1 files changed, 73 insertions, 21 deletions

diff --git a/C/CpuArch.c b/C/CpuArch.c
index 33f8a3a..d51b38a 100644
--- a/C/CpuArch.c
+++ b/C/CpuArch.c
@@ -1,5 +1,5 @@
 /* CpuArch.c -- CPU specific code
-2023-05-18 : Igor Pavlov : Public domain */
+2024-03-02 : Igor Pavlov : Public domain */
 
 #include "Precomp.h"
 
@@ -226,7 +226,7 @@ void __declspec(naked) Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
 DON'T remove Z7_NO_INLINE and Z7_FASTCALL for MY_cpuidex_HACK(): !!!
 */
 static
-Z7_NO_INLINE void Z7_FASTCALL MY_cpuidex_HACK(UInt32 subFunction, UInt32 func, int *CPUInfo)
+Z7_NO_INLINE void Z7_FASTCALL MY_cpuidex_HACK(Int32 subFunction, Int32 func, Int32 *CPUInfo)
 {
   UNUSED_VAR(subFunction)
   __cpuid(CPUInfo, func);
@@ -242,13 +242,13 @@ Z7_NO_INLINE
 #endif
 void Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
 {
-  MY_cpuidex((int *)p, (int)func, 0);
+  MY_cpuidex((Int32 *)p, (Int32)func, 0);
 }
 
 Z7_NO_INLINE
 UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
 {
-  int a[4];
+  Int32 a[4];
   MY_cpuidex(a, 0, 0);
   return a[0];
 }
@@ -384,7 +384,7 @@ BoolInt CPU_IsSupported_CMOV(void)
   UInt32 a[4];
   if (!x86cpuid_Func_1(&a[0]))
     return 0;
-  return (a[3] >> 15) & 1;
+  return (BoolInt)(a[3] >> 15) & 1;
 }
 
 BoolInt CPU_IsSupported_SSE(void)
@@ -393,7 +393,7 @@ BoolInt CPU_IsSupported_SSE(void)
   CHECK_SYS_SSE_SUPPORT
   if (!x86cpuid_Func_1(&a[0]))
     return 0;
-  return (a[3] >> 25) & 1;
+  return (BoolInt)(a[3] >> 25) & 1;
 }
 
 BoolInt CPU_IsSupported_SSE2(void)
@@ -402,7 +402,7 @@ BoolInt CPU_IsSupported_SSE2(void)
   CHECK_SYS_SSE_SUPPORT
   if (!x86cpuid_Func_1(&a[0]))
     return 0;
-  return (a[3] >> 26) & 1;
+  return (BoolInt)(a[3] >> 26) & 1;
 }
 
 #endif
@@ -419,17 +419,17 @@ static UInt32 x86cpuid_Func_1_ECX(void)
 
 BoolInt CPU_IsSupported_AES(void)
 {
-  return (x86cpuid_Func_1_ECX() >> 25) & 1;
+  return (BoolInt)(x86cpuid_Func_1_ECX() >> 25) & 1;
 }
 
 BoolInt CPU_IsSupported_SSSE3(void)
 {
-  return (x86cpuid_Func_1_ECX() >> 9) & 1;
+  return (BoolInt)(x86cpuid_Func_1_ECX() >> 9) & 1;
 }
 
 BoolInt CPU_IsSupported_SSE41(void)
 {
-  return (x86cpuid_Func_1_ECX() >> 19) & 1;
+  return (BoolInt)(x86cpuid_Func_1_ECX() >> 19) & 1;
 }
 
 BoolInt CPU_IsSupported_SHA(void)
@@ -441,7 +441,7 @@ BoolInt CPU_IsSupported_SHA(void)
   {
     UInt32 d[4];
     z7_x86_cpuid(d, 7);
-    return (d[1] >> 29) & 1;
+    return (BoolInt)(d[1] >> 29) & 1;
   }
 }
 
@@ -640,8 +640,8 @@ BoolInt CPU_IsSupported_AVX(void)
     const UInt32 bm = (UInt32)x86_xgetbv_0(MY_XCR_XFEATURE_ENABLED_MASK);
     // printf("\n=== XGetBV=%d\n", bm);
     return 1
-        & (bm >> 1)  // SSE state is supported (set by OS) for storing/restoring
-        & (bm >> 2); // AVX state is supported (set by OS) for storing/restoring
+        & (BoolInt)(bm >> 1)  // SSE state is supported (set by OS) for storing/restoring
+        & (BoolInt)(bm >> 2); // AVX state is supported (set by OS) for storing/restoring
   }
   // since Win7SP1: we can use GetEnabledXStateFeatures();
 }
@@ -658,10 +658,29 @@ BoolInt CPU_IsSupported_AVX2(void)
     z7_x86_cpuid(d, 7);
     // printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
     return 1
-      & (d[1] >> 5); // avx2
+      & (BoolInt)(d[1] >> 5); // avx2
   }
 }
 
+/*
+// fix it:
+BoolInt CPU_IsSupported_AVX512F_AVX512VL(void)
+{
+  if (!CPU_IsSupported_AVX())
+    return False;
+  if (z7_x86_cpuid_GetMaxFunc() < 7)
+    return False;
+  {
+    UInt32 d[4];
+    z7_x86_cpuid(d, 7);
+    // printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
+    return 1
+      & (BoolInt)(d[1] >> 16)  // avx512-f
+      & (BoolInt)(d[1] >> 31); // avx512-Vl
+  }
+}
+*/
+
 BoolInt CPU_IsSupported_VAES_AVX2(void)
 {
   if (!CPU_IsSupported_AVX())
@@ -673,9 +692,9 @@ BoolInt CPU_IsSupported_VAES_AVX2(void)
     z7_x86_cpuid(d, 7);
     // printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
     return 1
-      & (d[1] >> 5) // avx2
+      & (BoolInt)(d[1] >> 5) // avx2
       // & (d[1] >> 31) // avx512vl
-      & (d[2] >> 9); // vaes // VEX-256/EVEX
+      & (BoolInt)(d[2] >> 9); // vaes // VEX-256/EVEX
   }
 }
 
@@ -688,7 +707,7 @@ BoolInt CPU_IsSupported_PageGB(void)
     if (d[0] < 0x80000001)
       return False;
     z7_x86_cpuid(d, 0x80000001);
-    return (d[3] >> 26) & 1;
+    return (BoolInt)(d[3] >> 26) & 1;
   }
 }
 
@@ -760,32 +779,65 @@ BoolInt CPU_IsSupported_AES (void) { return APPLE_CRYPTO_SUPPORT_VAL; }
 
 #else // __APPLE__
 
-#include <sys/auxv.h>
+#if defined(__GLIBC__) && (__GLIBC__ * 100 + __GLIBC_MINOR__ >= 216)
+  #define Z7_GETAUXV_AVAILABLE
+#else
+// #pragma message("=== is not NEW GLIBC === ")
+  #if defined __has_include
+  #if __has_include (<sys/auxv.h>)
+// #pragma message("=== sys/auxv.h is avail=== ")
+    #define Z7_GETAUXV_AVAILABLE
+  #endif
+  #endif
+#endif
 
+#ifdef Z7_GETAUXV_AVAILABLE
+// #pragma message("=== Z7_GETAUXV_AVAILABLE === ")
+#include <sys/auxv.h>
 #define USE_HWCAP
+#endif
 
 #ifdef USE_HWCAP
 
+#if defined(__FreeBSD__)
+static unsigned long MY_getauxval(int aux)
+{
+  unsigned long val;
+  if (elf_aux_info(aux, &val, sizeof(val)))
+    return 0;
+  return val;
+}
+#else
+#define MY_getauxval  getauxval
+  #if defined __has_include
+  #if __has_include (<asm/hwcap.h>)
 #include <asm/hwcap.h>
+  #endif
+  #endif
+#endif
 
   #define MY_HWCAP_CHECK_FUNC_2(name1, name2) \
-  BoolInt CPU_IsSupported_ ## name1() { return (getauxval(AT_HWCAP)  & (HWCAP_  ## name2)) ? 1 : 0; }
+  BoolInt CPU_IsSupported_ ## name1(void) { return (MY_getauxval(AT_HWCAP)  & (HWCAP_  ## name2)); }
 
 #ifdef MY_CPU_ARM64
   #define MY_HWCAP_CHECK_FUNC(name) \
   MY_HWCAP_CHECK_FUNC_2(name, name)
+#if 1 || defined(__ARM_NEON)
+  BoolInt CPU_IsSupported_NEON(void) { return True; }
+#else
   MY_HWCAP_CHECK_FUNC_2(NEON, ASIMD)
+#endif
 // MY_HWCAP_CHECK_FUNC (ASIMD)
 #elif defined(MY_CPU_ARM)
   #define MY_HWCAP_CHECK_FUNC(name) \
-  BoolInt CPU_IsSupported_ ## name() { return (getauxval(AT_HWCAP2) & (HWCAP2_ ## name)) ? 1 : 0; }
+  BoolInt CPU_IsSupported_ ## name(void) { return (MY_getauxval(AT_HWCAP2) & (HWCAP2_ ## name)); }
   MY_HWCAP_CHECK_FUNC_2(NEON, NEON)
 #endif
 
 #else // USE_HWCAP
 
   #define MY_HWCAP_CHECK_FUNC(name) \
-  BoolInt CPU_IsSupported_ ## name() { return 0; }
+  BoolInt CPU_IsSupported_ ## name(void) { return 0; }
   MY_HWCAP_CHECK_FUNC(NEON)
 
 #endif // USE_HWCAP