Browse Source

use emms instead, add WIN guards

tags/v0.3.11^2
Matti Picus 5 years ago
parent
commit
403eb513a0
27 changed files with 87 additions and 29 deletions
  1. +3
    -1
      kernel/x86_64/amax.S
  2. +4
    -1
      kernel/x86_64/asum.S
  3. +4
    -1
      kernel/x86_64/dot.S
  4. +4
    -1
      kernel/x86_64/iamax.S
  5. +4
    -1
      kernel/x86_64/izamax.S
  6. +4
    -1
      kernel/x86_64/nrm2.S
  7. +4
    -1
      kernel/x86_64/qconjg.S
  8. +3
    -1
      kernel/x86_64/qdot.S
  9. +3
    -1
      kernel/x86_64/qgemm_kernel_2x2.S
  10. +3
    -1
      kernel/x86_64/qgemv_n.S
  11. +4
    -1
      kernel/x86_64/qgemv_t.S
  12. +3
    -1
      kernel/x86_64/qtrsm_kernel_LN_2x2.S
  13. +3
    -1
      kernel/x86_64/qtrsm_kernel_LT_2x2.S
  14. +3
    -2
      kernel/x86_64/qtrsm_kernel_RT_2x2.S
  15. +3
    -1
      kernel/x86_64/sum.S
  16. +3
    -1
      kernel/x86_64/xdot.S
  17. +3
    -1
      kernel/x86_64/xgemm3m_kernel_2x2.S
  18. +3
    -1
      kernel/x86_64/xgemm_kernel_1x1.S
  19. +3
    -1
      kernel/x86_64/xgemv_n.S
  20. +3
    -1
      kernel/x86_64/xgemv_t.S
  21. +3
    -1
      kernel/x86_64/xtrsm_kernel_LT_1x1.S
  22. +3
    -1
      kernel/x86_64/zamax.S
  23. +3
    -1
      kernel/x86_64/zasum.S
  24. +2
    -2
      kernel/x86_64/zdot.S
  25. +3
    -1
      kernel/x86_64/znrm2.S
  26. +3
    -1
      kernel/x86_64/zscal.S
  27. +3
    -1
      kernel/x86_64/zsum.S

+ 3
- 1
kernel/x86_64/amax.S View File

@@ -55,7 +55,9 @@
PROLOGUE
PROFCODE
fninit
#ifdef WINDOWS_ABI
emms
#endif

salq $BASE_SHIFT, INCX



+ 4
- 1
kernel/x86_64/asum.S View File

@@ -50,7 +50,10 @@
PROLOGUE
PROFCODE
fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M
jle .L999


+ 4
- 1
kernel/x86_64/dot.S View File

@@ -49,7 +49,10 @@

PROLOGUE
PROFCODE
fninit

#ifdef WINDOWS_ABI
emms
#endif

salq $BASE_SHIFT, INCX
salq $BASE_SHIFT, INCY


+ 4
- 1
kernel/x86_64/iamax.S View File

@@ -59,7 +59,10 @@

PROLOGUE
PROFCODE
fninit

#ifdef WINDOWS_ABI
emms
#endif

salq $BASE_SHIFT, INCX



+ 4
- 1
kernel/x86_64/izamax.S View File

@@ -59,7 +59,10 @@

PROLOGUE
PROFCODE
fninit

#ifdef WINDOWS_ABI
emms
#endif

salq $ZBASE_SHIFT, INCX



+ 4
- 1
kernel/x86_64/nrm2.S View File

@@ -50,7 +50,10 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M
jle .L999


+ 4
- 1
kernel/x86_64/qconjg.S View File

@@ -41,7 +41,10 @@

PROLOGUE
PROFCODE
fninit

#ifdef WINDOWS_ABI
emms
#endif

fldz
FLD 1 * SIZE(ARG1)


+ 3
- 1
kernel/x86_64/qdot.S View File

@@ -58,7 +58,9 @@

PROLOGUE

fninit
#ifdef WINDOWS_ABI
emms
#endif

pushl %edi
pushl %esi


+ 3
- 1
kernel/x86_64/qgemm_kernel_2x2.S View File

@@ -74,7 +74,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/qgemv_n.S View File

@@ -76,7 +76,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 4
- 1
kernel/x86_64/qgemv_t.S View File

@@ -75,7 +75,10 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)
movq %rbp, 8(%rsp)


+ 3
- 1
kernel/x86_64/qtrsm_kernel_LN_2x2.S View File

@@ -74,7 +74,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/qtrsm_kernel_LT_2x2.S View File

@@ -74,7 +74,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 2
kernel/x86_64/qtrsm_kernel_RT_2x2.S View File

@@ -74,8 +74,9 @@
PROLOGUE
PROFCODE

fninit

#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/sum.S View File

@@ -50,7 +50,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M


+ 3
- 1
kernel/x86_64/xdot.S View File

@@ -59,7 +59,9 @@

PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif


#define N %ebx


+ 3
- 1
kernel/x86_64/xgemm3m_kernel_2x2.S View File

@@ -78,7 +78,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/xgemm_kernel_1x1.S View File

@@ -97,7 +97,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/xgemv_n.S View File

@@ -76,7 +76,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/xgemv_t.S View File

@@ -75,7 +75,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/xtrsm_kernel_LT_1x1.S View File

@@ -90,7 +90,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

subq $STACKSIZE, %rsp
movq %rbx, 0(%rsp)


+ 3
- 1
kernel/x86_64/zamax.S View File

@@ -55,7 +55,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

salq $ZBASE_SHIFT, INCX



+ 3
- 1
kernel/x86_64/zasum.S View File

@@ -50,7 +50,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M


+ 2
- 2
kernel/x86_64/zdot.S View File

@@ -54,9 +54,9 @@
PROLOGUE
PROFCODE

fninit

#ifdef WINDOWS_ABI
emms

movq 40(%rsp), INCY
#endif



+ 3
- 1
kernel/x86_64/znrm2.S View File

@@ -50,7 +50,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M


+ 3
- 1
kernel/x86_64/zscal.S View File

@@ -50,7 +50,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

salq $ZBASE_SHIFT, INCX



+ 3
- 1
kernel/x86_64/zsum.S View File

@@ -50,7 +50,9 @@
PROLOGUE
PROFCODE

fninit
#ifdef WINDOWS_ABI
emms
#endif

fldz
testq M, M


Loading…
Cancel
Save