[AArch64][ARM][Clang] PerfMon Extension Added

[mubashar_]

Summary

Performance monitor architecture extension has been added to clang.

[lenary]

I'm happy with this patch, but you could update the description to be clearer that you're adding an option to allow PMU v3.4 to be enabled/disabled separately to the architecture - not adding the pmu extension itself.

[dmgreen]

Hmm. This appears to be mapping pmuv3p4 to "perfmon". But "perfmon" has been around a long time, not a new feature related to the pmuv3p4 update. It seems on the AArch64 side to control access to PMCCNTR, i.e. access to the base FEAT_PMUv3.

On the Arm side this is altering how HasV7Ops is define in https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/ARM/ARM.td#L515, which in turn is breaking Cortex-M cpus. It's probably a bug that HasV7Ops includes FeaturePerfMon, it shouldn't be included in HasV8MMainlineOps or ARMv7m if it's reading a system register with an mrc.

[mubashar_]

The command line option for Clang is now +pmuv3 to enable PerfMon.

[dmgreen]

LGTM. Thanks for the update.

llvm/include/llvm/Support/AArch64TargetParser.def

147

The formatting is a little off, for the subsequent columns.

clang/test/Driver/aarch64-perfmon.c

+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.4a+pmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-PERFMON %s
+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.2a+pmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-PERFMON %s
+// CHECK-PERFMON: "-target-feature" "+perfmon"
+
+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.4a+nopmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-NOPERFMON %s
+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.2a+nopmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-NOPERFMON %s
+// CHECK-NOPERFMON: "-target-feature" "-perfmon"
+
+// RUN: %clang -### -target aarch64-none-none-eabi                 %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.4a %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// RUN: %clang -### -target aarch64-none-none-eabi -march=armv8.2a %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// ABSENTPERFMON-NOT: "-target-feature" "+perfmon"
+// ABSENTPERFMON-NOT: "-target-feature" "-perfmon"
 No newline at end of file

clang/test/Driver/arm-perfmon.c

+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.4a+pmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-PERFMON %s
+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.2a+pmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-PERFMON %s
+// CHECK-PERFMON: "-target-feature" "+perfmon"
+
+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.4a+nopmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-NOPERFMON %s
+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.2a+nopmuv3p4 %s 2>&1 | FileCheck --check-prefix=CHECK-NOPERFMON %s
+// CHECK-NOPERFMON: "-target-feature" "-perfmon"
+
+// RUN: %clang -### -target arm-none-none-eabi                 %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.4a %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// RUN: %clang -### -target arm-none-none-eabi -march=armv8.2a %s 2>&1 | FileCheck %s --check-prefix=ABSENTPERFMON
+// ABSENTPERFMON-NOT: "-target-feature" "+perfmon"
+// ABSENTPERFMON-NOT: "-target-feature" "-perfmon"
 No newline at end of file

llvm/include/llvm/Support/AArch64TargetParser.h

 namespace llvm {
 
 class Triple;
 
 namespace AArch64 {
 
 // Arch extension modifiers for CPUs.
 enum ArchExtKind : uint64_t {
   AEK_INVALID =     0,
   AEK_NONE =        1,
   AEK_CRC =         1 << 1,
   AEK_CRYPTO =      1 << 2,
   AEK_FP =          1 << 3,
   AEK_SIMD =        1 << 4,
   AEK_FP16 =        1 << 5,
   AEK_PROFILE =     1 << 6,
   AEK_RAS =         1 << 7,
enum ArchExtKind : uint64_t {
   AEK_SSBS =        1 << 20,
   AEK_SB =          1 << 21,
   AEK_PREDRES =     1 << 22,
   AEK_SVE2 =        1 << 23,
   AEK_SVE2AES =     1 << 24,
   AEK_SVE2SM4 =     1 << 25,
   AEK_SVE2SHA3 =    1 << 26,
   AEK_SVE2BITPERM = 1 << 27,
   AEK_TME =         1 << 28,
   AEK_BF16 =        1 << 29,
   AEK_I8MM =        1 << 30,
   AEK_F32MM =       1ULL << 31,
   AEK_F64MM =       1ULL << 32,
   AEK_LS64 =        1ULL << 33,
   AEK_BRBE =        1ULL << 34,
   AEK_PAUTH =       1ULL << 35,
   AEK_FLAGM =       1ULL << 36,
   AEK_SME =         1ULL << 37,
   AEK_SMEF64 =      1ULL << 38,
   AEK_SMEI64 =      1ULL << 39,
+  AEK_PERFMON =     1ULL << 40,
 };
 
 enum class ArchKind {
 #define AARCH64_ARCH(NAME, ID, CPU_ATTR, SUB_ARCH, ARCH_ATTR, ARCH_FPU, ARCH_BASE_EXT) ID,
 #include "AArch64TargetParser.def"
 };
 
 const ARM::ArchNames<ArchKind> AArch64ARCHNames[] = {

llvm/include/llvm/Support/AArch64TargetParser.def

 AARCH64_ARCH_EXT_NAME("tme",          AArch64::AEK_TME,         "+tme",   "-tme")
 AARCH64_ARCH_EXT_NAME("ls64",         AArch64::AEK_LS64,        "+ls64",  "-ls64")
 AARCH64_ARCH_EXT_NAME("brbe",         AArch64::AEK_BRBE,        "+brbe",  "-brbe")
 AARCH64_ARCH_EXT_NAME("pauth",        AArch64::AEK_PAUTH,       "+pauth", "-pauth")
 AARCH64_ARCH_EXT_NAME("flagm",        AArch64::AEK_FLAGM,       "+flagm", "-flagm")
 AARCH64_ARCH_EXT_NAME("sme",          AArch64::AEK_SME,         "+sme",   "-sme")
 AARCH64_ARCH_EXT_NAME("sme-f64",      AArch64::AEK_SMEF64,      "+sme-f64", "-sme-f64")
 AARCH64_ARCH_EXT_NAME("sme-i64",      AArch64::AEK_SMEI64,      "+sme-i64", "-sme-i64")
+AARCH64_ARCH_EXT_NAME("pmuv3p4",      AArch64::AEK_PERFMON,     "+perfmon", "-perfmon")
 #undef AARCH64_ARCH_EXT_NAME
 
 #ifndef AARCH64_CPU_NAME
 #define AARCH64_CPU_NAME(NAME, ID, DEFAULT_FPU, IS_DEFAULT, DEFAULT_EXT)
 #endif
 AARCH64_CPU_NAME("cortex-a34", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC))
 AARCH64_CPU_NAME("cortex-a35", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,

llvm/include/llvm/Support/ARMTargetParser.h

 
 class Triple;
 
 namespace ARM {
 
 // Arch extension modifiers for CPUs.
 // Note that this is not the same as the AArch64 list
 enum ArchExtKind : uint64_t {
   AEK_INVALID =     0,
   AEK_NONE =        1,
   AEK_CRC =         1 << 1,
   AEK_CRYPTO =      1 << 2,
   AEK_FP =          1 << 3,
   AEK_HWDIVTHUMB =  1 << 4,
   AEK_HWDIVARM =    1 << 5,
   AEK_MP =          1 << 6,
   AEK_SIMD =        1 << 7,
enum ArchExtKind : uint64_t {
   AEK_CDECP1 =      1 << 23,
   AEK_CDECP2 =      1 << 24,
   AEK_CDECP3 =      1 << 25,
   AEK_CDECP4 =      1 << 26,
   AEK_CDECP5 =      1 << 27,
   AEK_CDECP6 =      1 << 28,
   AEK_CDECP7 =      1 << 29,
   AEK_PACBTI =      1 << 30,
+  AEK_PERFMON =     1ULL << 31,
   // Unsupported extensions.
   AEK_OS       =    1ULL << 59,
   AEK_IWMMXT   =    1ULL << 60,
   AEK_IWMMXT2  =    1ULL << 61,
   AEK_MAVERICK =    1ULL << 62,
   AEK_XSCALE   =    1ULL << 63,
 };
 
 // List of Arch Extension names.
 // FIXME: TableGen this.

llvm/include/llvm/Support/ARMTargetParser.def

 ARM_ARCH_EXT_NAME("cdecp1",   ARM::AEK_CDECP1,   "+cdecp1",  "-cdecp1")
 ARM_ARCH_EXT_NAME("cdecp2",   ARM::AEK_CDECP2,   "+cdecp2",  "-cdecp2")
 ARM_ARCH_EXT_NAME("cdecp3",   ARM::AEK_CDECP3,   "+cdecp3",  "-cdecp3")
 ARM_ARCH_EXT_NAME("cdecp4",   ARM::AEK_CDECP4,   "+cdecp4",  "-cdecp4")
 ARM_ARCH_EXT_NAME("cdecp5",   ARM::AEK_CDECP5,   "+cdecp5",  "-cdecp5")
 ARM_ARCH_EXT_NAME("cdecp6",   ARM::AEK_CDECP6,   "+cdecp6",  "-cdecp6")
 ARM_ARCH_EXT_NAME("cdecp7",   ARM::AEK_CDECP7,   "+cdecp7",  "-cdecp7")
 ARM_ARCH_EXT_NAME("pacbti",   ARM::AEK_PACBTI,   "+pacbti",  "-pacbti")
+ARM_ARCH_EXT_NAME("pmuv3p4",  ARM::AEK_PERFMON,  "+perfmon", "-perfmon")
 #undef ARM_ARCH_EXT_NAME
 
 #ifndef ARM_HW_DIV_NAME
 #define ARM_HW_DIV_NAME(NAME, ID)
 #endif
 ARM_HW_DIV_NAME("invalid", ARM::AEK_INVALID)
 ARM_HW_DIV_NAME("none", ARM::AEK_NONE)
 ARM_HW_DIV_NAME("thumb", ARM::AEK_HWDIVTHUMB)

llvm/lib/Support/AArch64TargetParser.cpp

bool AArch64::getExtensionFeatures(uint64_t Extensions,
   if (Extensions & AEK_FLAGM)
     Features.push_back("+flagm");
   if (Extensions & AArch64::AEK_SME)
     Features.push_back("+sme");
   if (Extensions & AArch64::AEK_SMEF64)
     Features.push_back("+sme-f64");
   if (Extensions & AArch64::AEK_SMEI64)
     Features.push_back("+sme-i64");
+  if (Extensions & AArch64::AEK_PERFMON)
+    Features.push_back("+perfmon");
 
   return true;
 }
 
 bool AArch64::getArchFeatures(AArch64::ArchKind AK,
                               std::vector<StringRef> &Features) {
   if (AK == ArchKind::ARMV8A)
     Features.push_back("+v8a");

llvm/unittests/Support/TargetParserTest.cpp

const char *ArchExt[][4] = {{"crc", "nocrc", "+crc", "-crc"},
                               {"os", "noos", nullptr, nullptr},
                               {"iwmmxt", "noiwmmxt", nullptr, nullptr},
                               {"iwmmxt2", "noiwmmxt2", nullptr, nullptr},
                               {"maverick", "maverick", nullptr, nullptr},
                               {"xscale", "noxscale", nullptr, nullptr},
                               {"sb", "nosb", "+sb", "-sb"},
                               {"i8mm", "noi8mm", "+i8mm", "-i8mm"},
                               {"mve", "nomve", "+mve", "-mve"},
-                              {"mve.fp", "nomve.fp", "+mve.fp", "-mve.fp"}};
+                              {"mve.fp", "nomve.fp", "+mve.fp", "-mve.fp"},
+                              {"pmuv3p4", "nopmuv3p4", "+perfmon", "-perfmon"}};
 
   for (unsigned i = 0; i < array_lengthof(ArchExt); i++) {
     EXPECT_EQ(StringRef(ArchExt[i][2]), ARM::getArchExtFeature(ArchExt[i][0]));
     EXPECT_EQ(StringRef(ArchExt[i][3]), ARM::getArchExtFeature(ArchExt[i][1]));
   }
 }
 
 static bool
TEST(TargetParserTest, testAArch64Extension) {
   EXPECT_FALSE(testAArch64Extension(
       "generic", AArch64::ArchKind::ARMV8_4A, "fp16"));
   EXPECT_FALSE(testAArch64Extension(
       "generic", AArch64::ArchKind::ARMV8_4A, "fp16fml"));
 }
 
 TEST(TargetParserTest, AArch64ExtensionFeatures) {
   std::vector<uint64_t> Extensions = {
     AArch64::AEK_CRC,      AArch64::AEK_CRYPTO,
     AArch64::AEK_FP,       AArch64::AEK_SIMD,
     AArch64::AEK_FP16,     AArch64::AEK_PROFILE,
     AArch64::AEK_RAS,      AArch64::AEK_LSE,
     AArch64::AEK_RDM,      AArch64::AEK_DOTPROD,
     AArch64::AEK_SVE,      AArch64::AEK_SVE2,
     AArch64::AEK_SVE2AES,  AArch64::AEK_SVE2SM4,
     AArch64::AEK_SVE2SHA3, AArch64::AEK_SVE2BITPERM,
     AArch64::AEK_RCPC,     AArch64::AEK_FP16FML,
     AArch64::AEK_SME,      AArch64::AEK_SMEF64,
-    AArch64::AEK_SMEI64 };
+    AArch64::AEK_SMEI64,   AArch64::AEK_PERFMON};
 
   std::vector<StringRef> Features;
 
   uint64_t ExtVal = 0;
   for (auto Ext : Extensions)
     ExtVal |= Ext;
 
   EXPECT_FALSE(AArch64::getExtensionFeatures(AArch64::AEK_INVALID, Features));
TEST(TargetParserTest, AArch64ExtensionFeatures) {
   EXPECT_TRUE(llvm::is_contained(Features, "+sve2"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sve2-aes"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sve2-sm4"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sve2-sha3"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sve2-bitperm"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sme"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sme-f64"));
   EXPECT_TRUE(llvm::is_contained(Features, "+sme-i64"));
+  EXPECT_TRUE(llvm::is_contained(Features, "+perfmon"));
 }
 
 TEST(TargetParserTest, AArch64ArchFeatures) {
   std::vector<StringRef> Features;
 
   for (auto AK : AArch64::ArchKinds)
     EXPECT_TRUE((AK == AArch64::ArchKind::INVALID)
                     ? !AArch64::getArchFeatures(AK, Features)
                     : AArch64::getArchFeatures(AK, Features));
 }
 
 TEST(TargetParserTest, AArch64ArchExtFeature) {
   const char *ArchExt[][4] = {{"crc", "nocrc", "+crc", "-crc"},
                               {"crypto", "nocrypto", "+crypto", "-crypto"},
                               {"flagm", "noflagm", "+flagm", "-flagm"},
                               {"fp", "nofp", "+fp-armv8", "-fp-armv8"},
                               {"simd", "nosimd", "+neon", "-neon"},
                               {"fp16", "nofp16", "+fullfp16", "-fullfp16"},
                               {"fp16fml", "nofp16fml", "+fp16fml", "-fp16fml"},
                               {"profile", "noprofile", "+spe", "-spe"},
                               {"ras", "noras", "+ras", "-ras"},
const char *ArchExt[][4] = {{"crc", "nocrc", "+crc", "-crc"},
                               {"sb", "nosb", "+sb", "-sb"},
                               {"predres", "nopredres", "+predres", "-predres"},
                               {"i8mm", "noi8mm", "+i8mm", "-i8mm"},
                               {"f32mm", "nof32mm", "+f32mm", "-f32mm"},
                               {"f64mm", "nof64mm", "+f64mm", "-f64mm"},
                               {"sme", "nosme", "+sme", "-sme"},
                               {"sme-f64", "nosme-f64", "+sme-f64", "-sme-f64"},
                               {"sme-i64", "nosme-i64", "+sme-i64", "-sme-i64"},
+                              {"pmuv3p4", "nopmuv3p4", "+perfmon", "-perfmon"},
 };
 
   for (unsigned i = 0; i < array_lengthof(ArchExt); i++) {
     EXPECT_EQ(StringRef(ArchExt[i][2]),
               AArch64::getArchExtFeature(ArchExt[i][0]));
     EXPECT_EQ(StringRef(ArchExt[i][3]),
               AArch64::getArchExtFeature(ArchExt[i][1]));
   }
 }
 
 } // namespace