[compiler-rt][AArch64][FMV] Use the hw.optional.arm.caps fast path (#95275)

MacOS 15.0 and iOS 18.0 added a new sysctl to fetch a bitvector of all
the hw.optional.arm.FEAT_*'s in one go. Using this has a perf advantage
over doing multiple round-trips to the kernel and back, but since it's
not present in older oses, we still need the slow fallback.

Author: jroelofs

compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc

@@ -2,6 +2,11 @@
 #if TARGET_OS_OSX || TARGET_OS_IPHONE
 #include <sys/sysctl.h>
 
+#if __has_include(<arm/cpu_capabilities_public.h>)
+#include <arm/cpu_capabilities_public.h>
+#define HAS_CPU_CAPABILITIES_PUBLIC_H 1
+#endif
+
 static bool isKnownAndSupported(const char *name) {
   int32_t val = 0;
   size_t size = sizeof(val);
@@ -10,6 +15,19 @@ static bool isKnownAndSupported(const char *name) {
   return val;
 }
 
+static uint64_t deriveImplicitFeatures(uint64_t features) {
+  // FEAT_SSBS2 implies FEAT_SSBS
+  if ((1ULL << FEAT_SSBS2) & features)
+    features |= (1ULL << FEAT_SSBS);
+
+  // FEAT_FP is always enabled
+  features |= (1ULL << FEAT_FP);
+
+  features |= (1ULL << FEAT_INIT);
+
+  return features;
+}
+
 void __init_cpu_features_resolver(void) {
   // On Darwin platforms, this may be called concurrently by multiple threads
   // because the resolvers that use it are called lazily at runtime (unlike on
@@ -21,6 +39,62 @@ void __init_cpu_features_resolver(void) {
 
   uint64_t features = 0;
 
+#ifdef HAS_CPU_CAPABILITIES_PUBLIC_H
+  uint8_t feats_bitvec[(CAP_BIT_NB + 7) / 8] = {0};
+  size_t len = sizeof(feats_bitvec);
+  // When hw.optional.arm.feats is available (macOS 15.0+, iOS 18.0+), use the
+  // fast path to get all the feature bits, otherwise fall back to the slow
+  // ~20-something sysctls path.
+  if (!sysctlbyname("hw.optional.arm.caps", &feats_bitvec, &len, 0, 0)) {
+
+#define CHECK_BIT(FROM, TO)                                                    \
+  do {                                                                         \
+    if (feats_bitvec[FROM / 8] & (1u << ((FROM) & 7))) {                       \
+      features |= (1ULL << TO);                                                \
+    }                                                                          \
+  } while (0)
+
+    CHECK_BIT(CAP_BIT_FEAT_FlagM, FEAT_FLAGM);
+    CHECK_BIT(CAP_BIT_FEAT_FlagM2, FEAT_FLAGM2);
+    CHECK_BIT(CAP_BIT_FEAT_FHM, FEAT_FP16FML);
+    CHECK_BIT(CAP_BIT_FEAT_DotProd, FEAT_DOTPROD);
+    CHECK_BIT(CAP_BIT_FEAT_SHA3, FEAT_SHA3);
+    CHECK_BIT(CAP_BIT_FEAT_RDM, FEAT_RDM);
+    CHECK_BIT(CAP_BIT_FEAT_LSE, FEAT_LSE);
+    CHECK_BIT(CAP_BIT_FEAT_SHA256, FEAT_SHA2);
+    CHECK_BIT(CAP_BIT_FEAT_SHA1, FEAT_SHA1);
+    CHECK_BIT(CAP_BIT_FEAT_AES, FEAT_AES);
+    CHECK_BIT(CAP_BIT_FEAT_PMULL, FEAT_PMULL);
+    CHECK_BIT(CAP_BIT_FEAT_SPECRES, FEAT_PREDRES);
+    CHECK_BIT(CAP_BIT_FEAT_SB, FEAT_SB);
+    CHECK_BIT(CAP_BIT_FEAT_FRINTTS, FEAT_FRINTTS);
+    CHECK_BIT(CAP_BIT_FEAT_LRCPC, FEAT_RCPC);
+    CHECK_BIT(CAP_BIT_FEAT_LRCPC2, FEAT_RCPC2);
+    CHECK_BIT(CAP_BIT_FEAT_FCMA, FEAT_FCMA);
+    CHECK_BIT(CAP_BIT_FEAT_JSCVT, FEAT_JSCVT);
+    CHECK_BIT(CAP_BIT_FEAT_DPB, FEAT_DPB);
+    CHECK_BIT(CAP_BIT_FEAT_DPB2, FEAT_DPB2);
+    CHECK_BIT(CAP_BIT_FEAT_BF16, FEAT_BF16);
+    CHECK_BIT(CAP_BIT_FEAT_I8MM, FEAT_I8MM);
+    CHECK_BIT(CAP_BIT_FEAT_DIT, FEAT_DIT);
+    CHECK_BIT(CAP_BIT_FEAT_FP16, FEAT_FP16);
+    CHECK_BIT(CAP_BIT_FEAT_SSBS, FEAT_SSBS2);
+    CHECK_BIT(CAP_BIT_FEAT_BTI, FEAT_BTI);
+    CHECK_BIT(CAP_BIT_AdvSIMD, FEAT_SIMD);
+    CHECK_BIT(CAP_BIT_CRC32, FEAT_CRC);
+    CHECK_BIT(CAP_BIT_FEAT_SME, FEAT_SME);
+    CHECK_BIT(CAP_BIT_FEAT_SME2, FEAT_SME2);
+    CHECK_BIT(CAP_BIT_FEAT_SME_F64F64, FEAT_SME_F64);
+    CHECK_BIT(CAP_BIT_FEAT_SME_I16I64, FEAT_SME_I64);
+
+    features = deriveImplicitFeatures(features);
+
+    __atomic_store(&__aarch64_cpu_features.features, &features,
+                   __ATOMIC_RELAXED);
+    return;
+  }
+#endif
+
   // https://developer.apple.com/documentation/kernel/1387446-sysctlbyname/determining_instruction_set_characteristics
   static const struct {
     const char *sysctl_name;
@@ -32,7 +106,6 @@ void __init_cpu_features_resolver(void) {
       {"hw.optional.arm.FEAT_DotProd", FEAT_DOTPROD},
       {"hw.optional.arm.FEAT_RDM", FEAT_RDM},
       {"hw.optional.arm.FEAT_LSE", FEAT_LSE},
-      {"hw.optional.floatingpoint", FEAT_FP},
       {"hw.optional.AdvSIMD", FEAT_SIMD},
       {"hw.optional.armv8_crc32", FEAT_CRC},
       {"hw.optional.arm.FEAT_SHA1", FEAT_SHA1},
@@ -62,7 +135,7 @@ void __init_cpu_features_resolver(void) {
     if (isKnownAndSupported(feature_checks[I].sysctl_name))
       features |= (1ULL << feature_checks[I].feature);
 
-  features |= (1ULL << FEAT_INIT);
+  features = deriveImplicitFeatures(features);
 
   __atomic_store(&__aarch64_cpu_features.features, &features,
                   __ATOMIC_RELAXED);