Skip to content

Commit a1197a2

Browse files
kinoshita-fjkinoshita-fj
authored
[AArch64] Add initial support for FUJITSU-MONAKA (#118432)
This patch adds initial support for FUJITSU-MONAKA CPU (-mcpu=fujitsu-monaka). The scheduling model will be corrected in the future.
1 parent fe0d0a9 commit a1197a2

File tree

9 files changed

+126
-3
lines changed

9 files changed

+126
-3
lines changed

‎clang/test/Driver/aarch64-fujitsu-monaka.c

Lines changed: 13 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,13 @@
1+
// RUN: %clang --target=aarch64 -mcpu=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=fujitsu-monaka %s
2+
// RUN: %clang --target=aarch64 -mlittle-endian -mcpu=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=fujitsu-monaka %s
3+
// RUN: %clang --target=aarch64 -mtune=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=fujitsu-monaka-TUNE %s
4+
// RUN: %clang --target=aarch64 -mlittle-endian -mtune=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=fujitsu-monaka-TUNE %s
5+
// fujitsu-monaka: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "fujitsu-monaka"
6+
// fujitsu-monaka-TUNE: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "generic"
7+
8+
// RUN: %clang --target=arm64 -mcpu=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-fujitsu-monaka %s
9+
// RUN: %clang --target=arm64 -mlittle-endian -mcpu=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-fujitsu-monaka %s
10+
// RUN: %clang --target=arm64 -mtune=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-fujitsu-monaka-TUNE %s
11+
// RUN: %clang --target=arm64 -mlittle-endian -mtune=fujitsu-monaka -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-fujitsu-monaka-TUNE %s
12+
// ARM64-fujitsu-monaka: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "fujitsu-monaka"
13+
// ARM64-fujitsu-monaka-TUNE: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "generic"

‎clang/test/Driver/print-enabled-extensions/aarch64-fujitsu-monaka.c

Lines changed: 82 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,82 @@
1+
// REQUIRES: aarch64-registered-target
2+
// RUN: %clang --target=aarch64 --print-enabled-extensions -mcpu=fujitsu-monaka | FileCheck --strict-whitespace --implicit-check-not=FEAT_ %s
3+
4+
// CHECK: Extensions enabled for the given AArch64 target
5+
// CHECK-EMPTY:
6+
// CHECK-NEXT: Architecture Feature(s) Description
7+
// CHECK-NEXT: FEAT_AES, FEAT_PMULL Enable AES support
8+
// CHECK-NEXT: FEAT_AMUv1 Enable Armv8.4-A Activity Monitors extension
9+
// CHECK-NEXT: FEAT_AMUv1p1 Enable Armv8.6-A Activity Monitors Virtualization support
10+
// CHECK-NEXT: FEAT_AdvSIMD Enable Advanced SIMD instructions
11+
// CHECK-NEXT: FEAT_BF16 Enable BFloat16 Extension
12+
// CHECK-NEXT: FEAT_BTI Enable Branch Target Identification
13+
// CHECK-NEXT: FEAT_CCIDX Enable Armv8.3-A Extend of the CCSIDR number of sets
14+
// CHECK-NEXT: FEAT_CLRBHB Enable Clear BHB instruction
15+
// CHECK-NEXT: FEAT_CRC32 Enable Armv8.0-A CRC-32 checksum instructions
16+
// CHECK-NEXT: FEAT_CSV2_2 Enable architectural speculation restriction
17+
// CHECK-NEXT: FEAT_DIT Enable Armv8.4-A Data Independent Timing instructions
18+
// CHECK-NEXT: FEAT_DPB Enable Armv8.2-A data Cache Clean to Point of Persistence
19+
// CHECK-NEXT: FEAT_DPB2 Enable Armv8.5-A Cache Clean to Point of Deep Persistence
20+
// CHECK-NEXT: FEAT_DotProd Enable dot product support
21+
// CHECK-NEXT: FEAT_ECV Enable enhanced counter virtualization extension
22+
// CHECK-NEXT: FEAT_ETE Enable Embedded Trace Extension
23+
// CHECK-NEXT: FEAT_FAMINMAX Enable FAMIN and FAMAX instructions
24+
// CHECK-NEXT: FEAT_FCMA Enable Armv8.3-A Floating-point complex number support
25+
// CHECK-NEXT: FEAT_FGT Enable fine grained virtualization traps extension
26+
// CHECK-NEXT: FEAT_FHM Enable FP16 FML instructions
27+
// CHECK-NEXT: FEAT_FP Enable Armv8.0-A Floating Point Extensions
28+
// CHECK-NEXT: FEAT_FP16 Enable half-precision floating-point data processing
29+
// CHECK-NEXT: FEAT_FP8 Enable FP8 instructions
30+
// CHECK-NEXT: FEAT_FP8DOT2 Enable FP8 2-way dot instructions
31+
// CHECK-NEXT: FEAT_FP8DOT4 Enable FP8 4-way dot instructions
32+
// CHECK-NEXT: FEAT_FP8FMA Enable Armv9.5-A FP8 multiply-add instructions
33+
// CHECK-NEXT: FEAT_FPAC Enable Armv8.3-A Pointer Authentication Faulting enhancement
34+
// CHECK-NEXT: FEAT_FRINTTS Enable FRInt[32|64][Z|X] instructions that round a floating-point number to an integer (in FP format) forcing it to fit into a 32- or 64-bit int
35+
// CHECK-NEXT: FEAT_FlagM Enable Armv8.4-A Flag Manipulation instructions
36+
// CHECK-NEXT: FEAT_FlagM2 Enable alternative NZCV format for floating point comparisons
37+
// CHECK-NEXT: FEAT_HBC Enable Armv8.8-A Hinted Conditional Branches Extension
38+
// CHECK-NEXT: FEAT_HCX Enable Armv8.7-A HCRX_EL2 system register
39+
// CHECK-NEXT: FEAT_I8MM Enable Matrix Multiply Int8 Extension
40+
// CHECK-NEXT: FEAT_JSCVT Enable Armv8.3-A JavaScript FP conversion instructions
41+
// CHECK-NEXT: FEAT_LOR Enable Armv8.1-A Limited Ordering Regions extension
42+
// CHECK-NEXT: FEAT_LRCPC Enable support for RCPC extension
43+
// CHECK-NEXT: FEAT_LRCPC2 Enable Armv8.4-A RCPC instructions with Immediate Offsets
44+
// CHECK-NEXT: FEAT_LS64, FEAT_LS64_V, FEAT_LS64_ACCDATA Enable Armv8.7-A LD64B/ST64B Accelerator Extension
45+
// CHECK-NEXT: FEAT_LSE Enable Armv8.1-A Large System Extension (LSE) atomic instructions
46+
// CHECK-NEXT: FEAT_LSE2 Enable Armv8.4-A Large System Extension 2 (LSE2) atomicity rules
47+
// CHECK-NEXT: FEAT_LUT Enable Lookup Table instructions
48+
// CHECK-NEXT: FEAT_MEC Enable Memory Encryption Contexts Extension
49+
// CHECK-NEXT: FEAT_MOPS Enable Armv8.8-A memcpy and memset acceleration instructions
50+
// CHECK-NEXT: FEAT_MPAM Enable Armv8.4-A Memory system Partitioning and Monitoring extension
51+
// CHECK-NEXT: FEAT_NMI, FEAT_GICv3_NMI Enable Armv8.8-A Non-maskable Interrupts
52+
// CHECK-NEXT: FEAT_NV, FEAT_NV2 Enable Armv8.4-A Nested Virtualization Enchancement
53+
// CHECK-NEXT: FEAT_PAN Enable Armv8.1-A Privileged Access-Never extension
54+
// CHECK-NEXT: FEAT_PAN2 Enable Armv8.2-A PAN s1e1R and s1e1W Variants
55+
// CHECK-NEXT: FEAT_PAuth Enable Armv8.3-A Pointer Authentication extension
56+
// CHECK-NEXT: FEAT_PMUv3 Enable Armv8.0-A PMUv3 Performance Monitors extension
57+
// CHECK-NEXT: FEAT_RAS, FEAT_RASv1p1 Enable Armv8.0-A Reliability, Availability and Serviceability Extensions
58+
// CHECK-NEXT: FEAT_RDM Enable Armv8.1-A Rounding Double Multiply Add/Subtract instructions
59+
// CHECK-NEXT: FEAT_RME Enable Realm Management Extension
60+
// CHECK-NEXT: FEAT_RNG Enable Random Number generation instructions
61+
// CHECK-NEXT: FEAT_SB Enable Armv8.5-A Speculation Barrier
62+
// CHECK-NEXT: FEAT_SEL2 Enable Armv8.4-A Secure Exception Level 2 extension
63+
// CHECK-NEXT: FEAT_SHA1, FEAT_SHA256 Enable SHA1 and SHA256 support
64+
// CHECK-NEXT: FEAT_SHA3, FEAT_SHA512 Enable SHA512 and SHA3 support
65+
// CHECK-NEXT: FEAT_SM4, FEAT_SM3 Enable SM3 and SM4 support
66+
// CHECK-NEXT: FEAT_SPECRES Enable Armv8.5-A execution and data prediction invalidation instructions
67+
// CHECK-NEXT: FEAT_SPECRES2 Enable Speculation Restriction Instruction
68+
// CHECK-NEXT: FEAT_SPEv1p2 Enable extra register in the Statistical Profiling Extension
69+
// CHECK-NEXT: FEAT_SSBS, FEAT_SSBS2 Enable Speculative Store Bypass Safe bit
70+
// CHECK-NEXT: FEAT_SVE Enable Scalable Vector Extension (SVE) instructions
71+
// CHECK-NEXT: FEAT_SVE2 Enable Scalable Vector Extension 2 (SVE2) instructions
72+
// CHECK-NEXT: FEAT_SVE_AES, FEAT_SVE_PMULL128 Enable SVE AES and quadword SVE polynomial multiply instructions
73+
// CHECK-NEXT: FEAT_SVE_BitPerm Enable bit permutation SVE2 instructions
74+
// CHECK-NEXT: FEAT_SVE_SHA3 Enable SHA3 SVE2 instructions
75+
// CHECK-NEXT: FEAT_SVE_SM4 Enable SM4 SVE2 instructions
76+
// CHECK-NEXT: FEAT_TLBIOS, FEAT_TLBIRANGE Enable Armv8.4-A TLB Range and Maintenance instructions
77+
// CHECK-NEXT: FEAT_TRBE Enable Trace Buffer Extension
78+
// CHECK-NEXT: FEAT_TRF Enable Armv8.4-A Trace extension
79+
// CHECK-NEXT: FEAT_UAO Enable Armv8.2-A UAO PState
80+
// CHECK-NEXT: FEAT_VHE Enable Armv8.1-A Virtual Host extension
81+
// CHECK-NEXT: FEAT_WFxT Enable Armv8.7-A WFET and WFIT instruction
82+
// CHECK-NEXT: FEAT_XS Enable Armv8.7-A limited-TLB-maintenance instruction

‎clang/test/Misc/target-invalid-cpu-note/aarch64.c

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -67,6 +67,7 @@
6767
// CHECK-SAME: {{^}}, exynos-m4
6868
// CHECK-SAME: {{^}}, exynos-m5
6969
// CHECK-SAME: {{^}}, falkor
70+
// CHECK-SAME: {{^}}, fujitsu-monaka
7071
// CHECK-SAME: {{^}}, generic
7172
// CHECK-SAME: {{^}}, grace
7273
// CHECK-SAME: {{^}}, kryo

‎llvm/lib/Target/AArch64/AArch64Processors.td

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -266,6 +266,14 @@ def TuneA64FX : SubtargetFeature<"a64fx", "ARMProcFamily", "A64FX",
266266
FeatureStorePairSuppress,
267267
FeaturePredictableSelectIsExpensive]>;
268268

269+
def TuneMONAKA : SubtargetFeature<"fujitsu-monaka", "ARMProcFamily", "MONAKA",
270+
"Fujitsu FUJITSU-MONAKA processors", [
271+
FeaturePredictableSelectIsExpensive,
272+
FeatureEnableSelectOptimize,
273+
FeaturePostRAScheduler,
274+
FeatureArithmeticBccFusion,
275+
]>;
276+
269277
def TuneCarmel : SubtargetFeature<"carmel", "ARMProcFamily", "Carmel",
270278
"Nvidia Carmel processors">;
271279

@@ -843,6 +851,12 @@ def ProcessorFeatures {
843851
FeatureSHA2, FeaturePerfMon, FeatureFullFP16,
844852
FeatureSVE, FeatureComplxNum,
845853
FeatureAES, FeatureCRC, FeatureLSE, FeatureRAS, FeatureRDM];
854+
list<SubtargetFeature> MONAKA = [HasV9_3aOps, FeaturePerfMon, FeatureCCIDX,
855+
FeatureFPAC, FeatureFP16FML, FeatureRandGen,
856+
FeatureSSBS, FeatureLS64, FeatureCLRBHB,
857+
FeatureSPECRES2, FeatureSVEAES, FeatureSVE2SM4,
858+
FeatureSVE2SHA3, FeatureSVE2BitPerm, FeatureETE,
859+
FeatureMEC, FeatureFP8DOT2];
846860
list<SubtargetFeature> Carmel = [HasV8_2aOps, FeatureNEON, FeatureSHA2, FeatureAES,
847861
FeatureFullFP16, FeatureCRC, FeatureLSE, FeatureRAS, FeatureRDM,
848862
FeatureFPARMv8];
@@ -1225,6 +1239,10 @@ def : ProcessorAlias<"apple-latest", "apple-m4">;
12251239
def : ProcessorModel<"a64fx", A64FXModel, ProcessorFeatures.A64FX,
12261240
[TuneA64FX]>;
12271241

1242+
// Fujitsu FUJITSU-MONAKA
1243+
def : ProcessorModel<"fujitsu-monaka", A64FXModel, ProcessorFeatures.MONAKA,
1244+
[TuneMONAKA]>;
1245+
12281246
// Nvidia Carmel
12291247
def : ProcessorModel<"carmel", NoSchedModel, ProcessorFeatures.Carmel,
12301248
[TuneCarmel]>;

‎llvm/lib/Target/AArch64/AArch64Subtarget.cpp

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -195,6 +195,9 @@ void AArch64Subtarget::initializeProperties(bool HasMinSize) {
195195
MaxPrefetchIterationsAhead = 4;
196196
VScaleForTuning = 4;
197197
break;
198+
case MONAKA:
199+
VScaleForTuning = 2;
200+
break;
198201
case AppleA7:
199202
case AppleA10:
200203
case AppleA11:

‎llvm/lib/TargetParser/Host.cpp

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -280,8 +280,9 @@ StringRef sys::detail::getHostCPUNameForARM(StringRef ProcCpuinfoContent) {
280280

281281
if (Implementer == "0x46") { // Fujitsu Ltd.
282282
return StringSwitch<const char *>(Part)
283-
.Case("0x001", "a64fx")
284-
.Default("generic");
283+
.Case("0x001", "a64fx")
284+
.Case("0x003", "fujitsu-monaka")
285+
.Default("generic");
285286
}
286287

287288
if (Implementer == "0x4e") { // NVIDIA Corporation

‎llvm/test/CodeGen/AArch64/cpus.ll

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -36,6 +36,7 @@
3636
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=tsv110 2>&1 | FileCheck %s
3737
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=apple-latest 2>&1 | FileCheck %s
3838
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=a64fx 2>&1 | FileCheck %s
39+
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=fujitsu-monaka 2>&1 | FileCheck %s
3940
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=ampere1 2>&1 | FileCheck %s
4041
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=ampere1a 2>&1 | FileCheck %s
4142
; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=ampere1b 2>&1 | FileCheck %s

‎llvm/unittests/TargetParser/Host.cpp

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -140,6 +140,9 @@ TEST(getLinuxHostCPUName, AArch64) {
140140
EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x51\n"
141141
"CPU part : 0x001"),
142142
"oryon-1");
143+
EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x46\n"
144+
"CPU part : 0x003"),
145+
"fujitsu-monaka");
143146

144147
// MSM8992/4 weirdness
145148
StringRef MSM8992ProcCpuInfo = R"(

‎llvm/unittests/TargetParser/TargetParserTest.cpp

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1158,13 +1158,14 @@ INSTANTIATE_TEST_SUITE_P(
11581158
AArch64CPUTestParams("thunderxt88", "armv8-a"),
11591159
AArch64CPUTestParams("tsv110", "armv8.2-a"),
11601160
AArch64CPUTestParams("a64fx", "armv8.2-a"),
1161+
AArch64CPUTestParams("fujitsu-monaka", "armv9.3-a"),
11611162
AArch64CPUTestParams("carmel", "armv8.2-a"),
11621163
AArch64CPUTestParams("saphira", "armv8.4-a"),
11631164
AArch64CPUTestParams("oryon-1", "armv8.6-a")),
11641165
AArch64CPUTestParams::PrintToStringParamName);
11651166

11661167
// Note: number of CPUs includes aliases.
1167-
static constexpr unsigned NumAArch64CPUArchs = 81;
1168+
static constexpr unsigned NumAArch64CPUArchs = 82;
11681169

11691170
TEST(TargetParserTest, testAArch64CPUArchList) {
11701171
SmallVector<StringRef, NumAArch64CPUArchs> List;

0 commit comments

Comments
 (0)