arno/hashcat
1
0
Fork 0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-22 08:08:10 +00:00
Code Issues Releases Wiki Activity

ROCm JiT learned how to use V_ADD3_U32 efficiently

Browse Source
This commit is contained in:
Jens Steube 2019-06-18 12:41:59 +02:00
parent 62d5d2dfbb
commit 6ec52bd342
1 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
OpenCL/inc_common.cl
View File

@ -898,6 +898,7 @@ DECLSPEC u32 hc_byte_perm_S (const u32 a, const u32 b, const int c)
#if HAS_VADD3
DECLSPEC u32x hc_add3 (const u32x a, const u32x b, const u32x c)
{
/*
u32x r = 0;
#if VECT_SIZE == 1
@ -947,15 +948,22 @@ DECLSPEC u32x hc_add3 (const u32x a, const u32x b, const u32x c)
#endif
return r;
*/
return a + b + c;
}
DECLSPEC u32 hc_add3_S (const u32 a, const u32 b, const u32 c)
{
/*
u32 r = 0;
__asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c));
return r;
*/
return a + b + c;
}
#else
DECLSPEC u32x hc_add3 (const u32x a, const u32x b, const u32x c)