Skip to content

Commit 9256485

Browse files
CarolineConcattoCarolineConcatto
authored
[Clang][LLVM][AArch64]Add new feature SSVE-BitPerm (#121947)
The 20204-12 ISA update release adds a new feature: FEAT_SSVE_BitPerm, which allows the sve-bitperm instructions to run in streaming mode. It also removes the requirement of FEAT_SVE2 for FEAT_SVE_BitPerm. The sve2-bitperm feature is now an alias for sve-bitperm and sve2. A new feature flag sve-bitperm is added to reflect the change that the instructions under FEAT_SVE_BitPerm are supported if: on non streaming mode with FEAT_SVE2 and FEAT_SVE_BitPerm or in streaming mode with FEAT_SME and FEAT_SSVE_BitPerm
1 parent 3d507a8 commit 9256485

31 files changed

+232
-152
lines changed

‎clang/include/clang/Basic/arm_sve.td‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1988,7 +1988,7 @@ def SVSM4E : SInst<"svsm4e[_{d}]", "ddd", "Ui", MergeNone, "aarch64_sve_sm
19881988
def SVSM4EKEY : SInst<"svsm4ekey[_{d}]", "ddd", "Ui", MergeNone, "aarch64_sve_sm4ekey", [IsOverloadNone]>;
19891989
}
19901990

1991-
let SVETargetGuard = "sve2-bitperm", SMETargetGuard = InvalidMode in {
1991+
let SVETargetGuard = "sve2,sve-bitperm", SMETargetGuard = InvalidMode in {
19921992
def SVBDEP : SInst<"svbdep[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_bdep_x">;
19931993
def SVBDEP_N : SInst<"svbdep[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_bdep_x">;
19941994
def SVBEXT : SInst<"svbext[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_bext_x">;

‎clang/lib/Basic/Targets/AArch64.cpp‎

Lines changed: 4 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -485,7 +485,7 @@ void AArch64TargetInfo::getTargetDefines(const LangOptions &Opts,
485485
if (HasSVE2 && HasSVEAES)
486486
Builder.defineMacro("__ARM_FEATURE_SVE2_AES", "1");
487487

488-
if (HasSVE2 && HasSVE2BitPerm)
488+
if (HasSVE2 && HasSVEBitPerm)
489489
Builder.defineMacro("__ARM_FEATURE_SVE2_BITPERM", "1");
490490

491491
if (HasSVE2 && HasSVE2SHA3)
@@ -769,7 +769,7 @@ bool AArch64TargetInfo::hasFeature(StringRef Feature) const {
769769
.Case("f64mm", FPU & SveMode && HasMatmulFP64)
770770
.Case("sve2", FPU & SveMode && HasSVE2)
771771
.Case("sve-aes", HasSVEAES)
772-
.Case("sve2-bitperm", FPU & SveMode && HasSVE2BitPerm)
772+
.Case("sve-bitperm", FPU & HasSVEBitPerm)
773773
.Case("sve2-sha3", FPU & SveMode && HasSVE2SHA3)
774774
.Case("sve2-sm4", FPU & SveMode && HasSVE2SM4)
775775
.Case("sve2p1", FPU & SveMode && HasSVE2p1)
@@ -881,12 +881,10 @@ bool AArch64TargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
881881
}
882882
if (Feature == "+sve-b16b16")
883883
HasSVEB16B16 = true;
884-
if (Feature == "+sve2-bitperm") {
884+
if (Feature == "+sve-bitperm") {
885885
FPU |= NeonMode;
886-
FPU |= SveMode;
887886
HasFullFP16 = true;
888-
HasSVE2 = true;
889-
HasSVE2BitPerm = true;
887+
HasSVEBitPerm = true;
890888
}
891889
if (Feature == "+f32mm") {
892890
FPU |= NeonMode;

‎clang/lib/Basic/Targets/AArch64.h‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -82,7 +82,7 @@ class LLVM_LIBRARY_VISIBILITY AArch64TargetInfo : public TargetInfo {
8282
bool HasSVE2SHA3 = false;
8383
bool HasSVE2SM4 = false;
8484
bool HasSVEB16B16 = false;
85-
bool HasSVE2BitPerm = false;
85+
bool HasSVEBitPerm = false;
8686
bool HasMatmulFP64 = false;
8787
bool HasMatmulFP32 = false;
8888
bool HasLSE = false;

‎clang/test/CodeGen/AArch64/fmv-dependencies.c‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -192,7 +192,7 @@ int caller() {
192192
// CHECK: attributes #[[sve]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sve,+v8a"
193193
// CHECK: attributes #[[sve2]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sve,+sve2,+v8a"
194194
// CHECK: attributes #[[sve2_aes]] = { {{.*}} "target-features"="+aes,+fp-armv8,+fullfp16,+neon,+outline-atomics,+sve,+sve-aes,+sve2,+sve2-aes,+v8a"
195-
// CHECK: attributes #[[sve2_bitperm]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sve,+sve2,+sve2-bitperm,+v8a"
195+
// CHECK: attributes #[[sve2_bitperm]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sve,+sve-bitperm,+sve2,+sve2-bitperm,+v8a"
196196
// CHECK: attributes #[[sve2_sha3]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sha2,+sha3,+sve,+sve2,+sve2-sha3,+v8a"
197197
// CHECK: attributes #[[sve2_sm4]] = { {{.*}} "target-features"="+fp-armv8,+fullfp16,+neon,+outline-atomics,+sm4,+sve,+sve2,+sve2-sm4,+v8a"
198198
// CHECK: attributes #[[wfxt]] = { {{.*}} "target-features"="+fp-armv8,+neon,+outline-atomics,+v8a,+wfxt"

‎clang/test/CodeGen/AArch64/sve2-intrinsics/acle_sve2_bdep.c‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
22
// REQUIRES: aarch64-registered-target
33

4-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
4+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
88

99
#include <arm_sve.h>
1010

‎clang/test/CodeGen/AArch64/sve2-intrinsics/acle_sve2_bext.c‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
22
// REQUIRES: aarch64-registered-target
33

4-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
4+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
88

99
#include <arm_sve.h>
1010

‎clang/test/CodeGen/AArch64/sve2-intrinsics/acle_sve2_bgrp.c‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
22
// REQUIRES: aarch64-registered-target
33

4-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5-
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7-
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
4+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5+
// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
6+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
7+
// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve-bitperm -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
88

99
#include <arm_sve.h>
1010

‎clang/test/CodeGen/AArch64/targetattr.c‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -204,7 +204,7 @@ void applem4() {}
204204
// CHECK: attributes #[[ATTR1]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rdm,+sve,+v8.1a,+v8.2a,+v8a" }
205205
// CHECK: attributes #[[ATTR2]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+fp-armv8,+fullfp16,+lse,+neon,+ras,+rdm,+sve,+sve2,+v8.1a,+v8.2a,+v8a" }
206206
// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+bti,+ccidx,+complxnum,+crc,+dit,+dotprod,+flagm,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+neon,+pauth,+predres,+ras,+rcpc,+rdm,+sb,+ssbs,+sve,+sve2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8.6a,+v8a" }
207-
// CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a710" "target-features"="+bf16,+bti,+ccidx,+complxnum,+crc,+dit,+dotprod,+ete,+flagm,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+mte,+neon,+pauth,+perfmon,+predres,+ras,+rcpc,+rdm,+sb,+ssbs,+sve,+sve2,+sve2-bitperm,+trbe,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8a,+v9a" }
207+
// CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a710" "target-features"="+bf16,+bti,+ccidx,+complxnum,+crc,+dit,+dotprod,+ete,+flagm,+fp-armv8,+fp16fml,+fullfp16,+i8mm,+jsconv,+lse,+mte,+neon,+pauth,+perfmon,+predres,+ras,+rcpc,+rdm,+sb,+ssbs,+sve,+sve-bitperm,+sve2,+trbe,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8a,+v9a" }
208208
// CHECK: attributes #[[ATTR5]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "tune-cpu"="cortex-a710" }
209209
// CHECK: attributes #[[ATTR6]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" "target-features"="+ete,+fp-armv8,+neon,+trbe,+v8a" }
210210
// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "tune-cpu"="generic" }

‎clang/test/Driver/aarch64-implied-sme-features.c‎

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -51,4 +51,7 @@
5151
// SME-SUBFEATURE-CONFLICT-REV: "-target-feature" "+bf16"{{.*}} "-target-feature" "+sme" "-target-feature" "+sme-i16i64"
5252

5353
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+ssve-aes %s -### 2>&1 | FileCheck %s --check-prefix=SVE-AES
54-
// SVE-AES: "-target-feature" "+sme" "-target-feature" "+sme2" "-target-feature" "+ssve-aes" "-target-feature" "+sve-aes"
54+
// SVE-AES: "-target-feature" "+sme" "-target-feature" "+sme2" "-target-feature" "+ssve-aes" "-target-feature" "+sve-aes"
55+
56+
+// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+ssve-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=SVE-BITPERM
57+
+// SVE-BITPERM: "-target-feature" "+sme" "-target-feature" "+sme2" "-target-feature" "+ssve-bitperm" "-target-feature" "+sve-bitperm"

‎clang/test/Driver/aarch64-implied-sve-features.c‎

Lines changed: 11 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -23,17 +23,24 @@
2323
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve+sve2 %s -### 2>&1 | FileCheck %s --check-prefix=SVE-SVE2
2424
// SVE-SVE2: "-target-feature" "+sve" "-target-feature" "+sve2"
2525

26+
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=SVE-BITPERM
27+
// SVE-BITPERM: "-target-feature" "+sve-bitperm"
28+
2629
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve2-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=SVE2-BITPERM
27-
// SVE2-BITPERM: "-target-feature" "+sve" "-target-feature" "+sve2" "-target-feature" "+sve2-bitperm"
30+
// SVE2-BITPERM: "-target-feature" "+sve" "-target-feature" "+sve-bitperm" "-target-feature" "+sve2" "-target-feature" "+sve2-bitperm"
2831

2932
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+nosve2-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=NOSVE2-BITPERM
33+
// NOSVE2-BITPERM-NOT: "-target-feature" "+sve-bitperm"
3034
// NOSVE2-BITPERM-NOT: "-target-feature" "+sve2-bitperm"
3135
// NOSVE2-BITPERM-NOT: "-target-feature" "+sve2"
3236
// NOSVE2-BITPERM-NOT: "-target-feature" "+sve"
3337
// NOSVE2-BITPERM-NOT: sve2-bitperm"
3438

39+
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve-bitperm+nosve-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=SVE-BITPERM-REVERT
40+
// SVE-BITPERM-REVERT: "-target-feature" "-sve-bitperm"
41+
3542
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve2-bitperm+nosve2-bitperm %s -### 2>&1 | FileCheck %s --check-prefix=SVE2-BITPERM-REVERT
36-
// SVE2-BITPERM-REVERT: "-target-feature" "+sve" "-target-feature" "+sve2" "-target-feature" "-sve2-bitperm"
43+
// SVE2-BITPERM-REVERT: "-target-feature" "+sve" "-target-feature" "-sve-bitperm" "-target-feature" "-sve2" "-target-feature" "-sve2-bitperm"
3744

3845
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve2-aes+nosve2-aes %s -### 2>&1 | FileCheck %s --check-prefix=SVE2-AES-REVERT
3946
// SVE2-AES-REVERT: "-target-feature" "+sve" "-target-feature" "-sve-aes" "-target-feature" "+sve2" "-target-feature" "-sve2-aes"
@@ -57,7 +64,7 @@
5764
// SVE2-SM4: "-target-feature" "+sve" "-target-feature" "+sve2" "-target-feature" "+sve2-sm4"
5865

5966
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve2-bitperm+nosve2-aes %s -### 2>&1 | FileCheck %s --check-prefix=SVE2-SUBFEATURE-MIX
60-
// SVE2-SUBFEATURE-MIX: "-target-feature" "+sve" "-target-feature" "+sve2" "-target-feature" "+sve2-bitperm"
67+
// SVE2-SUBFEATURE-MIX: "-target-feature" "+sve" "-target-feature" "+sve-bitperm" "-target-feature" "+sve2" "-target-feature" "+sve2-bitperm"
6168
// SVE2-SUBFEATURE-NOT: sve2-aes
6269

6370
// RUN: %clang --target=aarch64-linux-gnu -march=armv8-a+sve2-sm4+nosve2 %s -### 2>&1 | FileCheck %s --check-prefix=SVE2-SUBFEATURE-CONFLICT
@@ -72,6 +79,7 @@
7279
// SVE-SUBFEATURE-CONFLICT-REV: "-target-feature" "+sve" "-target-feature" "+sve-aes" "-target-feature" "+sve2" "-target-feature" "+sve2-aes"
7380

7481
// RUN: %clang --target=aarch64-linux-gnu -mcpu=neoverse-n2+nosve2 %s -### 2>&1 | FileCheck %s --check-prefix=SVE-MCPU-FEATURES
82+
// SVE-MCPU-FEATURES-NOT: "-target-feature" "+sve-bitperm"
7583
// SVE-MCPU-FEATURES-NOT: "-target-feature" "+sve2-bitperm"
7684
// SVE-MCPU-FEATURES-NOT: "-target-feature" "+sve2"
7785
// SVE-MCPU-FEATURES: "-target-feature" "+sve"

0 commit comments

Comments
 (0)