aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fix handling of __clock() builtin for "generic" targets.

Browse Source
This commit is contained in:
Matt Pharr
2013-08-20 09:04:52 -07:00
parent d976da7559
commit e7f067d70c
7 changed files with 118 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
examples/intrinsics/generic-16.h
View File

@@ -1759,3 +1759,23 @@ static FORCEINLINE uint64_t __atomic_cmpxchg(uint64_t *p, uint64_t cmpval,
return __sync_val_compare_and_swap(p, cmpval, newval);
#endif
}
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif // !WIN32

20
examples/intrinsics/generic-32.h
View File

@@ -1827,3 +1827,23 @@ static FORCEINLINE uint64_t __atomic_cmpxchg(uint64_t *p, uint64_t cmpval,
return __sync_val_compare_and_swap(p, cmpval, newval);
#endif
}
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif // WIN32
#undef FORCEINLINE

20
examples/intrinsics/generic-64.h
View File

@@ -1960,3 +1960,23 @@ static FORCEINLINE uint64_t __atomic_cmpxchg(uint64_t *p, uint64_t cmpval,
return __sync_val_compare_and_swap(p, cmpval, newval);
#endif
}
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif
#undef FORCEINLINE

21
examples/intrinsics/knc.h
View File

@@ -2121,9 +2121,24 @@ static FORCEINLINE uint64_t __atomic_cmpxchg(uint64_t *p, uint64_t cmpval,
#endif
}
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif // !WIN32
#undef FORCEINLINE
#undef PRE_ALIGN
#undef POST_ALIGN

19
examples/intrinsics/knc2x.h
View File

@@ -2055,7 +2055,24 @@ static FORCEINLINE void __aos_to_soa4_float(float *ptr, __vec32_f *out0, __vec32
}
*/
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif // !WIN32
#undef FORCEINLINE
#undef PRE_ALIGN
#undef POST_ALIGN

20
examples/intrinsics/sse4.h
View File

@@ -4000,6 +4000,22 @@ static FORCEINLINE uint64_t __atomic_cmpxchg(uint64_t *p, uint64_t cmpval,
#endif
}
#ifdef WIN32
#include <windows.h>
#define __clock __rdtsc
#else // WIN32
static FORCEINLINE uint64_t __clock() {
uint32_t low, high;
#ifdef __x86_64
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%rax", "%rbx", "%rcx", "%rdx" );
#else
__asm__ __volatile__ ("xorl %%eax,%%eax \n cpuid"
::: "%eax", "%ebx", "%ecx", "%edx" );
#endif
__asm__ __volatile__ ("rdtsc" : "=a" (low), "=d" (high));
return (uint64_t)high << 32 | low;
}
#endif // !WIN32
#undef FORCEINLINE