpu" "gfx1201"
 
 // GFX9_GENERIC:      "-target-cpu" "gfx9-generic"
+// GFX9_4_GENERIC:      "-target-cpu" "gfx9-4-generic"
 // GFX10_1_GENERIC:   "-target-cpu" "gfx10-1-generic"
 // GFX10_3_GENERIC:   "-target-cpu" "gfx10-3-generic"
 // GFX11_GENERIC:     "-target-cpu" "gfx11-generic"
diff --git a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
index b3ddbd53a0391b..cae31797c1febe 100644
--- a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
+++ b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
@@ -70,6 +70,7 @@
 // CHECK-SAME: {{^}}, gfx1200
 // CHECK-SAME: {{^}}, gfx1201
 // CHECK-SAME: {{^}}, gfx9-generic
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx10-1-generic
 // CHECK-SAME: {{^}}, gfx10-3-generic
 // CHECK-SAME: {{^}}, gfx11-generic
diff --git a/clang/test/Misc/target-invalid-cpu-note/nvptx.c b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
index a59e1c6fab1c49..44fe07065b2428 100644
--- a/clang/test/Misc/target-invalid-cpu-note/nvptx.c
+++ b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
@@ -50,6 +50,7 @@
 // CHECK-SAME: {{^}}, gfx909
 // CHECK-SAME: {{^}}, gfx90a
 // CHECK-SAME: {{^}}, gfx90c
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx940
 // CHECK-SAME: {{^}}, gfx941
 // CHECK-SAME: {{^}}, gfx942
diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 5b83ea428c0bff..92d072bec6b57d 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -576,6 +576,11 @@ Generic processor code objects are versioned. See :ref:`amdgpu-generic-processor
                                                                                                   - ``v_dot2_f32_f16``
 
 
+     ``gfx9-4-generic``   ``amdgcn``     - ``gfx940``      - xnack            - Absolute flat   No restrictions.
+                                         - ``gfx941``                           scratch
+                                         - ``gfx942``
+
+
      ``gfx10-1-generic``  ``amdgcn``     - ``gfx1010``     - xnack            - Absolute flat   - The following instructions are
                                          - ``gfx1011``     - wavefrontsize64    scratch           not available on ``gfx1011``
                                          - ``gfx1012``     - cumode                               and ``gfx1012``
diff --git a/llvm/include/llvm/BinaryFormat/ELF.h b/llvm/include/llvm/BinaryFormat/ELF.h
index c591a96232f115..6c05ea7208e1f1 100644
--- a/llvm/include/llvm/BinaryFormat/ELF.h
+++ b/llvm/include/llvm/BinaryFormat/ELF.h
@@ -822,11 +822,12 @@ enum : unsigned {
   EF_AMDGPU_MACH_AMDGCN_RESERVED_0X57   = 0x057,
   EF_AMDGPU_MACH_AMDGCN_GFX1153         = 0x058,
   EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC   = 0x059,
+  EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC  = 0x05f,
   // clang-format on
 
   // First/last AMDGCN-based processors.
   EF_AMDGPU_MACH_AMDGCN_FIRST = EF_AMDGPU_MACH_AMDGCN_GFX600,
-  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC,
+  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC,
 
   // Indicates if the "xnack" target feature is enabled for all code contained
   // in the object.
diff --git a/llvm/include/llvm/TargetParser/TargetParser.h b/llvm/include/llvm/TargetParser/TargetParser.h
index ae86ff39083d89..c6db4dfd7f5159 100644
--- a/llvm/include/llvm/TargetParser/TargetParser.h
+++ b/llvm/include/llvm/TargetParser/TargetParser.h
@@ -119,9 +119,10 @@ enum GPUKind : uint32_t {
   GK_GFX10_3_GENERIC = 194,
   GK_GFX11_GENERIC = 195,
   GK_GFX12_GENERIC = 196,
+  GK_GFX9_4_GENERIC = 197,
 
   GK_AMDGCN_GENERIC_FIRST = GK_GFX9_GENERIC,
-  GK_AMDGCN_GENERIC_LAST = GK_GFX12_GENERIC,
+  GK_AMDGCN_GENERIC_LAST = GK_GFX9_4_GENERIC,
 };
 
 /// Instruction set architecture version.
diff --git a/llvm/lib/Object/ELFObjectFile.cpp b/llvm/lib/Object/ELFObjectFile.cpp
index 5096877d2a4b00..9dc39936ffd8bb 100644
--- a/llvm/lib/Object/ELFObjectFile.cpp
+++ b/llvm/lib/Object/ELFObjectFile.cpp
@@ -602,6 +602,8 @@ StringRef ELFObjectFileBase::getAMDGPUCPUName() const {
   // Generic AMDGCN targets
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:
     return "gfx9-generic";
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:
+    return "gfx9-4-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:
     return "gfx10-1-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:
diff --git a/llvm/lib/ObjectYAML/ELFYAML.cpp b/llvm/lib/ObjectYAML/ELFYAML.cpp
index e97248cbcf5682..76c6c496c5e6bc 100644
--- a/llvm/lib/ObjectYAML/ELFYAML.cpp
+++ b/llvm/lib/ObjectYAML/ELFYAML.cpp
@@ -631,6 +631,7 @@ void ScalarBitSetTraits<ELFYAML::ELF_EF>::bitset(IO &IO,
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1200, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1201, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC, EF_AMDGPU_MACH);
+    BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC, EF_AMDGPU_MACH);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 910f5e06a6f3c4..2f8d6f1c7f8065 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1478,6 +1478,10 @@ def FeatureISAVersion9_4_2 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
     [])>;
 
+def FeatureISAVersion9_4_Generic : FeatureSet<
+  !listconcat(FeatureISAVersion9_4_Common.Features,
+    [FeatureRequiresCOV6])>;
+
 def FeatureISAVersion10_Common : FeatureSet<
   [FeatureGFX10,
    FeatureLDSBankCount32,
diff --git a/llvm/lib/Target/AMDGPU/GCNProcessors.td b/llvm/lib/Target/AMDGPU/GCNProcessors.td
index 547941633fda61..e12a6127b17063 100644
--- a/llvm/lib/Target/AMDGPU/GCNProcessors.td
+++ b/llvm/lib/Target/AMDGPU/GCNProcessors.td
@@ -209,6 +209,11 @@ def : ProcessorModel<"gfx9-generic", SIQuarterSpeedModel,
   FeatureISAVersion9_Generic.Features
 >;
 
+// [gfx940, gfx941, gfx942]
+def : ProcessorModel<"gfx9-4-generic", SIQuarterSpeedModel,
+  FeatureISAVersion9_4_Generic.Features
+>;
+
 //===----------------------------------------------------------------------===//
 // GCN GFX10.
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
index 5c625c3d83ff1b..507725b91a9bee 100644
--- a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
@@ -119,6 +119,7 @@ StringRef AMDGPUTargetStreamer::getArchNameFromElfMach(unsigned ElfMach) {
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200: AK = GK_GFX1200; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201: AK = GK_GFX1201; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:     AK = GK_GFX9_GENERIC; break;
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:   AK = GK_GFX9_4_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:  AK = GK_GFX10_1_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:  AK = GK_GFX10_3_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC:    AK = GK_GFX11_GENERIC; break;
@@ -204,6 +205,7 @@ unsigned AMDGPUTargetStreamer::getElfMach(StringRef GPU) {
   case GK_GFX1200: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200;
   case GK_GFX1201: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201;
   case GK_GFX9_GENERIC:     return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC;
+  case GK_GFX9_4_GENERIC:   return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC;
   case GK_GFX10_1_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC;
   case GK_GFX10_3_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC;
   case GK_GFX11_GENERIC:    return ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC;
@@ -821,6 +823,9 @@ unsigned AMDGPUTargetELFStreamer::getEFlagsV6() {
     case AMDGPU::GK_GFX9_GENERIC:
       Version = GenericVersion::GFX9;
       break;
+    case AMDGPU::GK_GFX9_4_GENERIC:
+      Version = GenericVersion::GFX9_4;
+      break;
     case AMDGPU::GK_GFX10_1_GENERIC:
       Version = GenericVersion::GFX10_1;
       break;
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index beebe320b2cf3a..88a6d75b72c7d0 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -46,6 +46,7 @@ struct IsaVersion;
 /// within a generic family.
 namespace GenericVersion {
 static constexpr unsigned GFX9 = 1;
+static constexpr unsigned GFX9_4 = 1;
 static constexpr unsigned GFX10_1 = 1;
 static constexpr unsigned GFX10_3 = 1;
 static constexpr unsigned GFX11 = 1;
diff --git a/llvm/lib/TargetParser/TargetParser.cpp b/llvm/lib/TargetParser/TargetParser.cpp
index 059d7b4f5ff2d0..e7a979eaedb75c 100644
--- a/llvm/lib/TargetParser/TargetParser.cpp
+++ b/llvm/lib/TargetParser/TargetParser.cpp
@@ -130,6 +130,7 @@ constexpr GPUInfo AMDGCNGPUs[] = {
     {{"gfx1201"},   {"gfx1201"}, GK_GFX1201, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
 
     {{"gfx9-generic"},      {"gfx9-generic"},    GK_GFX9_GENERIC,    FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK},
+    {{"gfx9-4-generic"},    {"gfx9-4-generic"},  GK_GFX9_4_GENERIC,  FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK|FEATURE_SRAMECC},
     {{"gfx10-1-generic"},   {"gfx10-1-generic"}, GK_GFX10_1_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_XNACK|FEATURE_WGP},
     {{"gfx10-3-generic"},   {"gfx10-3-generic"}, GK_GFX10_3_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
     {{"gfx11-generic"},     {"gfx11-generic"},   GK_GFX11_GENERIC,   FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
@@ -156,6 +157,8 @@ StringRef llvm::AMDGPU::getArchFamilyNameAMDGCN(GPUKind AK) {
   switch (AK) {
   case AMDGPU::GK_GFX9_GENERIC:
     return "gfx9";
+  case AMDGPU::GK_GFX9_4_GENERIC:
+    return "gfx9";
   case AMDGPU::GK_GFX10_1_GENERIC:
   case AMDGPU::GK_GFX10_3_GENERIC:
     return "gfx10";
@@ -296,6 +299,7 @@ AMDGPU::IsaVersion AMDGPU::getIsaVersion(StringRef GPU) {
   // TODO: Split up this API depending on its caller so
   // generic target handling is more obvious and less risky.
   case GK_GFX9_GENERIC:    return {9, 0, 0};
+  case GK_GFX9_4_GENERIC:  return {9, 4, 0};
   case GK_GFX10_1_GENERIC: return {10, 1, 0};
   case GK_GFX10_3_GENERIC: return {10, 3, 0};
   case GK_GFX11_GENERIC:   return {11, 0, 3};
@@ -466,6 +470,7 @@ void AMDGPU::fillAMDGPUFeatureMap(StringRef GPU, const Triple &T,
     case GK_GFX942:
     case GK_GFX941:
     case GK_GFX940:
+    case GK_GFX9_4_GENERIC:
       Features["gfx940-insts"] = true;
       Features["fp8-insts"] = true;
       Features["fp8-conversion-insts"] = true;
diff --git a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
index 5986d2d38ef1ad..4eac26e853c2a0 100644
--- a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
+++ b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
@@ -112,6 +112,8 @@
 
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_XNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_NOXNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-3-generic < %s | FileCheck --check-prefixes=GFX10_3_GENERIC %s
@@ -210,6 +212,8 @@
 
 ; GFX9_GENERIC_NOXNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack-"
 ; GFX9_GENERIC_XNACK:       .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack+"
+; GFX9_4_GENERIC_NOXNACK:   .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack-"
+; GFX9_4_GENERIC_XNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack+"
 ; GFX10_1_GENERIC_NOXNACK:  .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack-"
 ; GFX10_1_GENERIC_XNACK:    .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack+"
 ; GFX10_3_GENERIC:          .amdgcn_target "amdgcn-amd-amdhsa--gfx10-3-generic"
diff --git a/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll b/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll
index 662de47413654f..10ef4dc280d091 100644
--- a/llvm...
[truncated]

llvmbot · 2024-11-06T18:19:28Z

@llvm/pr-subscribers-clang-driver

Author: Shilei Tian (shiltian)

Changes

Patch is 145.20 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/115190.diff

33 Files Affected:

(modified) clang/include/clang/Basic/Cuda.h (+1)
(modified) clang/lib/Basic/Cuda.cpp (+1)
(modified) clang/lib/Basic/Targets/NVPTX.cpp (+1)
(modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+1)
(modified) clang/test/CodeGenOpenCL/amdgpu-features.cl (+3)
(modified) clang/test/Driver/amdgpu-macros.cl (+1)
(modified) clang/test/Driver/amdgpu-mcpu.cl (+2)
(modified) clang/test/Misc/target-invalid-cpu-note/amdgcn.c (+1)
(modified) clang/test/Misc/target-invalid-cpu-note/nvptx.c (+1)
(modified) llvm/docs/AMDGPUUsage.rst (+5)
(modified) llvm/include/llvm/BinaryFormat/ELF.h (+2-1)
(modified) llvm/include/llvm/TargetParser/TargetParser.h (+2-1)
(modified) llvm/lib/Object/ELFObjectFile.cpp (+2)
(modified) llvm/lib/ObjectYAML/ELFYAML.cpp (+1)
(modified) llvm/lib/Target/AMDGPU/AMDGPU.td (+4)
(modified) llvm/lib/Target/AMDGPU/GCNProcessors.td (+5)
(modified) llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp (+5)
(modified) llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h (+1)
(modified) llvm/lib/TargetParser/TargetParser.cpp (+5)
(modified) llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll (+443)
(modified) llvm/test/CodeGen/AMDGPU/dst-sel-hazard.mir (+1)
(modified) llvm/test/CodeGen/AMDGPU/elf-header-flags-mach.ll (+2)
(modified) llvm/test/CodeGen/AMDGPU/generic-targets-require-v6.ll (+3)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.large.mir (+1144)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.small.mir (+496)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx940.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.update.dpp.gfx90a.ll (+1)
(modified) llvm/test/CodeGen/AMDGPU/no-corresponding-integer-type.ll (+1)
(modified) llvm/test/Object/AMDGPU/elf-header-flags-mach.yaml (+7)
(modified) llvm/test/tools/llvm-objdump/ELF/AMDGPU/subtarget.ll (+6)
(modified) llvm/test/tools/llvm-readobj/ELF/AMDGPU/elf-headers.test (+3)
(modified) llvm/tools/llvm-readobj/ELFDumper.cpp (+1)

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index 7b4f435dc39f29..721e8981af6ffc 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -103,6 +103,7 @@ enum class OffloadArch {
   GFX909,
   GFX90a,
   GFX90c,
+  GFX9_4_GENERIC,
   GFX940,
   GFX941,
   GFX942,
diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index d765baef913e2f..59c932468cd891 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -121,6 +121,7 @@ static const OffloadArchToStringMap arch_names[] = {
     GFX(909),  // gfx909
     GFX(90a),  // gfx90a
     GFX(90c),  // gfx90c
+    {OffloadArch::GFX9_4_GENERIC, "gfx9-4-generic", "compute_amdgcn"},
     GFX(940),  // gfx940
     GFX(941),  // gfx941
     GFX(942),  // gfx942
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index e0bd0b096324d8..0897032c4b8546 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -205,6 +205,7 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 598b946ad88dbb..43dc0e62284602 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -2301,6 +2301,7 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/test/CodeGenOpenCL/amdgpu-features.cl b/clang/test/CodeGenOpenCL/amdgpu-features.cl
index f3473346baae5a..692f5103724342 100644
--- a/clang/test/CodeGenOpenCL/amdgpu-features.cl
+++ b/clang/test/CodeGenOpenCL/amdgpu-features.cl
@@ -56,6 +56,8 @@
 
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize64 -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103-W64 %s
 
+// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx9-4-generic -emit-llvm -o - %s | FileCheck --check-prefix=GFX9_4_Generic %s
+
 // NOCPU-NOT: "target-features"
 // NOCPU-WAVE32: "target-features"="+wavefrontsize32"
 // NOCPU-WAVE64: "target-features"="+wavefrontsize64"
@@ -85,6 +87,7 @@
 // GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX942: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX9_4_Generic: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
diff --git a/clang/test/Driver/amdgpu-macros.cl b/clang/test/Driver/amdgpu-macros.cl
index dd5a4483e4d607..9352c59ef28b7f 100644
--- a/clang/test/Driver/amdgpu-macros.cl
+++ b/clang/test/Driver/amdgpu-macros.cl
@@ -133,6 +133,7 @@
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1201 -DFAMILY=GFX12
 
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_1_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_3_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx11_generic -DFAMILY=GFX11
diff --git a/clang/test/Driver/amdgpu-mcpu.cl b/clang/test/Driver/amdgpu-mcpu.cl
index 42ce33db6eec07..3c090b0254e782 100644
--- a/clang/test/Driver/amdgpu-mcpu.cl
+++ b/clang/test/Driver/amdgpu-mcpu.cl
@@ -118,6 +118,7 @@
 // RUN: %clang -### -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefix=GFX1201 %s
 
 // RUN: %clang -### -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefix=GFX9_GENERIC %s
+// RUN: %clang -### -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefix=GFX9_4_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefix=GFX10_1_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefix=GFX10_3_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefix=GFX11_GENERIC %s
@@ -172,6 +173,7 @@
 // GFX1201:   "-target-cpu" "gfx1201"
 
 // GFX9_GENERIC:      "-target-cpu" "gfx9-generic"
+// GFX9_4_GENERIC:      "-target-cpu" "gfx9-4-generic"
 // GFX10_1_GENERIC:   "-target-cpu" "gfx10-1-generic"
 // GFX10_3_GENERIC:   "-target-cpu" "gfx10-3-generic"
 // GFX11_GENERIC:     "-target-cpu" "gfx11-generic"
diff --git a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
index b3ddbd53a0391b..cae31797c1febe 100644
--- a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
+++ b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
@@ -70,6 +70,7 @@
 // CHECK-SAME: {{^}}, gfx1200
 // CHECK-SAME: {{^}}, gfx1201
 // CHECK-SAME: {{^}}, gfx9-generic
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx10-1-generic
 // CHECK-SAME: {{^}}, gfx10-3-generic
 // CHECK-SAME: {{^}}, gfx11-generic
diff --git a/clang/test/Misc/target-invalid-cpu-note/nvptx.c b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
index a59e1c6fab1c49..44fe07065b2428 100644
--- a/clang/test/Misc/target-invalid-cpu-note/nvptx.c
+++ b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
@@ -50,6 +50,7 @@
 // CHECK-SAME: {{^}}, gfx909
 // CHECK-SAME: {{^}}, gfx90a
 // CHECK-SAME: {{^}}, gfx90c
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx940
 // CHECK-SAME: {{^}}, gfx941
 // CHECK-SAME: {{^}}, gfx942
diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 5b83ea428c0bff..92d072bec6b57d 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -576,6 +576,11 @@ Generic processor code objects are versioned. See :ref:`amdgpu-generic-processor
                                                                                                   - ``v_dot2_f32_f16``
 
 
+     ``gfx9-4-generic``   ``amdgcn``     - ``gfx940``      - xnack            - Absolute flat   No restrictions.
+                                         - ``gfx941``                           scratch
+                                         - ``gfx942``
+
+
      ``gfx10-1-generic``  ``amdgcn``     - ``gfx1010``     - xnack            - Absolute flat   - The following instructions are
                                          - ``gfx1011``     - wavefrontsize64    scratch           not available on ``gfx1011``
                                          - ``gfx1012``     - cumode                               and ``gfx1012``
diff --git a/llvm/include/llvm/BinaryFormat/ELF.h b/llvm/include/llvm/BinaryFormat/ELF.h
index c591a96232f115..6c05ea7208e1f1 100644
--- a/llvm/include/llvm/BinaryFormat/ELF.h
+++ b/llvm/include/llvm/BinaryFormat/ELF.h
@@ -822,11 +822,12 @@ enum : unsigned {
   EF_AMDGPU_MACH_AMDGCN_RESERVED_0X57   = 0x057,
   EF_AMDGPU_MACH_AMDGCN_GFX1153         = 0x058,
   EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC   = 0x059,
+  EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC  = 0x05f,
   // clang-format on
 
   // First/last AMDGCN-based processors.
   EF_AMDGPU_MACH_AMDGCN_FIRST = EF_AMDGPU_MACH_AMDGCN_GFX600,
-  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC,
+  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC,
 
   // Indicates if the "xnack" target feature is enabled for all code contained
   // in the object.
diff --git a/llvm/include/llvm/TargetParser/TargetParser.h b/llvm/include/llvm/TargetParser/TargetParser.h
index ae86ff39083d89..c6db4dfd7f5159 100644
--- a/llvm/include/llvm/TargetParser/TargetParser.h
+++ b/llvm/include/llvm/TargetParser/TargetParser.h
@@ -119,9 +119,10 @@ enum GPUKind : uint32_t {
   GK_GFX10_3_GENERIC = 194,
   GK_GFX11_GENERIC = 195,
   GK_GFX12_GENERIC = 196,
+  GK_GFX9_4_GENERIC = 197,
 
   GK_AMDGCN_GENERIC_FIRST = GK_GFX9_GENERIC,
-  GK_AMDGCN_GENERIC_LAST = GK_GFX12_GENERIC,
+  GK_AMDGCN_GENERIC_LAST = GK_GFX9_4_GENERIC,
 };
 
 /// Instruction set architecture version.
diff --git a/llvm/lib/Object/ELFObjectFile.cpp b/llvm/lib/Object/ELFObjectFile.cpp
index 5096877d2a4b00..9dc39936ffd8bb 100644
--- a/llvm/lib/Object/ELFObjectFile.cpp
+++ b/llvm/lib/Object/ELFObjectFile.cpp
@@ -602,6 +602,8 @@ StringRef ELFObjectFileBase::getAMDGPUCPUName() const {
   // Generic AMDGCN targets
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:
     return "gfx9-generic";
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:
+    return "gfx9-4-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:
     return "gfx10-1-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:
diff --git a/llvm/lib/ObjectYAML/ELFYAML.cpp b/llvm/lib/ObjectYAML/ELFYAML.cpp
index e97248cbcf5682..76c6c496c5e6bc 100644
--- a/llvm/lib/ObjectYAML/ELFYAML.cpp
+++ b/llvm/lib/ObjectYAML/ELFYAML.cpp
@@ -631,6 +631,7 @@ void ScalarBitSetTraits<ELFYAML::ELF_EF>::bitset(IO &IO,
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1200, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1201, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC, EF_AMDGPU_MACH);
+    BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC, EF_AMDGPU_MACH);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 910f5e06a6f3c4..2f8d6f1c7f8065 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1478,6 +1478,10 @@ def FeatureISAVersion9_4_2 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
     [])>;
 
+def FeatureISAVersion9_4_Generic : FeatureSet<
+  !listconcat(FeatureISAVersion9_4_Common.Features,
+    [FeatureRequiresCOV6])>;
+
 def FeatureISAVersion10_Common : FeatureSet<
   [FeatureGFX10,
    FeatureLDSBankCount32,
diff --git a/llvm/lib/Target/AMDGPU/GCNProcessors.td b/llvm/lib/Target/AMDGPU/GCNProcessors.td
index 547941633fda61..e12a6127b17063 100644
--- a/llvm/lib/Target/AMDGPU/GCNProcessors.td
+++ b/llvm/lib/Target/AMDGPU/GCNProcessors.td
@@ -209,6 +209,11 @@ def : ProcessorModel<"gfx9-generic", SIQuarterSpeedModel,
   FeatureISAVersion9_Generic.Features
 >;
 
+// [gfx940, gfx941, gfx942]
+def : ProcessorModel<"gfx9-4-generic", SIQuarterSpeedModel,
+  FeatureISAVersion9_4_Generic.Features
+>;
+
 //===----------------------------------------------------------------------===//
 // GCN GFX10.
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
index 5c625c3d83ff1b..507725b91a9bee 100644
--- a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
@@ -119,6 +119,7 @@ StringRef AMDGPUTargetStreamer::getArchNameFromElfMach(unsigned ElfMach) {
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200: AK = GK_GFX1200; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201: AK = GK_GFX1201; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:     AK = GK_GFX9_GENERIC; break;
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:   AK = GK_GFX9_4_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:  AK = GK_GFX10_1_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:  AK = GK_GFX10_3_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC:    AK = GK_GFX11_GENERIC; break;
@@ -204,6 +205,7 @@ unsigned AMDGPUTargetStreamer::getElfMach(StringRef GPU) {
   case GK_GFX1200: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200;
   case GK_GFX1201: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201;
   case GK_GFX9_GENERIC:     return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC;
+  case GK_GFX9_4_GENERIC:   return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC;
   case GK_GFX10_1_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC;
   case GK_GFX10_3_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC;
   case GK_GFX11_GENERIC:    return ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC;
@@ -821,6 +823,9 @@ unsigned AMDGPUTargetELFStreamer::getEFlagsV6() {
     case AMDGPU::GK_GFX9_GENERIC:
       Version = GenericVersion::GFX9;
       break;
+    case AMDGPU::GK_GFX9_4_GENERIC:
+      Version = GenericVersion::GFX9_4;
+      break;
     case AMDGPU::GK_GFX10_1_GENERIC:
       Version = GenericVersion::GFX10_1;
       break;
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index beebe320b2cf3a..88a6d75b72c7d0 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -46,6 +46,7 @@ struct IsaVersion;
 /// within a generic family.
 namespace GenericVersion {
 static constexpr unsigned GFX9 = 1;
+static constexpr unsigned GFX9_4 = 1;
 static constexpr unsigned GFX10_1 = 1;
 static constexpr unsigned GFX10_3 = 1;
 static constexpr unsigned GFX11 = 1;
diff --git a/llvm/lib/TargetParser/TargetParser.cpp b/llvm/lib/TargetParser/TargetParser.cpp
index 059d7b4f5ff2d0..e7a979eaedb75c 100644
--- a/llvm/lib/TargetParser/TargetParser.cpp
+++ b/llvm/lib/TargetParser/TargetParser.cpp
@@ -130,6 +130,7 @@ constexpr GPUInfo AMDGCNGPUs[] = {
     {{"gfx1201"},   {"gfx1201"}, GK_GFX1201, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
 
     {{"gfx9-generic"},      {"gfx9-generic"},    GK_GFX9_GENERIC,    FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK},
+    {{"gfx9-4-generic"},    {"gfx9-4-generic"},  GK_GFX9_4_GENERIC,  FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK|FEATURE_SRAMECC},
     {{"gfx10-1-generic"},   {"gfx10-1-generic"}, GK_GFX10_1_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_XNACK|FEATURE_WGP},
     {{"gfx10-3-generic"},   {"gfx10-3-generic"}, GK_GFX10_3_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
     {{"gfx11-generic"},     {"gfx11-generic"},   GK_GFX11_GENERIC,   FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
@@ -156,6 +157,8 @@ StringRef llvm::AMDGPU::getArchFamilyNameAMDGCN(GPUKind AK) {
   switch (AK) {
   case AMDGPU::GK_GFX9_GENERIC:
     return "gfx9";
+  case AMDGPU::GK_GFX9_4_GENERIC:
+    return "gfx9";
   case AMDGPU::GK_GFX10_1_GENERIC:
   case AMDGPU::GK_GFX10_3_GENERIC:
     return "gfx10";
@@ -296,6 +299,7 @@ AMDGPU::IsaVersion AMDGPU::getIsaVersion(StringRef GPU) {
   // TODO: Split up this API depending on its caller so
   // generic target handling is more obvious and less risky.
   case GK_GFX9_GENERIC:    return {9, 0, 0};
+  case GK_GFX9_4_GENERIC:  return {9, 4, 0};
   case GK_GFX10_1_GENERIC: return {10, 1, 0};
   case GK_GFX10_3_GENERIC: return {10, 3, 0};
   case GK_GFX11_GENERIC:   return {11, 0, 3};
@@ -466,6 +470,7 @@ void AMDGPU::fillAMDGPUFeatureMap(StringRef GPU, const Triple &T,
     case GK_GFX942:
     case GK_GFX941:
     case GK_GFX940:
+    case GK_GFX9_4_GENERIC:
       Features["gfx940-insts"] = true;
       Features["fp8-insts"] = true;
       Features["fp8-conversion-insts"] = true;
diff --git a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
index 5986d2d38ef1ad..4eac26e853c2a0 100644
--- a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
+++ b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
@@ -112,6 +112,8 @@
 
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_XNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_NOXNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-3-generic < %s | FileCheck --check-prefixes=GFX10_3_GENERIC %s
@@ -210,6 +212,8 @@
 
 ; GFX9_GENERIC_NOXNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack-"
 ; GFX9_GENERIC_XNACK:       .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack+"
+; GFX9_4_GENERIC_NOXNACK:   .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack-"
+; GFX9_4_GENERIC_XNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack+"
 ; GFX10_1_GENERIC_NOXNACK:  .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack-"
 ; GFX10_1_GENERIC_XNACK:    .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack+"
 ; GFX10_3_GENERIC:          .amdgcn_target "amdgcn-amd-amdhsa--gfx10-3-generic"
diff --git a/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll b/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll
index 662de47413654f..10ef4dc280d091 100644
--- a/llvm...
[truncated]

llvmbot · 2024-11-06T18:19:29Z

@llvm/pr-subscribers-llvm-binary-utilities

Author: Shilei Tian (shiltian)

Changes

Patch is 145.20 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/115190.diff

33 Files Affected:

(modified) clang/include/clang/Basic/Cuda.h (+1)
(modified) clang/lib/Basic/Cuda.cpp (+1)
(modified) clang/lib/Basic/Targets/NVPTX.cpp (+1)
(modified) clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp (+1)
(modified) clang/test/CodeGenOpenCL/amdgpu-features.cl (+3)
(modified) clang/test/Driver/amdgpu-macros.cl (+1)
(modified) clang/test/Driver/amdgpu-mcpu.cl (+2)
(modified) clang/test/Misc/target-invalid-cpu-note/amdgcn.c (+1)
(modified) clang/test/Misc/target-invalid-cpu-note/nvptx.c (+1)
(modified) llvm/docs/AMDGPUUsage.rst (+5)
(modified) llvm/include/llvm/BinaryFormat/ELF.h (+2-1)
(modified) llvm/include/llvm/TargetParser/TargetParser.h (+2-1)
(modified) llvm/lib/Object/ELFObjectFile.cpp (+2)
(modified) llvm/lib/ObjectYAML/ELFYAML.cpp (+1)
(modified) llvm/lib/Target/AMDGPU/AMDGPU.td (+4)
(modified) llvm/lib/Target/AMDGPU/GCNProcessors.td (+5)
(modified) llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp (+5)
(modified) llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h (+1)
(modified) llvm/lib/TargetParser/TargetParser.cpp (+5)
(modified) llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll (+443)
(modified) llvm/test/CodeGen/AMDGPU/dst-sel-hazard.mir (+1)
(modified) llvm/test/CodeGen/AMDGPU/elf-header-flags-mach.ll (+2)
(modified) llvm/test/CodeGen/AMDGPU/generic-targets-require-v6.ll (+3)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.large.mir (+1144)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.small.mir (+496)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx940.ll (+4)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.update.dpp.gfx90a.ll (+1)
(modified) llvm/test/CodeGen/AMDGPU/no-corresponding-integer-type.ll (+1)
(modified) llvm/test/Object/AMDGPU/elf-header-flags-mach.yaml (+7)
(modified) llvm/test/tools/llvm-objdump/ELF/AMDGPU/subtarget.ll (+6)
(modified) llvm/test/tools/llvm-readobj/ELF/AMDGPU/elf-headers.test (+3)
(modified) llvm/tools/llvm-readobj/ELFDumper.cpp (+1)

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index 7b4f435dc39f29..721e8981af6ffc 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -103,6 +103,7 @@ enum class OffloadArch {
   GFX909,
   GFX90a,
   GFX90c,
+  GFX9_4_GENERIC,
   GFX940,
   GFX941,
   GFX942,
diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index d765baef913e2f..59c932468cd891 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -121,6 +121,7 @@ static const OffloadArchToStringMap arch_names[] = {
     GFX(909),  // gfx909
     GFX(90a),  // gfx90a
     GFX(90c),  // gfx90c
+    {OffloadArch::GFX9_4_GENERIC, "gfx9-4-generic", "compute_amdgcn"},
     GFX(940),  // gfx940
     GFX(941),  // gfx941
     GFX(942),  // gfx942
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index e0bd0b096324d8..0897032c4b8546 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -205,6 +205,7 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 598b946ad88dbb..43dc0e62284602 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -2301,6 +2301,7 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
       case OffloadArch::GFX909:
       case OffloadArch::GFX90a:
       case OffloadArch::GFX90c:
+      case OffloadArch::GFX9_4_GENERIC:
       case OffloadArch::GFX940:
       case OffloadArch::GFX941:
       case OffloadArch::GFX942:
diff --git a/clang/test/CodeGenOpenCL/amdgpu-features.cl b/clang/test/CodeGenOpenCL/amdgpu-features.cl
index f3473346baae5a..692f5103724342 100644
--- a/clang/test/CodeGenOpenCL/amdgpu-features.cl
+++ b/clang/test/CodeGenOpenCL/amdgpu-features.cl
@@ -56,6 +56,8 @@
 
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize64 -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103-W64 %s
 
+// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx9-4-generic -emit-llvm -o - %s | FileCheck --check-prefix=GFX9_4_Generic %s
+
 // NOCPU-NOT: "target-features"
 // NOCPU-WAVE32: "target-features"="+wavefrontsize32"
 // NOCPU-WAVE64: "target-features"="+wavefrontsize64"
@@ -85,6 +87,7 @@
 // GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX942: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX9_4_Generic: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
diff --git a/clang/test/Driver/amdgpu-macros.cl b/clang/test/Driver/amdgpu-macros.cl
index dd5a4483e4d607..9352c59ef28b7f 100644
--- a/clang/test/Driver/amdgpu-macros.cl
+++ b/clang/test/Driver/amdgpu-macros.cl
@@ -133,6 +133,7 @@
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1201 -DFAMILY=GFX12
 
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx9_generic -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_1_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx10_3_generic -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx11_generic -DFAMILY=GFX11
diff --git a/clang/test/Driver/amdgpu-mcpu.cl b/clang/test/Driver/amdgpu-mcpu.cl
index 42ce33db6eec07..3c090b0254e782 100644
--- a/clang/test/Driver/amdgpu-mcpu.cl
+++ b/clang/test/Driver/amdgpu-mcpu.cl
@@ -118,6 +118,7 @@
 // RUN: %clang -### -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefix=GFX1201 %s
 
 // RUN: %clang -### -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefix=GFX9_GENERIC %s
+// RUN: %clang -### -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefix=GFX9_4_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefix=GFX10_1_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefix=GFX10_3_GENERIC %s
 // RUN: %clang -### -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefix=GFX11_GENERIC %s
@@ -172,6 +173,7 @@
 // GFX1201:   "-target-cpu" "gfx1201"
 
 // GFX9_GENERIC:      "-target-cpu" "gfx9-generic"
+// GFX9_4_GENERIC:      "-target-cpu" "gfx9-4-generic"
 // GFX10_1_GENERIC:   "-target-cpu" "gfx10-1-generic"
 // GFX10_3_GENERIC:   "-target-cpu" "gfx10-3-generic"
 // GFX11_GENERIC:     "-target-cpu" "gfx11-generic"
diff --git a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
index b3ddbd53a0391b..cae31797c1febe 100644
--- a/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
+++ b/clang/test/Misc/target-invalid-cpu-note/amdgcn.c
@@ -70,6 +70,7 @@
 // CHECK-SAME: {{^}}, gfx1200
 // CHECK-SAME: {{^}}, gfx1201
 // CHECK-SAME: {{^}}, gfx9-generic
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx10-1-generic
 // CHECK-SAME: {{^}}, gfx10-3-generic
 // CHECK-SAME: {{^}}, gfx11-generic
diff --git a/clang/test/Misc/target-invalid-cpu-note/nvptx.c b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
index a59e1c6fab1c49..44fe07065b2428 100644
--- a/clang/test/Misc/target-invalid-cpu-note/nvptx.c
+++ b/clang/test/Misc/target-invalid-cpu-note/nvptx.c
@@ -50,6 +50,7 @@
 // CHECK-SAME: {{^}}, gfx909
 // CHECK-SAME: {{^}}, gfx90a
 // CHECK-SAME: {{^}}, gfx90c
+// CHECK-SAME: {{^}}, gfx9-4-generic
 // CHECK-SAME: {{^}}, gfx940
 // CHECK-SAME: {{^}}, gfx941
 // CHECK-SAME: {{^}}, gfx942
diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 5b83ea428c0bff..92d072bec6b57d 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -576,6 +576,11 @@ Generic processor code objects are versioned. See :ref:`amdgpu-generic-processor
                                                                                                   - ``v_dot2_f32_f16``
 
 
+     ``gfx9-4-generic``   ``amdgcn``     - ``gfx940``      - xnack            - Absolute flat   No restrictions.
+                                         - ``gfx941``                           scratch
+                                         - ``gfx942``
+
+
      ``gfx10-1-generic``  ``amdgcn``     - ``gfx1010``     - xnack            - Absolute flat   - The following instructions are
                                          - ``gfx1011``     - wavefrontsize64    scratch           not available on ``gfx1011``
                                          - ``gfx1012``     - cumode                               and ``gfx1012``
diff --git a/llvm/include/llvm/BinaryFormat/ELF.h b/llvm/include/llvm/BinaryFormat/ELF.h
index c591a96232f115..6c05ea7208e1f1 100644
--- a/llvm/include/llvm/BinaryFormat/ELF.h
+++ b/llvm/include/llvm/BinaryFormat/ELF.h
@@ -822,11 +822,12 @@ enum : unsigned {
   EF_AMDGPU_MACH_AMDGCN_RESERVED_0X57   = 0x057,
   EF_AMDGPU_MACH_AMDGCN_GFX1153         = 0x058,
   EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC   = 0x059,
+  EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC  = 0x05f,
   // clang-format on
 
   // First/last AMDGCN-based processors.
   EF_AMDGPU_MACH_AMDGCN_FIRST = EF_AMDGPU_MACH_AMDGCN_GFX600,
-  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC,
+  EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC,
 
   // Indicates if the "xnack" target feature is enabled for all code contained
   // in the object.
diff --git a/llvm/include/llvm/TargetParser/TargetParser.h b/llvm/include/llvm/TargetParser/TargetParser.h
index ae86ff39083d89..c6db4dfd7f5159 100644
--- a/llvm/include/llvm/TargetParser/TargetParser.h
+++ b/llvm/include/llvm/TargetParser/TargetParser.h
@@ -119,9 +119,10 @@ enum GPUKind : uint32_t {
   GK_GFX10_3_GENERIC = 194,
   GK_GFX11_GENERIC = 195,
   GK_GFX12_GENERIC = 196,
+  GK_GFX9_4_GENERIC = 197,
 
   GK_AMDGCN_GENERIC_FIRST = GK_GFX9_GENERIC,
-  GK_AMDGCN_GENERIC_LAST = GK_GFX12_GENERIC,
+  GK_AMDGCN_GENERIC_LAST = GK_GFX9_4_GENERIC,
 };
 
 /// Instruction set architecture version.
diff --git a/llvm/lib/Object/ELFObjectFile.cpp b/llvm/lib/Object/ELFObjectFile.cpp
index 5096877d2a4b00..9dc39936ffd8bb 100644
--- a/llvm/lib/Object/ELFObjectFile.cpp
+++ b/llvm/lib/Object/ELFObjectFile.cpp
@@ -602,6 +602,8 @@ StringRef ELFObjectFileBase::getAMDGPUCPUName() const {
   // Generic AMDGCN targets
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:
     return "gfx9-generic";
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:
+    return "gfx9-4-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:
     return "gfx10-1-generic";
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:
diff --git a/llvm/lib/ObjectYAML/ELFYAML.cpp b/llvm/lib/ObjectYAML/ELFYAML.cpp
index e97248cbcf5682..76c6c496c5e6bc 100644
--- a/llvm/lib/ObjectYAML/ELFYAML.cpp
+++ b/llvm/lib/ObjectYAML/ELFYAML.cpp
@@ -631,6 +631,7 @@ void ScalarBitSetTraits<ELFYAML::ELF_EF>::bitset(IO &IO,
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1200, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1201, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC, EF_AMDGPU_MACH);
+    BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC, EF_AMDGPU_MACH);
     BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC, EF_AMDGPU_MACH);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 910f5e06a6f3c4..2f8d6f1c7f8065 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1478,6 +1478,10 @@ def FeatureISAVersion9_4_2 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
     [])>;
 
+def FeatureISAVersion9_4_Generic : FeatureSet<
+  !listconcat(FeatureISAVersion9_4_Common.Features,
+    [FeatureRequiresCOV6])>;
+
 def FeatureISAVersion10_Common : FeatureSet<
   [FeatureGFX10,
    FeatureLDSBankCount32,
diff --git a/llvm/lib/Target/AMDGPU/GCNProcessors.td b/llvm/lib/Target/AMDGPU/GCNProcessors.td
index 547941633fda61..e12a6127b17063 100644
--- a/llvm/lib/Target/AMDGPU/GCNProcessors.td
+++ b/llvm/lib/Target/AMDGPU/GCNProcessors.td
@@ -209,6 +209,11 @@ def : ProcessorModel<"gfx9-generic", SIQuarterSpeedModel,
   FeatureISAVersion9_Generic.Features
 >;
 
+// [gfx940, gfx941, gfx942]
+def : ProcessorModel<"gfx9-4-generic", SIQuarterSpeedModel,
+  FeatureISAVersion9_4_Generic.Features
+>;
+
 //===----------------------------------------------------------------------===//
 // GCN GFX10.
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
index 5c625c3d83ff1b..507725b91a9bee 100644
--- a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
@@ -119,6 +119,7 @@ StringRef AMDGPUTargetStreamer::getArchNameFromElfMach(unsigned ElfMach) {
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200: AK = GK_GFX1200; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201: AK = GK_GFX1201; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:     AK = GK_GFX9_GENERIC; break;
+  case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:   AK = GK_GFX9_4_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:  AK = GK_GFX10_1_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:  AK = GK_GFX10_3_GENERIC; break;
   case ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC:    AK = GK_GFX11_GENERIC; break;
@@ -204,6 +205,7 @@ unsigned AMDGPUTargetStreamer::getElfMach(StringRef GPU) {
   case GK_GFX1200: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200;
   case GK_GFX1201: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201;
   case GK_GFX9_GENERIC:     return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC;
+  case GK_GFX9_4_GENERIC:   return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC;
   case GK_GFX10_1_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC;
   case GK_GFX10_3_GENERIC:  return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC;
   case GK_GFX11_GENERIC:    return ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC;
@@ -821,6 +823,9 @@ unsigned AMDGPUTargetELFStreamer::getEFlagsV6() {
     case AMDGPU::GK_GFX9_GENERIC:
       Version = GenericVersion::GFX9;
       break;
+    case AMDGPU::GK_GFX9_4_GENERIC:
+      Version = GenericVersion::GFX9_4;
+      break;
     case AMDGPU::GK_GFX10_1_GENERIC:
       Version = GenericVersion::GFX10_1;
       break;
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index beebe320b2cf3a..88a6d75b72c7d0 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -46,6 +46,7 @@ struct IsaVersion;
 /// within a generic family.
 namespace GenericVersion {
 static constexpr unsigned GFX9 = 1;
+static constexpr unsigned GFX9_4 = 1;
 static constexpr unsigned GFX10_1 = 1;
 static constexpr unsigned GFX10_3 = 1;
 static constexpr unsigned GFX11 = 1;
diff --git a/llvm/lib/TargetParser/TargetParser.cpp b/llvm/lib/TargetParser/TargetParser.cpp
index 059d7b4f5ff2d0..e7a979eaedb75c 100644
--- a/llvm/lib/TargetParser/TargetParser.cpp
+++ b/llvm/lib/TargetParser/TargetParser.cpp
@@ -130,6 +130,7 @@ constexpr GPUInfo AMDGCNGPUs[] = {
     {{"gfx1201"},   {"gfx1201"}, GK_GFX1201, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
 
     {{"gfx9-generic"},      {"gfx9-generic"},    GK_GFX9_GENERIC,    FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK},
+    {{"gfx9-4-generic"},    {"gfx9-4-generic"},  GK_GFX9_4_GENERIC,  FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK|FEATURE_SRAMECC},
     {{"gfx10-1-generic"},   {"gfx10-1-generic"}, GK_GFX10_1_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_XNACK|FEATURE_WGP},
     {{"gfx10-3-generic"},   {"gfx10-3-generic"}, GK_GFX10_3_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
     {{"gfx11-generic"},     {"gfx11-generic"},   GK_GFX11_GENERIC,   FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
@@ -156,6 +157,8 @@ StringRef llvm::AMDGPU::getArchFamilyNameAMDGCN(GPUKind AK) {
   switch (AK) {
   case AMDGPU::GK_GFX9_GENERIC:
     return "gfx9";
+  case AMDGPU::GK_GFX9_4_GENERIC:
+    return "gfx9";
   case AMDGPU::GK_GFX10_1_GENERIC:
   case AMDGPU::GK_GFX10_3_GENERIC:
     return "gfx10";
@@ -296,6 +299,7 @@ AMDGPU::IsaVersion AMDGPU::getIsaVersion(StringRef GPU) {
   // TODO: Split up this API depending on its caller so
   // generic target handling is more obvious and less risky.
   case GK_GFX9_GENERIC:    return {9, 0, 0};
+  case GK_GFX9_4_GENERIC:  return {9, 4, 0};
   case GK_GFX10_1_GENERIC: return {10, 1, 0};
   case GK_GFX10_3_GENERIC: return {10, 3, 0};
   case GK_GFX11_GENERIC:   return {11, 0, 3};
@@ -466,6 +470,7 @@ void AMDGPU::fillAMDGPUFeatureMap(StringRef GPU, const Triple &T,
     case GK_GFX942:
     case GK_GFX941:
     case GK_GFX940:
+    case GK_GFX9_4_GENERIC:
       Features["gfx940-insts"] = true;
       Features["fp8-insts"] = true;
       Features["fp8-conversion-insts"] = true;
diff --git a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
index 5986d2d38ef1ad..4eac26e853c2a0 100644
--- a/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
+++ b/llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll
@@ -112,6 +112,8 @@
 
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_XNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_NOXNACK %s
+; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_4_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_NOXNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-1-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX10_1_GENERIC_XNACK %s
 ; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx10-3-generic < %s | FileCheck --check-prefixes=GFX10_3_GENERIC %s
@@ -210,6 +212,8 @@
 
 ; GFX9_GENERIC_NOXNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack-"
 ; GFX9_GENERIC_XNACK:       .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack+"
+; GFX9_4_GENERIC_NOXNACK:   .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack-"
+; GFX9_4_GENERIC_XNACK:     .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack+"
 ; GFX10_1_GENERIC_NOXNACK:  .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack-"
 ; GFX10_1_GENERIC_XNACK:    .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack+"
 ; GFX10_3_GENERIC:          .amdgcn_target "amdgcn-amd-amdhsa--gfx10-3-generic"
diff --git a/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll b/llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll
index 662de47413654f..10ef4dc280d091 100644
--- a/llvm...
[truncated]

llvm/lib/TargetParser/TargetParser.cpp

llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.large.mir

llvm/lib/Target/AMDGPU/VOP3PInstructions.td

llvm/lib/TargetParser/TargetParser.cpp

llvm/docs/AMDGPUUsage.rst

clang/lib/Basic/Cuda.cpp

arsenm

Test that the unsupported clang builtins are errors?

arsenm · 2024-11-08T18:48:41Z

clang/test/CodeGenOpenCL/amdgpu-features.cl

@@ -85,6 +87,7 @@
 // GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX942: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX9_4_Generic: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-conversion-insts,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"


we really should stop adding this spam in clang

shiltian · 2024-11-10T20:58:49Z

Test that the unsupported clang builtins are errors?

Done.

shiltian · 2024-11-12T20:09:28Z

gentle ping

clang/test/CodeGenOpenCL/builtins-amdgcn-gfx9-4-generic-err.cl

shiltian · 2024-11-13T04:06:47Z

Merge activity

Nov 12, 11:06 PM EST: A user started a stack merge that includes this pull request via Graphite.
Nov 12, 11:08 PM EST: Graphite rebased this pull request as part of a merge.
Nov 12, 11:11 PM EST: A user merged this pull request with Graphite.

This patch introduces a new generic target, `gfx9-4-generic`. Since it doesn’t support FP8 and XF32-related instructions, the patch includes several code reorganizations to accommodate these changes. (cherry picked from commit de0fd64) Change-Id: I0ee078d6e653f6fd78f2efbb66adf27e5f41fa53

shiltian requested review from jayfoad, arsenm, scchan, rampitec, kzhuravl and Pierre-vh November 6, 2024 18:19

shiltian commented Nov 6, 2024

View reviewed changes

llvm/lib/TargetParser/TargetParser.cpp Show resolved Hide resolved

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from a24e413 to defb12c Compare November 6, 2024 18:21

rampitec reviewed Nov 6, 2024

View reviewed changes

llvm/test/CodeGen/AMDGPU/div-rem-by-constant-64.ll Outdated Show resolved Hide resolved

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.opt.exp.large.mir Outdated Show resolved Hide resolved

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from defb12c to 7c36223 Compare November 6, 2024 19:35

llvmbot added the mc Machine (object) code label Nov 6, 2024

shiltian force-pushed the users/shiltian/gfx9-4-generic branch 3 times, most recently from 02ca83a to c3a5591 Compare November 6, 2024 21:30

shiltian changed the base branch from main to users/shiltian/xf32-insts November 6, 2024 21:30

This was referenced Nov 6, 2024

[NFC][AMDGPU] Guard FP8 related instructions properly #115211

Merged

[AMDGPU] Introduce a "new" target feature xf32-insts #115214

Merged

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from c3a5591 to c70cc83 Compare November 6, 2024 21:35

shiltian commented Nov 6, 2024

View reviewed changes

llvm/lib/Target/AMDGPU/VOP3PInstructions.td Outdated Show resolved Hide resolved

shiltian commented Nov 6, 2024

View reviewed changes

llvm/lib/TargetParser/TargetParser.cpp Outdated Show resolved Hide resolved

shiltian force-pushed the users/shiltian/gfx9-4-generic branch 2 times, most recently from 081d98d to 2962968 Compare November 6, 2024 21:47

kzhuravl reviewed Nov 7, 2024

View reviewed changes

llvm/docs/AMDGPUUsage.rst Show resolved Hide resolved

shiltian force-pushed the users/shiltian/xf32-insts branch from 4077a19 to 1bf7b52 Compare November 7, 2024 17:17

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from 2962968 to ca91f78 Compare November 7, 2024 17:17

shiltian force-pushed the users/shiltian/xf32-insts branch from 1bf7b52 to c8a38d2 Compare November 7, 2024 17:21

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from ca91f78 to f605ddb Compare November 7, 2024 17:22

shiltian force-pushed the users/shiltian/xf32-insts branch from c8a38d2 to 45be0db Compare November 7, 2024 17:39

Base automatically changed from users/shiltian/xf32-insts to main November 7, 2024 17:41

shiltian force-pushed the users/shiltian/gfx9-4-generic branch 3 times, most recently from 36713d3 to 2fa6523 Compare November 8, 2024 04:43

arsenm reviewed Nov 8, 2024

View reviewed changes

clang/lib/Basic/Cuda.cpp Show resolved Hide resolved

arsenm reviewed Nov 8, 2024

View reviewed changes

shiltian force-pushed the users/shiltian/gfx9-4-generic branch 3 times, most recently from 629e974 to ee359e3 Compare November 10, 2024 20:56

arsenm approved these changes Nov 12, 2024

View reviewed changes

clang/test/CodeGenOpenCL/builtins-amdgcn-gfx9-4-generic-err.cl Outdated Show resolved Hide resolved

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from ee359e3 to 43ef3b9 Compare November 13, 2024 03:56

[AMDGPU] Introduce a new generic target gfx9-4-generic

15a0171

shiltian force-pushed the users/shiltian/gfx9-4-generic branch from 43ef3b9 to 15a0171 Compare November 13, 2024 04:08

shiltian merged commit de0fd64 into main Nov 13, 2024
6 of 8 checks passed

shiltian deleted the users/shiltian/gfx9-4-generic branch November 13, 2024 04:11

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[AMDGPU] Introduce a new generic target `gfx9-4-generic` #115190

[AMDGPU] Introduce a new generic target `gfx9-4-generic` #115190

shiltian commented Nov 6, 2024 •

edited

Loading

shiltian commented Nov 6, 2024 •

edited

Loading

llvmbot commented Nov 6, 2024 •

edited

Loading

llvmbot commented Nov 6, 2024

llvmbot commented Nov 6, 2024

arsenm left a comment

arsenm Nov 8, 2024

shiltian commented Nov 10, 2024

shiltian commented Nov 12, 2024

shiltian commented Nov 13, 2024 •

edited

Loading

[AMDGPU] Introduce a new generic target gfx9-4-generic #115190

[AMDGPU] Introduce a new generic target gfx9-4-generic #115190

Conversation

shiltian commented Nov 6, 2024 • edited Loading

shiltian commented Nov 6, 2024 • edited Loading

llvmbot commented Nov 6, 2024 • edited Loading

llvmbot commented Nov 6, 2024

llvmbot commented Nov 6, 2024

arsenm left a comment

Choose a reason for hiding this comment

arsenm Nov 8, 2024

Choose a reason for hiding this comment

shiltian commented Nov 10, 2024

shiltian commented Nov 12, 2024

shiltian commented Nov 13, 2024 • edited Loading

Merge activity

[AMDGPU] Introduce a new generic target `gfx9-4-generic` #115190

[AMDGPU] Introduce a new generic target `gfx9-4-generic` #115190

shiltian commented Nov 6, 2024 •

edited

Loading

shiltian commented Nov 6, 2024 •

edited

Loading

llvmbot commented Nov 6, 2024 •

edited

Loading

shiltian commented Nov 13, 2024 •

edited

Loading