added check for x86intrin.h (fixed #137), changed to builtin function…

…s for cpu detection (fixes #138 #140 and #141)

src/Makefile.am

@@ -37,8 +37,7 @@ fast_parsimony.c \
 stepwise.c \
 random.c \
 phylip.c \
-hardware.c \
-init.c
+hardware.c
 
 libpll_la_CFLAGS = $(AM_CFLAGS)
 

src/hardware.c

@@ -21,101 +21,83 @@
 
 #include "pll.h"
 
-#ifndef __PPC__
-#define cpuid(f1, f2, a, b, c, d)                                \
-  __asm__ __volatile__ ("cpuid"                                  \
-                        : "=a" (a), "=b" (b), "=c" (c), "=d" (d) \
-                        : "a" (f1), "c" (f2));
-#endif
-
 static void cpu_features_detect()
 {
-  unsigned int a,b,c,d;
-
-  memset(pll_hardware,0,sizeof(pll_hardware_t));
-
-#ifdef __PPC__
-  pll_hardware->altivec_present = 1;
-#else
-
-  cpuid(0,0,a,b,c,d);
-  unsigned int maxlevel = a & 0xff;
-
-  if (maxlevel >= 1)
-  {
-    cpuid(1,0,a,b,c,d);
-    pll_hardware->mmx_present    = (d >> 23) & 1;
-    pll_hardware->sse_present    = (d >> 25) & 1;
-    pll_hardware->sse2_present   = (d >> 26) & 1;
-    pll_hardware->sse3_present   = (c >>  0) & 1;
-    pll_hardware->ssse3_present  = (c >>  9) & 1;
-    pll_hardware->sse41_present  = (c >> 19) & 1;
-    pll_hardware->sse42_present  = (c >> 20) & 1;
-    pll_hardware->popcnt_present = (c >> 23) & 1;
-    pll_hardware->avx_present    = (c >> 28) & 1;
-
-    if (maxlevel >= 7)
-    {
-      cpuid(7,0,a,b,c,d);
-      pll_hardware->avx2_present = (b >> 5) & 1;
-    }
-  }
+  memset(&pll_hardware,0,sizeof(pll_hardware_t));
+
+  pll_hardware.init            = 1;
+#if defined(__PPC__)
+  pll_hardware.altivec_present = __builtin_cpu_supports("altivec");
+#elif defined(__x86_64__) || defined(__i386__)
+  pll_hardware.mmx_present     = __builtin_cpu_supports("mmx");
+  pll_hardware.sse_present     = __builtin_cpu_supports("sse");
+  pll_hardware.sse2_present    = __builtin_cpu_supports("sse2");
+  pll_hardware.sse3_present    = __builtin_cpu_supports("sse3");
+  pll_hardware.ssse3_present   = __builtin_cpu_supports("ssse3");
+  pll_hardware.sse41_present   = __builtin_cpu_supports("sse4.1");
+  pll_hardware.sse42_present   = __builtin_cpu_supports("sse4.2");
+  pll_hardware.popcnt_present  = __builtin_cpu_supports("popcnt");
+  pll_hardware.avx_present     = __builtin_cpu_supports("avx");
+  pll_hardware.avx2_present    = __builtin_cpu_supports("avx2");
 #endif
 }
 
 static void cpu_features_show()
 {
-  if (!pll_hardware)
-  {
-    /* TODO: Add proper error control after we figure out
-       cross-platform compatibility */
-    return;
-  }
-
   fprintf(stderr, "CPU features:");
-  if (pll_hardware->altivec_present)
+  if (pll_hardware.altivec_present)
     fprintf(stderr, " altivec");
-  if (pll_hardware->mmx_present)
+  if (pll_hardware.mmx_present)
     fprintf(stderr, " mmx");
-  if (pll_hardware->sse_present)
+  if (pll_hardware.sse_present)
     fprintf(stderr, " sse");
-  if (pll_hardware->sse2_present)
+  if (pll_hardware.sse2_present)
     fprintf(stderr, " sse2");
-  if (pll_hardware->sse3_present)
+  if (pll_hardware.sse3_present)
     fprintf(stderr, " sse3");
-  if (pll_hardware->ssse3_present)
+  if (pll_hardware.ssse3_present)
     fprintf(stderr, " ssse3");
-  if (pll_hardware->sse41_present)
+  if (pll_hardware.sse41_present)
     fprintf(stderr, " sse4.1");
-  if (pll_hardware->sse42_present)
+  if (pll_hardware.sse42_present)
     fprintf(stderr, " sse4.2");
-  if (pll_hardware->popcnt_present)
+  if (pll_hardware.popcnt_present)
     fprintf(stderr, " popcnt");
-  if (pll_hardware->avx_present)
+  if (pll_hardware.avx_present)
     fprintf(stderr, " avx");
-  if (pll_hardware->avx2_present)
+  if (pll_hardware.avx2_present)
     fprintf(stderr, " avx2");
   fprintf(stderr, "\n");
 }
 
 PLL_EXPORT int pll_hardware_probe()
 {
   /* probe cpu features */
-  if (!pll_hardware)
-  {
-    if (!(pll_hardware = (pll_hardware_t *)calloc(1,sizeof(pll_hardware_t))))
-    {
-      pll_errno = PLL_ERROR_MEM_ALLOC;
-      snprintf(pll_errmsg, 200, "Unable to allocate enough memory.");
-      return PLL_FAILURE;
-    }
-  }
   cpu_features_detect();
 
   return PLL_SUCCESS;
 }
 
 PLL_EXPORT void pll_hardware_dump()
 {
+  if (!pll_hardware.init)
+    pll_hardware_probe();
+
   cpu_features_show(); 
 }
+
+PLL_EXPORT void pll_hardware_ignore()
+{
+  pll_hardware.init            = 1;
+  pll_hardware.altivec_present = 1;
+  pll_hardware.mmx_present     = 1;
+  pll_hardware.sse_present     = 1;
+  pll_hardware.sse2_present    = 1;
+  pll_hardware.sse3_present    = 1;
+  pll_hardware.ssse3_present   = 1;
+  pll_hardware.sse41_present   = 1;
+  pll_hardware.sse42_present   = 1;
+  pll_hardware.popcnt_present  = 1;
+  pll_hardware.avx_present     = 1;
+  pll_hardware.avx2_present    = 1;
+}

src/pll.c

@@ -24,7 +24,7 @@
 __thread int pll_errno;
 __thread char pll_errmsg[200] = {0};
 
-pll_hardware_t * pll_hardware = NULL;
+pll_hardware_t pll_hardware = {0,0,0,0,0,0,0,0,0,0,0,0};
 
 static void dealloc_partition_data(pll_partition_t * partition);
 

src/pll.h

@@ -26,12 +26,15 @@
 #include <stdint.h>
 #include <string.h>
 #include <ctype.h>
-#include <x86intrin.h>
 
 #ifdef HAVE_CONFIG_H
 #include "config.h"
 #endif
 
+#ifdef HAVE_X86INTRIN_H
+#include <x86intrin.h>
+#endif
+
 /* platform specific */
 
 #if (!defined(__APPLE__) && !defined(__WIN32__) && !defined(__WIN64__))
@@ -49,7 +52,8 @@
 #define PLL_MIN(a,b) ((a) < (b) ? (a) : (b))
 #define PLL_MAX(a,b) ((a) > (b) ? (a) : (b))
 #define PLL_SWAP(x,y) do { __typeof__ (x) _t = x; x = y; y = _t; } while(0)
-#define PLL_STAT(x) (pll_hardware && pll_hardware->x)
+#define PLL_STAT(x) ((pll_hardware.init || pll_hardware_probe()) \
+                     && pll_hardware.x)
 
 /* constants */
 
@@ -163,6 +167,7 @@
 
 typedef struct pll_hardware_s
 {
+  int init;
   /* cpu features */
   int altivec_present;
   int mmx_present;
@@ -451,7 +456,7 @@ struct pll_random_data
 
 PLL_EXPORT extern __thread int pll_errno;
 PLL_EXPORT extern __thread char pll_errmsg[200];
-PLL_EXPORT extern pll_hardware_t * pll_hardware;
+PLL_EXPORT extern pll_hardware_t pll_hardware;
 
 PLL_EXPORT extern const unsigned int pll_map_bin[256];
 PLL_EXPORT extern const unsigned int pll_map_nt[256];
@@ -1875,12 +1880,7 @@ PLL_EXPORT int pll_hardware_probe(void);
 
 PLL_EXPORT void pll_hardware_dump();
 
-/* functions in init.c */
-
-PLL_EXPORT void pll_init(void) __attribute__((constructor));
-
-PLL_EXPORT void pll_fini(void) __attribute__((destructor));
-
+PLL_EXPORT void pll_hardware_ignore();
 
 #ifdef __cplusplus
 } /* extern "C" */