Skip to content

Commit 36713d3

Browse files
shiltianshiltian
committed
[AMDGPU] Introduce a new generic target gfx9-4-generic 
1 parent bfa3ffb commit 36713d3

33 files changed

+203
-10
lines changed

‎clang/include/clang/Basic/Cuda.h

+1
Original file line numberDiff line numberDiff line change
@@ -103,6 +103,7 @@ enum class OffloadArch {
103103
GFX909,
104104
GFX90a,
105105
GFX90c,
106+
GFX9_4_GENERIC,
106107
GFX940,
107108
GFX941,
108109
GFX942,

‎clang/lib/Basic/Cuda.cpp

+1
Original file line numberDiff line numberDiff line change
@@ -121,6 +121,7 @@ static const OffloadArchToStringMap arch_names[] = {
121121
GFX(909), // gfx909
122122
GFX(90a), // gfx90a
123123
GFX(90c), // gfx90c
124+
{OffloadArch::GFX9_4_GENERIC, "gfx9-4-generic", "compute_amdgcn"},
124125
GFX(940), // gfx940
125126
GFX(941), // gfx941
126127
GFX(942), // gfx942

‎clang/lib/Basic/Targets/NVPTX.cpp

+1
Original file line numberDiff line numberDiff line change
@@ -205,6 +205,7 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
205205
case OffloadArch::GFX909:
206206
case OffloadArch::GFX90a:
207207
case OffloadArch::GFX90c:
208+
case OffloadArch::GFX9_4_GENERIC:
208209
case OffloadArch::GFX940:
209210
case OffloadArch::GFX941:
210211
case OffloadArch::GFX942:

‎clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

+1
Original file line numberDiff line numberDiff line change
@@ -2301,6 +2301,7 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
23012301
case OffloadArch::GFX909:
23022302
case OffloadArch::GFX90a:
23032303
case OffloadArch::GFX90c:
2304+
case OffloadArch::GFX9_4_GENERIC:
23042305
case OffloadArch::GFX940:
23052306
case OffloadArch::GFX941:
23062307
case OffloadArch::GFX942:

‎clang/test/CodeGenOpenCL/amdgpu-features.cl

+3
Original file line numberDiff line numberDiff line change
@@ -56,6 +56,8 @@
5656

5757
// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize64 -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103-W64 %s
5858

59+
// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx9-4-generic -emit-llvm -o - %s | FileCheck --check-prefix=GFX9_4_Generic %s
60+
5961
// NOCPU-NOT: "target-features"
6062
// NOCPU-WAVE32: "target-features"="+wavefrontsize32"
6163
// NOCPU-WAVE64: "target-features"="+wavefrontsize64"
@@ -85,6 +87,7 @@
8587
// 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"
8688
// 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"
8789
// 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"
90+
// 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"
8891
// GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
8992
// 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"
9093
// 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"

‎clang/test/Driver/amdgpu-macros.cl

+1
Original file line numberDiff line numberDiff line change
@@ -133,6 +133,7 @@
133133
// 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
134134

135135
// 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
136+
// 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
136137
// 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
137138
// 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
138139
// 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

‎clang/test/Driver/amdgpu-mcpu.cl

+2
Original file line numberDiff line numberDiff line change
@@ -118,6 +118,7 @@
118118
// RUN: %clang -### -target amdgcn -mcpu=gfx1201 %s 2>&1 | FileCheck --check-prefix=GFX1201 %s
119119

120120
// RUN: %clang -### -target amdgcn -mcpu=gfx9-generic %s 2>&1 | FileCheck --check-prefix=GFX9_GENERIC %s
121+
// RUN: %clang -### -target amdgcn -mcpu=gfx9-4-generic %s 2>&1 | FileCheck --check-prefix=GFX9_4_GENERIC %s
121122
// RUN: %clang -### -target amdgcn -mcpu=gfx10-1-generic %s 2>&1 | FileCheck --check-prefix=GFX10_1_GENERIC %s
122123
// RUN: %clang -### -target amdgcn -mcpu=gfx10-3-generic %s 2>&1 | FileCheck --check-prefix=GFX10_3_GENERIC %s
123124
// RUN: %clang -### -target amdgcn -mcpu=gfx11-generic %s 2>&1 | FileCheck --check-prefix=GFX11_GENERIC %s
@@ -172,6 +173,7 @@
172173
// GFX1201: "-target-cpu" "gfx1201"
173174

174175
// GFX9_GENERIC: "-target-cpu" "gfx9-generic"
176+
// GFX9_4_GENERIC: "-target-cpu" "gfx9-4-generic"
175177
// GFX10_1_GENERIC: "-target-cpu" "gfx10-1-generic"
176178
// GFX10_3_GENERIC: "-target-cpu" "gfx10-3-generic"
177179
// GFX11_GENERIC: "-target-cpu" "gfx11-generic"

‎clang/test/Misc/target-invalid-cpu-note/amdgcn.c

+1
Original file line numberDiff line numberDiff line change
@@ -70,6 +70,7 @@
7070
// CHECK-SAME: {{^}}, gfx1200
7171
// CHECK-SAME: {{^}}, gfx1201
7272
// CHECK-SAME: {{^}}, gfx9-generic
73+
// CHECK-SAME: {{^}}, gfx9-4-generic
7374
// CHECK-SAME: {{^}}, gfx10-1-generic
7475
// CHECK-SAME: {{^}}, gfx10-3-generic
7576
// CHECK-SAME: {{^}}, gfx11-generic

‎clang/test/Misc/target-invalid-cpu-note/nvptx.c

+1
Original file line numberDiff line numberDiff line change
@@ -50,6 +50,7 @@
5050
// CHECK-SAME: {{^}}, gfx909
5151
// CHECK-SAME: {{^}}, gfx90a
5252
// CHECK-SAME: {{^}}, gfx90c
53+
// CHECK-SAME: {{^}}, gfx9-4-generic
5354
// CHECK-SAME: {{^}}, gfx940
5455
// CHECK-SAME: {{^}}, gfx941
5556
// CHECK-SAME: {{^}}, gfx942

‎llvm/docs/AMDGPUUsage.rst

+6
Original file line numberDiff line numberDiff line change
@@ -576,6 +576,12 @@ Generic processor code objects are versioned. See :ref:`amdgpu-generic-processor
576576
- ``v_dot2_f32_f16``
577577

578578

579+
``gfx9-4-generic`` ``amdgcn`` - ``gfx940`` - xnack - Absolute flat FP8 and BF8 instructions,
580+
- ``gfx941`` - sramecc scratch FP8 and BF8 conversion instructions,
581+
- ``gfx942`` as well as instructions with XF32 format support
582+
are not available.
583+
584+
579585
``gfx10-1-generic`` ``amdgcn`` - ``gfx1010`` - xnack - Absolute flat - The following instructions are
580586
- ``gfx1011`` - wavefrontsize64 scratch not available on ``gfx1011``
581587
- ``gfx1012`` - cumode and ``gfx1012``

‎llvm/include/llvm/BinaryFormat/ELF.h

+2-1
Original file line numberDiff line numberDiff line change
@@ -822,11 +822,12 @@ enum : unsigned {
822822
EF_AMDGPU_MACH_AMDGCN_RESERVED_0X57 = 0x057,
823823
EF_AMDGPU_MACH_AMDGCN_GFX1153 = 0x058,
824824
EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC = 0x059,
825+
EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC = 0x05f,
825826
// clang-format on
826827

827828
// First/last AMDGCN-based processors.
828829
EF_AMDGPU_MACH_AMDGCN_FIRST = EF_AMDGPU_MACH_AMDGCN_GFX600,
829-
EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC,
830+
EF_AMDGPU_MACH_AMDGCN_LAST = EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC,
830831

831832
// Indicates if the "xnack" target feature is enabled for all code contained
832833
// in the object.

‎llvm/include/llvm/TargetParser/TargetParser.h

+2-1
Original file line numberDiff line numberDiff line change
@@ -119,9 +119,10 @@ enum GPUKind : uint32_t {
119119
GK_GFX10_3_GENERIC = 194,
120120
GK_GFX11_GENERIC = 195,
121121
GK_GFX12_GENERIC = 196,
122+
GK_GFX9_4_GENERIC = 197,
122123

123124
GK_AMDGCN_GENERIC_FIRST = GK_GFX9_GENERIC,
124-
GK_AMDGCN_GENERIC_LAST = GK_GFX12_GENERIC,
125+
GK_AMDGCN_GENERIC_LAST = GK_GFX9_4_GENERIC,
125126
};
126127

127128
/// Instruction set architecture version.

‎llvm/lib/Object/ELFObjectFile.cpp

+2
Original file line numberDiff line numberDiff line change
@@ -602,6 +602,8 @@ StringRef ELFObjectFileBase::getAMDGPUCPUName() const {
602602
// Generic AMDGCN targets
603603
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:
604604
return "gfx9-generic";
605+
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:
606+
return "gfx9-4-generic";
605607
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:
606608
return "gfx10-1-generic";
607609
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:

‎llvm/lib/ObjectYAML/ELFYAML.cpp

+1
Original file line numberDiff line numberDiff line change
@@ -631,6 +631,7 @@ void ScalarBitSetTraits<ELFYAML::ELF_EF>::bitset(IO &IO,
631631
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1200, EF_AMDGPU_MACH);
632632
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX1201, EF_AMDGPU_MACH);
633633
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC, EF_AMDGPU_MACH);
634+
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC, EF_AMDGPU_MACH);
634635
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC, EF_AMDGPU_MACH);
635636
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC, EF_AMDGPU_MACH);
636637
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC, EF_AMDGPU_MACH);

‎llvm/lib/Target/AMDGPU/AMDGPU.td

+24-7
Original file line numberDiff line numberDiff line change
@@ -1451,11 +1451,7 @@ def FeatureISAVersion9_4_Common : FeatureSet<
14511451
FeatureDPALU_DPP,
14521452
FeaturePackedFP32Ops,
14531453
FeatureMAIInsts,
1454-
FeatureFP8Insts,
1455-
FeatureFP8ConversionInsts,
1456-
FeatureCvtFP8VOP1Bug,
14571454
FeaturePkFmacF16Inst,
1458-
FeatureXF32Insts,
14591455
FeatureAtomicFaddRtnInsts,
14601456
FeatureAtomicFaddNoRtnInsts,
14611457
FeatureAtomicBufferGlobalPkAddF16Insts,
@@ -1476,15 +1472,36 @@ def FeatureISAVersion9_4_Common : FeatureSet<
14761472

14771473
def FeatureISAVersion9_4_0 : FeatureSet<
14781474
!listconcat(FeatureISAVersion9_4_Common.Features,
1479-
[FeatureForceStoreSC0SC1])>;
1475+
[
1476+
FeatureForceStoreSC0SC1,
1477+
FeatureFP8Insts,
1478+
FeatureFP8ConversionInsts,
1479+
FeatureCvtFP8VOP1Bug,
1480+
FeatureXF32Insts
1481+
])>;
14801482

14811483
def FeatureISAVersion9_4_1 : FeatureSet<
14821484
!listconcat(FeatureISAVersion9_4_Common.Features,
1483-
[FeatureForceStoreSC0SC1])>;
1485+
[
1486+
FeatureForceStoreSC0SC1,
1487+
FeatureFP8Insts,
1488+
FeatureFP8ConversionInsts,
1489+
FeatureCvtFP8VOP1Bug,
1490+
FeatureXF32Insts
1491+
])>;
14841492

14851493
def FeatureISAVersion9_4_2 : FeatureSet<
14861494
!listconcat(FeatureISAVersion9_4_Common.Features,
1487-
[])>;
1495+
[
1496+
FeatureFP8Insts,
1497+
FeatureFP8ConversionInsts,
1498+
FeatureCvtFP8VOP1Bug,
1499+
FeatureXF32Insts
1500+
])>;
1501+
1502+
def FeatureISAVersion9_4_Generic : FeatureSet<
1503+
!listconcat(FeatureISAVersion9_4_Common.Features,
1504+
[FeatureRequiresCOV6])>;
14881505

14891506
def FeatureISAVersion10_Common : FeatureSet<
14901507
[FeatureGFX10,

‎llvm/lib/Target/AMDGPU/GCNProcessors.td

+5
Original file line numberDiff line numberDiff line change
@@ -209,6 +209,11 @@ def : ProcessorModel<"gfx9-generic", SIQuarterSpeedModel,
209209
FeatureISAVersion9_Generic.Features
210210
>;
211211

212+
// [gfx940, gfx941, gfx942]
213+
def : ProcessorModel<"gfx9-4-generic", SIQuarterSpeedModel,
214+
FeatureISAVersion9_4_Generic.Features
215+
>;
216+
212217
//===----------------------------------------------------------------------===//
213218
// GCN GFX10.
214219
//===----------------------------------------------------------------------===//

‎llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp

+5
Original file line numberDiff line numberDiff line change
@@ -119,6 +119,7 @@ StringRef AMDGPUTargetStreamer::getArchNameFromElfMach(unsigned ElfMach) {
119119
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200: AK = GK_GFX1200; break;
120120
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201: AK = GK_GFX1201; break;
121121
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC: AK = GK_GFX9_GENERIC; break;
122+
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC: AK = GK_GFX9_4_GENERIC; break;
122123
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC: AK = GK_GFX10_1_GENERIC; break;
123124
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC: AK = GK_GFX10_3_GENERIC; break;
124125
case ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC: AK = GK_GFX11_GENERIC; break;
@@ -204,6 +205,7 @@ unsigned AMDGPUTargetStreamer::getElfMach(StringRef GPU) {
204205
case GK_GFX1200: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1200;
205206
case GK_GFX1201: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX1201;
206207
case GK_GFX9_GENERIC: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC;
208+
case GK_GFX9_4_GENERIC: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC;
207209
case GK_GFX10_1_GENERIC: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC;
208210
case GK_GFX10_3_GENERIC: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC;
209211
case GK_GFX11_GENERIC: return ELF::EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC;
@@ -821,6 +823,9 @@ unsigned AMDGPUTargetELFStreamer::getEFlagsV6() {
821823
case AMDGPU::GK_GFX9_GENERIC:
822824
Version = GenericVersion::GFX9;
823825
break;
826+
case AMDGPU::GK_GFX9_4_GENERIC:
827+
Version = GenericVersion::GFX9_4;
828+
break;
824829
case AMDGPU::GK_GFX10_1_GENERIC:
825830
Version = GenericVersion::GFX10_1;
826831
break;

‎llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h

+1
Original file line numberDiff line numberDiff line change
@@ -46,6 +46,7 @@ struct IsaVersion;
4646
/// within a generic family.
4747
namespace GenericVersion {
4848
static constexpr unsigned GFX9 = 1;
49+
static constexpr unsigned GFX9_4 = 1;
4950
static constexpr unsigned GFX10_1 = 1;
5051
static constexpr unsigned GFX10_3 = 1;
5152
static constexpr unsigned GFX11 = 1;

‎llvm/lib/TargetParser/TargetParser.cpp

+7-1
Original file line numberDiff line numberDiff line change
@@ -130,6 +130,7 @@ constexpr GPUInfo AMDGCNGPUs[] = {
130130
{{"gfx1201"}, {"gfx1201"}, GK_GFX1201, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
131131

132132
{{"gfx9-generic"}, {"gfx9-generic"}, GK_GFX9_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK},
133+
{{"gfx9-4-generic"}, {"gfx9-4-generic"}, GK_GFX9_4_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_XNACK|FEATURE_SRAMECC},
133134
{{"gfx10-1-generic"}, {"gfx10-1-generic"}, GK_GFX10_1_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_XNACK|FEATURE_WGP},
134135
{{"gfx10-3-generic"}, {"gfx10-3-generic"}, GK_GFX10_3_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
135136
{{"gfx11-generic"}, {"gfx11-generic"}, GK_GFX11_GENERIC, FEATURE_FAST_FMA_F32|FEATURE_FAST_DENORMAL_F32|FEATURE_WAVE32|FEATURE_WGP},
@@ -155,6 +156,7 @@ const GPUInfo *getArchEntry(AMDGPU::GPUKind AK, ArrayRef<GPUInfo> Table) {
155156
StringRef llvm::AMDGPU::getArchFamilyNameAMDGCN(GPUKind AK) {
156157
switch (AK) {
157158
case AMDGPU::GK_GFX9_GENERIC:
159+
case AMDGPU::GK_GFX9_4_GENERIC:
158160
return "gfx9";
159161
case AMDGPU::GK_GFX10_1_GENERIC:
160162
case AMDGPU::GK_GFX10_3_GENERIC:
@@ -296,6 +298,7 @@ AMDGPU::IsaVersion AMDGPU::getIsaVersion(StringRef GPU) {
296298
// TODO: Split up this API depending on its caller so
297299
// generic target handling is more obvious and less risky.
298300
case GK_GFX9_GENERIC: return {9, 0, 0};
301+
case GK_GFX9_4_GENERIC: return {9, 4, 0};
299302
case GK_GFX10_1_GENERIC: return {10, 1, 0};
300303
case GK_GFX10_3_GENERIC: return {10, 3, 0};
301304
case GK_GFX11_GENERIC: return {11, 0, 3};
@@ -466,9 +469,12 @@ void AMDGPU::fillAMDGPUFeatureMap(StringRef GPU, const Triple &T,
466469
case GK_GFX942:
467470
case GK_GFX941:
468471
case GK_GFX940:
469-
Features["gfx940-insts"] = true;
470472
Features["fp8-insts"] = true;
471473
Features["fp8-conversion-insts"] = true;
474+
Features["xf32-insts"] = true;
475+
[[fallthrough]];
476+
case GK_GFX9_4_GENERIC:
477+
Features["gfx940-insts"] = true;
472478
Features["atomic-ds-pk-add-16-insts"] = true;
473479
Features["atomic-flat-pk-add-16-insts"] = true;
474480
Features["atomic-global-pk-add-bf16-inst"] = true;

‎llvm/test/CodeGen/AMDGPU/directive-amdgcn-target.ll

+4
Original file line numberDiff line numberDiff line change
@@ -112,6 +112,8 @@
112112

113113
; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=-xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_NOXNACK %s
114114
; RUN: llc --amdhsa-code-object-version=6 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-generic -mattr=+xnack < %s | FileCheck --check-prefixes=GFX9_GENERIC_XNACK %s
115+
; 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
116+
; 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
115117
; 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
116118
; 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
117119
; 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 @@
210212

211213
; GFX9_GENERIC_NOXNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack-"
212214
; GFX9_GENERIC_XNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx9-generic:xnack+"
215+
; GFX9_4_GENERIC_NOXNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack-"
216+
; GFX9_4_GENERIC_XNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx9-4-generic:xnack+"
213217
; GFX10_1_GENERIC_NOXNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack-"
214218
; GFX10_1_GENERIC_XNACK: .amdgcn_target "amdgcn-amd-amdhsa--gfx10-1-generic:xnack+"
215219
; GFX10_3_GENERIC: .amdgcn_target "amdgcn-amd-amdhsa--gfx10-3-generic"

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

+1
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,7 @@
11
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
22
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -o - %s | FileCheck -check-prefixes=GFX9 %s
33
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -o - %s | FileCheck -check-prefixes=GFX942 %s
4+
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx9-4-generic --amdhsa-code-object-version=6 -o - %s | FileCheck -check-prefixes=GFX942 %s
45
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -o - %s | FileCheck -check-prefixes=GFX1030 %s
56

67
; Sample test to check how we deal with division/modulos by 64 bit constants.

‎llvm/test/CodeGen/AMDGPU/dst-sel-hazard.mir

+1
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
22
# RUN: llc -mtriple=amdgcn -mcpu=gfx942 -run-pass post-RA-hazard-rec -o - %s | FileCheck -check-prefix=HAZARD %s
3+
# RUN: llc -mtriple=amdgcn -mcpu=gfx9-4-generic -run-pass post-RA-hazard-rec -o - %s | FileCheck -check-prefix=HAZARD %s
34
# RUN: llc -mtriple=amdgcn -mcpu=gfx90a -run-pass post-RA-hazard-rec -o - %s | FileCheck -check-prefix=NOHAZARD %s
45

56
---

‎llvm/test/CodeGen/AMDGPU/elf-header-flags-mach.ll

+2
Original file line numberDiff line numberDiff line change
@@ -80,6 +80,7 @@
8080
; RUN: llc -filetype=obj -mtriple=amdgcn -mcpu=gfx1201 < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX1201 %s
8181

8282
; RUN: llc -filetype=obj --amdhsa-code-object-version=6 -mtriple=amdgcn -mcpu=gfx9-generic < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX9_GENERIC %s
83+
; RUN: llc -filetype=obj --amdhsa-code-object-version=6 -mtriple=amdgcn -mcpu=gfx9-4-generic < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX9_4_GENERIC %s
8384
; RUN: llc -filetype=obj --amdhsa-code-object-version=6 -mtriple=amdgcn -mcpu=gfx10-1-generic < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX10_1_GENERIC %s
8485
; RUN: llc -filetype=obj --amdhsa-code-object-version=6 -mtriple=amdgcn -mcpu=gfx10-3-generic < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX10_3_GENERIC %s
8586
; RUN: llc -filetype=obj --amdhsa-code-object-version=6 -mtriple=amdgcn -mcpu=gfx11-generic < %s | llvm-readobj --file-header - | FileCheck --check-prefixes=ALL,ARCH-GCN,GFX11_GENERIC %s
@@ -161,6 +162,7 @@
161162
; GFX1201: EF_AMDGPU_MACH_AMDGCN_GFX1201 (0x4E)
162163

163164
; GFX9_GENERIC: EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC (0x51)
165+
; GFX9_4_GENERIC: EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC (0x5F)
164166
; GFX10_1_GENERIC: EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC (0x52)
165167
; GFX10_3_GENERIC: EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC (0x53)
166168
; GFX11_GENERIC: EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC (0x54)

0 commit comments

Comments
 (0)