floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Replace with ARM64 intrinsics

This commit is contained in:
Nursultan Zarlyk 2022-06-09 18:49:49 +02:00
parent 1bb7993a97
commit 1dfc4e6150
1 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
common_arm64.h
View File

@ -35,9 +35,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifdef C_MSVC #ifdef C_MSVC
#include <intrin.h> #include <intrin.h>
#define MB do {} while (0) #define MB __dmb(_ARM64_BARRIER_ISH)
#define WMB do {} while (0) #define WMB __dmb(_ARM64_BARRIER_ISHST)
#define RMB #define RMB __dmb(_ARM64_BARRIER_ISHLD)
#else #else
#define MB __asm__ __volatile__ ("dmb ish" : : : "memory") #define MB __asm__ __volatile__ ("dmb ish" : : : "memory")
#define WMB __asm__ __volatile__ ("dmb ishst" : : : "memory") #define WMB __asm__ __volatile__ ("dmb ishst" : : : "memory")
@ -92,7 +92,12 @@ static void __inline blas_lock(volatile BLASULONG *address){
#if !defined(OS_DARWIN) && !defined (OS_ANDROID) #if !defined(OS_DARWIN) && !defined (OS_ANDROID)
static __inline BLASULONG rpcc(void){ static __inline BLASULONG rpcc(void){
#ifdef C_MSVC #ifdef C_MSVC
return __rdtsc(); const int64_t pmccntr_el0 = (((3 & 1) << 14) | // op0
((3 & 7) << 11) | // op1
((9 & 15) << 7) | // crn
((13 & 15) << 3) | // crm
((0 & 7) << 0)); // op2
return _ReadStatusReg(pmccntr_el0);
#else #else
BLASULONG ret = 0; BLASULONG ret = 0;
blasint shift; blasint shift;