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

Use ALIGN_5 instead of .algin 32 in assembly kernel. Added ALIGN_5 for 32-bit OSX.

This commit is contained in:
Zhang Xianyi 2013-07-01 16:09:05 +08:00
parent 3ce2c62b0b
commit 57944538b6
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
common_x86.h
View File

@ -335,6 +335,7 @@ REALNAME:
#define ALIGN_2 .align 2 #define ALIGN_2 .align 2
#define ALIGN_3 .align 3 #define ALIGN_3 .align 3
#define ALIGN_4 .align 4 #define ALIGN_4 .align 4
#define ALIGN_5 .align 5
#define ffreep fstp #define ffreep fstp
#endif #endif
@ -356,11 +357,10 @@ REALNAME:
#ifndef ALIGN_6 #ifndef ALIGN_6
#define ALIGN_6 .align 64 #define ALIGN_6 .align 64
#endif
// ffreep %st(0). // ffreep %st(0).
// Because Clang didn't support ffreep, we directly use the opcode. // Because Clang didn't support ffreep, we directly use the opcode.
// Please check out http://www.sandpile.org/x86/opc_fpu.htm // Please check out http://www.sandpile.org/x86/opc_fpu.htm
#ifndef ffreep #ifndef ffreep
#define ffreep .byte 0xdf, 0xc0 # #define ffreep .byte 0xdf, 0xc0 #
#endif #endif
#endif

2
kernel/x86_64/dgemv_n_bulldozer.S
View File

@ -313,7 +313,7 @@
decq I decq I
jle .L14 jle .L14
.align 32 ALIGN_5
.L13: .L13: