Skip to content

Commit 02cafa8

Browse files
SpencerAbsontru
SpencerAbson
authored and
tru
committed
[AArch64] Adopt updated B16B16 target flags
The enablement of SVE/SME non-widening BFloat16 instructions was recently changed in response to an architecture update, in which: - FEAT_SVE_B16B16 was weakened - FEAT_SME_B16B16 was introduced New flags, 'sve-b16b16' and 'sme-b16b16' were introduced to replace the existing 'b16b16'. This was acheived in the below two patches. - #101480 - #102501 Ideally, the interface change introduced here will be valid in LLVM-19. We do not see it necessary to back-port the entire change, but just to add 'sme-b16b16' and 'sve-b16b16' as aliases to the existing (and unchanged) 'b16b16' and 'sme2' flags which together cover all of these features. The predication of Bf16 variants of svmin/svminnm and svmax/svmaxnm is also fixed in this change.
1 parent 9e90c40 commit 02cafa8

File tree

4 files changed

+46
-6
lines changed

4 files changed

+46
-6
lines changed

‎clang/include/clang/Basic/arm_sve.td

Lines changed: 21 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -2116,7 +2116,7 @@ def SVFCLAMP_BF : SInst<"svclamp[_{d}]", "dddd", "b", MergeNone, "aarch64_sve_
21162116
multiclass MinMaxIntr<string i, string zm, string mul, string t> {
21172117
def SVS # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "csil", MergeNone, "aarch64_sve_s" # i # zm # "_" # mul, [IsStreaming], []>;
21182118
def SVU # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "UcUsUiUl", MergeNone, "aarch64_sve_u" # i # zm # "_" # mul, [IsStreaming], []>;
2119-
def SVF # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "bhfd", MergeNone, "aarch64_sve_f" # i # zm # "_" # mul, [IsStreaming], []>;
2119+
def SVF # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "hfd", MergeNone, "aarch64_sve_f" # i # zm # "_" # mul, [IsStreaming], []>;
21202120
}
21212121

21222122
let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
@@ -2134,11 +2134,11 @@ let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
21342134
}
21352135

21362136
multiclass SInstMinMaxByVector<string name> {
2137-
def NAME # _SINGLE_X2 : SInst<"sv" # name # "nm[_single_{d}_x2]", "22d", "bhfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x2", [IsStreaming], []>;
2138-
def NAME # _SINGLE_X4 : SInst<"sv" # name # "nm[_single_{d}_x4]", "44d", "bhfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x4", [IsStreaming], []>;
2137+
def NAME # _SINGLE_X2 : SInst<"sv" # name # "nm[_single_{d}_x2]", "22d", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x2", [IsStreaming], []>;
2138+
def NAME # _SINGLE_X4 : SInst<"sv" # name # "nm[_single_{d}_x4]", "44d", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x4", [IsStreaming], []>;
21392139

2140-
def NAME # _X2 : SInst<"sv" # name # "nm[_{d}_x2]", "222", "bhfd", MergeNone, "aarch64_sve_f" # name # "nm_x2", [IsStreaming], []>;
2141-
def NAME # _X4 : SInst<"sv" # name # "nm[_{d}_x4]", "444", "bhfd", MergeNone, "aarch64_sve_f" # name # "nm_x4", [IsStreaming], []>;
2140+
def NAME # _X2 : SInst<"sv" # name # "nm[_{d}_x2]", "222", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_x2", [IsStreaming], []>;
2141+
def NAME # _X4 : SInst<"sv" # name # "nm[_{d}_x4]", "444", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_x4", [IsStreaming], []>;
21422142
}
21432143

21442144
let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
@@ -2172,9 +2172,25 @@ let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
21722172
def SVFCLAMP_X4 : SInst<"svclamp[_single_{d}_x4]", "44dd", "hfd", MergeNone, "aarch64_sve_fclamp_single_x4", [IsStreaming], []>;
21732173
}
21742174

2175+
multiclass BfSingleMultiVector<string name> {
2176+
def NAME # _SINGLE_X2 : SInst<"sv" # name # "[_single_{d}_x2]", "22d", "b", MergeNone, "aarch64_sve_f" # name # "_single_x2", [IsStreaming], []>;
2177+
def NAME # _SINGLE_X4 : SInst<"sv" # name # "[_single_{d}_x4]", "44d", "b", MergeNone, "aarch64_sve_f" # name # "_single_x4", [IsStreaming], []>;
2178+
2179+
def NAME # _X2 : SInst<"sv" # name # "[_{d}_x2]", "222", "b", MergeNone, "aarch64_sve_f" # name # "_x2", [IsStreaming], []>;
2180+
def NAME # _X4 : SInst<"sv" # name # "[_{d}_x4]", "444", "b", MergeNone, "aarch64_sve_f" # name # "_x4", [IsStreaming], []>;
2181+
}
2182+
21752183
let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2,b16b16"in {
21762184
def SVBFCLAMP_X2 : SInst<"svclamp[_single_{d}_x2]", "22dd", "b", MergeNone, "aarch64_sve_bfclamp_single_x2", [IsStreaming], []>;
21772185
def SVBFCLAMP_X4 : SInst<"svclamp[_single_{d}_x4]", "44dd", "b", MergeNone, "aarch64_sve_bfclamp_single_x4", [IsStreaming], []>;
2186+
2187+
// bfmin, bfmax (single, multi)
2188+
defm SVBFMIN : BfSingleMultiVector<"min">;
2189+
defm SVBFMAX : BfSingleMultiVector<"max">;
2190+
2191+
// bfminnm, bfmaxnm (single, multi)
2192+
defm SVBFMINNM : BfSingleMultiVector<"minnm">;
2193+
defm SVBFMAXNM : BfSingleMultiVector<"maxnm">;
21782194
}
21792195

21802196
let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {

‎clang/test/Driver/print-supported-extensions-aarch64.c

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -55,6 +55,7 @@
5555
// CHECK-NEXT: sha3 FEAT_SHA3, FEAT_SHA512 Enable SHA512 and SHA3 support
5656
// CHECK-NEXT: sm4 FEAT_SM4, FEAT_SM3 Enable SM3 and SM4 support
5757
// CHECK-NEXT: sme FEAT_SME Enable Scalable Matrix Extension (SME)
58+
// CHECK-NEXT: sme-b16b16 FEAT_SME_B16B16 Enable SME2.1 ZA-targeting non-widening BFloat16 instructions
5859
// CHECK-NEXT: sme-f16f16 FEAT_SME_F16F16 Enable SME non-widening Float16 instructions
5960
// CHECK-NEXT: sme-f64f64 FEAT_SME_F64F64 Enable Scalable Matrix Extension (SME) F64F64 instructions
6061
// CHECK-NEXT: sme-f8f16 FEAT_SME_F8F16 Enable Scalable Matrix Extension (SME) F8F16 instructions
@@ -71,6 +72,7 @@
7172
// CHECK-NEXT: ssve-fp8dot4 FEAT_SSVE_FP8DOT4 Enable SVE2 FP8 4-way dot product instructions
7273
// CHECK-NEXT: ssve-fp8fma FEAT_SSVE_FP8FMA Enable SVE2 FP8 multiply-add instructions
7374
// CHECK-NEXT: sve FEAT_SVE Enable Scalable Vector Extension (SVE) instructions
75+
// CHECK-NEXT: sve-b16b16 FEAT_SVE_B16B16 Enable SVE2 non-widening and SME2 Z-targeting non-widening BFloat16 instructions
7476
// CHECK-NEXT: sve2 FEAT_SVE2 Enable Scalable Vector Extension 2 (SVE2) instructions
7577
// CHECK-NEXT: sve2-aes FEAT_SVE_AES, FEAT_SVE_PMULL128 Enable AES SVE2 instructions
7678
// CHECK-NEXT: sve2-bitperm FEAT_SVE_BitPerm Enable bit permutation SVE2 instructions

‎llvm/lib/Target/AArch64/AArch64Features.td

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -438,6 +438,15 @@ def FeatureSVE2p1: ExtensionWithMArch<"sve2p1", "SVE2p1", "FEAT_SVE2p1",
438438
def FeatureB16B16 : ExtensionWithMArch<"b16b16", "B16B16", "FEAT_SVE_B16B16",
439439
"Enable SVE2.1 or SME2.1 non-widening BFloat16 to BFloat16 instructions", [FeatureBF16]>;
440440

441+
// FeatureSVEB16B16 and FeatureSMEB16B16 act as aliases for {FeatureB16B16}, and
442+
// {FeatureB16B16, FeatureSME2} respectively. This allows LLVM-20 interfacing programs
443+
// that use '+sve-b16b16' and '+sme-b16b16' to compile in LLVM-19.
444+
def FeatureSVEB16B16 : ExtensionWithMArch<"sve-b16b16", "SVEB16B16", "FEAT_SVE_B16B16",
445+
"Enable SVE2 non-widening and SME2 Z-targeting non-widening BFloat16 instructions", [FeatureB16B16]>;
446+
447+
def FeatureSMEB16B16 : ExtensionWithMArch<"sme-b16b16", "SMEB16B16", "FEAT_SME_B16B16",
448+
"Enable SME2.1 ZA-targeting non-widening BFloat16 instructions", [FeatureSME2, FeatureB16B16]>;
449+
441450
def FeatureSMEF16F16 : ExtensionWithMArch<"sme-f16f16", "SMEF16F16", "FEAT_SME_F16F16",
442451
"Enable SME non-widening Float16 instructions", [FeatureSME2]>;
443452

‎llvm/unittests/TargetParser/TargetParserTest.cpp

Lines changed: 14 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -2005,6 +2005,7 @@ TEST(TargetParserTest, AArch64ExtensionFeatures) {
20052005
AArch64::AEK_CPA, AArch64::AEK_PAUTHLR,
20062006
AArch64::AEK_TLBIW, AArch64::AEK_JSCVT,
20072007
AArch64::AEK_FCMA, AArch64::AEK_FP8,
2008+
AArch64::AEK_SMEB16B16, AArch64::AEK_SVEB16B16,
20082009

20092010
};
20102011

@@ -2043,6 +2044,7 @@ TEST(TargetParserTest, AArch64ExtensionFeatures) {
20432044
EXPECT_TRUE(llvm::is_contained(Features, "+sve2-sha3"));
20442045
EXPECT_TRUE(llvm::is_contained(Features, "+sve2-bitperm"));
20452046
EXPECT_TRUE(llvm::is_contained(Features, "+sve2p1"));
2047+
EXPECT_TRUE(llvm::is_contained(Features, "+sve-b16b16"));
20462048
EXPECT_TRUE(llvm::is_contained(Features, "+b16b16"));
20472049
EXPECT_TRUE(llvm::is_contained(Features, "+rcpc"));
20482050
EXPECT_TRUE(llvm::is_contained(Features, "+rand"));
@@ -2063,6 +2065,7 @@ TEST(TargetParserTest, AArch64ExtensionFeatures) {
20632065
EXPECT_TRUE(llvm::is_contained(Features, "+sme-f64f64"));
20642066
EXPECT_TRUE(llvm::is_contained(Features, "+sme-i16i64"));
20652067
EXPECT_TRUE(llvm::is_contained(Features, "+sme-f16f16"));
2068+
EXPECT_TRUE(llvm::is_contained(Features, "+sme-b16b16"));
20662069
EXPECT_TRUE(llvm::is_contained(Features, "+sme2"));
20672070
EXPECT_TRUE(llvm::is_contained(Features, "+sme2p1"));
20682071
EXPECT_TRUE(llvm::is_contained(Features, "+hbc"));
@@ -2188,6 +2191,7 @@ TEST(TargetParserTest, AArch64ArchExtFeature) {
21882191
{"lse", "nolse", "+lse", "-lse"},
21892192
{"rdm", "nordm", "+rdm", "-rdm"},
21902193
{"sve", "nosve", "+sve", "-sve"},
2194+
{"sve-b16b16", "nosve-b16b16", "+sve-b16b16", "-sve-b16b16"},
21912195
{"sve2", "nosve2", "+sve2", "-sve2"},
21922196
{"sve2-aes", "nosve2-aes", "+sve2-aes", "-sve2-aes"},
21932197
{"sve2-sm4", "nosve2-sm4", "+sve2-sm4", "-sve2-sm4"},
@@ -2212,6 +2216,7 @@ TEST(TargetParserTest, AArch64ArchExtFeature) {
22122216
{"sme-f64f64", "nosme-f64f64", "+sme-f64f64", "-sme-f64f64"},
22132217
{"sme-i16i64", "nosme-i16i64", "+sme-i16i64", "-sme-i16i64"},
22142218
{"sme-f16f16", "nosme-f16f16", "+sme-f16f16", "-sme-f16f16"},
2219+
{"sme-b16b16", "nosme-b16b16", "+sme-b16b16", "-sme-b16b16"},
22152220
{"sme2", "nosme2", "+sme2", "-sme2"},
22162221
{"sme2p1", "nosme2p1", "+sme2p1", "-sme2p1"},
22172222
{"hbc", "nohbc", "+hbc", "-hbc"},
@@ -2452,6 +2457,12 @@ AArch64ExtensionDependenciesBaseArchTestParams
24522457
{AArch64::ARMV8A, {"nobf16", "b16b16"}, {"bf16", "b16b16"}, {}},
24532458
{AArch64::ARMV8A, {"b16b16", "nobf16"}, {}, {"bf16", "b16b16"}},
24542459

2460+
// b16b16 -> {sve-b16b16, sme-b16b16}
2461+
{AArch64::ARMV8A, {"nob16b16", "sve-b16b16"}, {"b16b16", "sve-b16b16"}, {}},
2462+
{AArch64::ARMV8A, {"sve-b16b16", "nob16b16"}, {}, {"sve-b16b16", "b16b16"}},
2463+
{AArch64::ARMV8A, {"nob16b16", "sme-b16b16"}, {"b16b16", "sme-b16b16"}, {}},
2464+
{AArch64::ARMV8A, {"sme-b16b16", "nob16b16"}, {}, {"b16b16", "sme-b16b16"}},
2465+
24552466
// sve -> {sve2, f32mm, f64mm}
24562467
{AArch64::ARMV8A, {"nosve", "sve2"}, {"sve", "sve2"}, {}},
24572468
{AArch64::ARMV8A, {"sve2", "nosve"}, {}, {"sve", "sve2"}},
@@ -2491,7 +2502,7 @@ AArch64ExtensionDependenciesBaseArchTestParams
24912502
{AArch64::ARMV8A, {"sme-fa64", "nosme"}, {}, {"sme", "sme-fa64"}},
24922503

24932504
// sme2 -> {sme2p1, ssve-fp8fma, ssve-fp8dot2, ssve-fp8dot4, sme-f8f16,
2494-
// sme-f8f32}
2505+
// sme-f8f32, sme-b16b16}
24952506
{AArch64::ARMV8A, {"nosme2", "sme2p1"}, {"sme2", "sme2p1"}, {}},
24962507
{AArch64::ARMV8A, {"sme2p1", "nosme2"}, {}, {"sme2", "sme2p1"}},
24972508
{AArch64::ARMV8A,
@@ -2522,6 +2533,8 @@ AArch64ExtensionDependenciesBaseArchTestParams
25222533
{AArch64::ARMV8A, {"sme-f8f16", "nosme2"}, {}, {"sme2", "sme-f8f16"}},
25232534
{AArch64::ARMV8A, {"nosme2", "sme-f8f32"}, {"sme2", "sme-f8f32"}, {}},
25242535
{AArch64::ARMV8A, {"sme-f8f32", "nosme2"}, {}, {"sme2", "sme-f8f32"}},
2536+
{AArch64::ARMV8A, {"nosme2", "sme-b16b16"}, {"sme2", "sme-b16b16"}, {}},
2537+
{AArch64::ARMV8A, {"sme-b16b16", "nosme2"}, {}, {"sme2", "sme-b16b16"}},
25252538

25262539
// fp8 -> {sme-f8f16, sme-f8f32}
25272540
{AArch64::ARMV8A, {"nofp8", "sme-f8f16"}, {"fp8", "sme-f8f16"}, {}},

0 commit comments

Comments
 (0)