[compiler-rt] Add cpu model init for Windows. (#111961)

Author: DanielKristofKiss

compiler-rt/lib/builtins/cpu_model/aarch64.c

@@ -76,6 +76,8 @@ struct {
 #elif defined(__linux__) && __has_include(<sys/auxv.h>)
 #include "aarch64/fmv/mrs.inc"
 #include "aarch64/fmv/getauxval.inc"
+#elif defined(_WIN32)
+#include "aarch64/fmv/windows.inc"
 #else
 #include "aarch64/fmv/unimplemented.inc"
 #endif

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

@@ -0,0 +1,85 @@
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+#include <processthreadsapi.h>
+#include <stdint.h>
+
+#ifndef PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE 43 
+#endif
+#ifndef PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE 44
+#endif
+#ifndef PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE 45
+#endif
+#ifndef PF_ARM_SVE_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_INSTRUCTIONS_AVAILABLE 46
+#endif
+#ifndef PF_ARM_SVE2_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE2_INSTRUCTIONS_AVAILABLE 47
+#endif
+#ifndef PF_ARM_SVE_PMULL128_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_PMULL128_INSTRUCTIONS_AVAILABLE 50
+#endif
+#ifndef PF_ARM_SVE_SHA3_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_SHA3_INSTRUCTIONS_AVAILABLE 55
+#endif
+#ifndef PF_ARM_SVE_SM4_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_SM4_INSTRUCTIONS_AVAILABLE 56
+#endif
+#ifndef PF_ARM_SVE_I8MM_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_I8MM_INSTRUCTIONS_AVAILABLE 57
+#endif
+#ifndef PF_ARM_SVE_F32MM_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_F32MM_INSTRUCTIONS_AVAILABLE 58
+#endif
+#ifndef PF_ARM_SVE_F64MM_INSTRUCTIONS_AVAILABLE
+#define PF_ARM_SVE_F64MM_INSTRUCTIONS_AVAILABLE 59
+#endif
+
+void __init_cpu_features_resolver(unsigned long hwcap,
+                                  const __ifunc_arg_t *arg) {}
+
+void CONSTRUCTOR_ATTRIBUTE __init_cpu_features(void) {
+  if (__atomic_load_n(&__aarch64_cpu_features.features, __ATOMIC_RELAXED))
+    return;
+
+#define setCPUFeature(F) features |= 1ULL << F
+
+  uint64_t features = 0;
+
+  setCPUFeature(FEAT_INIT);
+  setCPUFeature(FEAT_FP);
+
+  // https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-isprocessorfeaturepresent
+  if (IsProcessorFeaturePresent(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE)) {
+    setCPUFeature(FEAT_SHA2);
+    setCPUFeature(FEAT_PMULL);
+  }
+
+  static const struct ProcessFeatureToFeatMap_t {
+    int WinApiFeature;
+    enum CPUFeatures CPUFeature;
+  } FeatMap[] = {
+      {PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE, FEAT_CRC},
+      {PF_ARM_V81_ATOMIC_INSTRUCTIONS_AVAILABLE, FEAT_LSE},
+      {PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE, FEAT_DOTPROD},
+      {PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE, FEAT_JSCVT},
+      {PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE, FEAT_RCPC},
+      {PF_ARM_SVE_INSTRUCTIONS_AVAILABLE, FEAT_SVE},
+      {PF_ARM_SVE2_INSTRUCTIONS_AVAILABLE, FEAT_SVE2},
+      {PF_ARM_SVE_PMULL128_INSTRUCTIONS_AVAILABLE, FEAT_SVE_PMULL128},
+      {PF_ARM_SVE_SHA3_INSTRUCTIONS_AVAILABLE, FEAT_SVE_SHA3},
+      {PF_ARM_SVE_SM4_INSTRUCTIONS_AVAILABLE, FEAT_SVE_SM4},
+      {PF_ARM_SVE_F32MM_INSTRUCTIONS_AVAILABLE, FEAT_SVE_F32MM},
+      {PF_ARM_SVE_F64MM_INSTRUCTIONS_AVAILABLE, FEAT_SVE_F64MM},
+      // There is no I8MM flag, but when SVE_I8MM is available, I8MM is too.
+      {PF_ARM_SVE_I8MM_INSTRUCTIONS_AVAILABLE, FEAT_I8MM},
+  };
+
+  for (size_t I = 0, E = sizeof(FeatMap) / sizeof(FeatMap[0]); I != E; ++I)
+    if (IsProcessorFeaturePresent(FeatMap[I].WinApiFeature))
+      setCPUFeature(FeatMap[I].CPUFeature);
+
+  __atomic_store(&__aarch64_cpu_features.features, &features, __ATOMIC_RELAXED);
+}

compiler-rt/lib/builtins/cpu_model/cpu_model.h

@@ -31,7 +31,15 @@
 // We're choosing init priority 90 to force our constructors to run before any
 // constructors in the end user application (starting at priority 101). This
 // value matches the libgcc choice for the same functions.
-#define CONSTRUCTOR_ATTRIBUTE __attribute__((constructor(90)))
+#ifdef _WIN32
+// Contructor that replaces the ifunc runs currently with prio 10, see
+// the LowerIFuncPass. The resolver of FMV depends on the cpu features so set
+// the priority to 9.
+#define CONSTRUCTOR_PRIORITY 9
+#else
+#define CONSTRUCTOR_PRIORITY 90
+#endif
+#define CONSTRUCTOR_ATTRIBUTE __attribute__((constructor(CONSTRUCTOR_PRIORITY)))
 #else
 // FIXME: For MSVC, we should make a function pointer global in .CRT$X?? so that
 // this runs during initialization.