OSchip
/
OpenBLAS
Not watched
1
0
Fork 0
Code
Releases 66 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
Browse Source

Merge 80f7c5daf5 into 544b069e85

pull/1760/merge
Andrew GitHub 7 years ago
parent
544b069e85 80f7c5daf5
commit
bc024fe00a
No known key found for this signature in database GPG Key ID: 4AEE18F83AFDEB23
72 changed files with 72 additions and 72 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +1
    -1
      common_thread.h
  2. +1
    -1
      interface/axpy.c
  3. +1
    -1
      interface/gbmv.c
  4. +1
    -1
      interface/gemm.c
  5. +1
    -1
      interface/gemv.c
  6. +1
    -1
      interface/ger.c
  7. +1
    -1
      interface/lapack/gesv.c
  8. +1
    -1
      interface/lapack/getrf.c
  9. +1
    -1
      interface/lapack/getrs.c
  10. +1
    -1
      interface/lapack/laswp.c
  11. +1
    -1
      interface/lapack/lauum.c
  12. +1
    -1
      interface/lapack/potrf.c
  13. +1
    -1
      interface/lapack/potri.c
  14. +1
    -1
      interface/lapack/trtri.c
  15. +1
    -1
      interface/lapack/zgetrf.c
  16. +1
    -1
      interface/lapack/zgetrs.c
  17. +1
    -1
      interface/lapack/zlaswp.c
  18. +1
    -1
      interface/lapack/zlauum.c
  19. +1
    -1
      interface/lapack/zpotrf.c
  20. +1
    -1
      interface/lapack/zpotri.c
  21. +1
    -1
      interface/lapack/ztrtri.c
  22. +1
    -1
      interface/sbmv.c
  23. +1
    -1
      interface/scal.c
  24. +1
    -1
      interface/spmv.c
  25. +1
    -1
      interface/spr.c
  26. +1
    -1
      interface/spr2.c
  27. +1
    -1
      interface/symm.c
  28. +1
    -1
      interface/symv.c
  29. +1
    -1
      interface/syr.c
  30. +1
    -1
      interface/syr2.c
  31. +1
    -1
      interface/syr2k.c
  32. +1
    -1
      interface/syrk.c
  33. +1
    -1
      interface/tbmv.c
  34. +1
    -1
      interface/tpmv.c
  35. +1
    -1
      interface/trmv.c
  36. +1
    -1
      interface/trsm.c
  37. +1
    -1
      interface/zaxpy.c
  38. +1
    -1
      interface/zgbmv.c
  39. +1
    -1
      interface/zgemv.c
  40. +1
    -1
      interface/zger.c
  41. +1
    -1
      interface/zhbmv.c
  42. +1
    -1
      interface/zhemv.c
  43. +1
    -1
      interface/zher.c
  44. +1
    -1
      interface/zher2.c
  45. +1
    -1
      interface/zhpmv.c
  46. +1
    -1
      interface/zhpr.c
  47. +1
    -1
      interface/zhpr2.c
  48. +1
    -1
      interface/zsbmv.c
  49. +1
    -1
      interface/zscal.c
  50. +1
    -1
      interface/zspmv.c
  51. +1
    -1
      interface/zspr.c
  52. +1
    -1
      interface/zspr2.c
  53. +1
    -1
      interface/zswap.c
  54. +1
    -1
      interface/zsymv.c
  55. +1
    -1
      interface/zsyr.c
  56. +1
    -1
      interface/zsyr2.c
  57. +1
    -1
      interface/ztbmv.c
  58. +1
    -1
      interface/ztpmv.c
  59. +1
    -1
      interface/ztrmv.c
  60. +1
    -1
      kernel/arm64/casum_thunderx2t99.c
  61. +1
    -1
      kernel/arm64/copy_thunderx2t99.c
  62. +1
    -1
      kernel/arm64/dasum_thunderx2t99.c
  63. +1
    -1
      kernel/arm64/dot_thunderx2t99.c
  64. +1
    -1
      kernel/arm64/dznrm2_thunderx2t99.c
  65. +1
    -1
      kernel/arm64/dznrm2_thunderx2t99_fast.c
  66. +1
    -1
      kernel/arm64/iamax_thunderx2t99.c
  67. +1
    -1
      kernel/arm64/izamax_thunderx2t99.c
  68. +1
    -1
      kernel/arm64/sasum_thunderx2t99.c
  69. +1
    -1
      kernel/arm64/scnrm2_thunderx2t99.c
  70. +1
    -1
      kernel/arm64/zasum_thunderx2t99.c
  71. +1
    -1
      kernel/arm64/zdot_thunderx2t99.c
  72. +1
    -1
      kernel/x86_64/ddot.c

+ 1
- 1
common_thread.h View File

@@ -129,7 +129,7 @@ typedef struct blas_queue {

extern int blas_server_avail;

static __inline int num_cpu_avail(int level) {
static __inline int num_cpu_avail(void) {

#ifdef USE_OPENMP
int openmp_nthreads=0;


+ 1
- 1
interface/axpy.c View File

@@ -91,7 +91,7 @@ void CNAME(blasint n, FLOAT alpha, FLOAT *x, blasint incx, FLOAT *y, blasint inc
if (incx == 0 || incy == 0 || n <= MULTI_THREAD_MINIMAL)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/gbmv.c View File

@@ -227,7 +227,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/gemm.c View File

@@ -413,7 +413,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANS
if ( MNK <= (SMP_THRESHOLD_MIN * (double) GEMM_MULTITHREAD_THRESHOLD) )
args.nthreads = 1;
else
args.nthreads = num_cpu_avail(3);
args.nthreads = num_cpu_avail();
args.common = NULL;

if (args.nthreads == 1) {


+ 1
- 1
interface/gemv.c View File

@@ -223,7 +223,7 @@ void CNAME(enum CBLAS_ORDER order,
if ( 1L * m * n < 2304L * GEMM_MULTITHREAD_THRESHOLD )
nthreads = 1;
else
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/ger.c View File

@@ -176,7 +176,7 @@ void CNAME(enum CBLAS_ORDER order,
#ifdef SMPTEST
// Threshold chosen so that speed-up is > 1 on a Xeon E5-2630
if(1L * m * n > 2048L * GEMM_MULTITHREAD_THRESHOLD)
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();
else
nthreads = 1;



+ 1
- 1
interface/lapack/gesv.c View File

@@ -114,7 +114,7 @@ int NAME(blasint *N, blasint *NRHS, FLOAT *a, blasint *ldA, blasint *ipiv,

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/getrf.c View File

@@ -95,7 +95,7 @@ int NAME(blasint *M, blasint *N, FLOAT *a, blasint *ldA, blasint *ipiv, blasint

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/getrs.c View File

@@ -126,7 +126,7 @@ int NAME(char *TRANS, blasint *N, blasint *NRHS, FLOAT *a, blasint *ldA,

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/laswp.c View File

@@ -77,7 +77,7 @@ int NAME(blasint *N, FLOAT *a, blasint *LDA, blasint *K1, blasint *K2, blasint *
flag = (incx < 0);

#ifdef SMP
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/lauum.c View File

@@ -112,7 +112,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/potrf.c View File

@@ -112,7 +112,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/potri.c View File

@@ -121,7 +121,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/trtri.c View File

@@ -127,7 +127,7 @@ int NAME(char *UPLO, char *DIAG, blasint *N, FLOAT *a, blasint *ldA, blasint *In
#endif

#ifdef SMP
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zgetrf.c View File

@@ -95,7 +95,7 @@ int NAME(blasint *M, blasint *N, FLOAT *a, blasint *ldA, blasint *ipiv, blasint

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zgetrs.c View File

@@ -125,7 +125,7 @@ int NAME(char *TRANS, blasint *N, blasint *NRHS, FLOAT *a, blasint *ldA,
#endif

#ifdef SMP
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zlaswp.c View File

@@ -78,7 +78,7 @@ int NAME(blasint *N, FLOAT *a, blasint *LDA, blasint *K1, blasint *K2, blasint *
flag = (incx < 0);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zlauum.c View File

@@ -112,7 +112,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zpotrf.c View File

@@ -112,7 +112,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/zpotri.c View File

@@ -120,7 +120,7 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){
#endif

#ifdef SMP
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/lapack/ztrtri.c View File

@@ -125,7 +125,7 @@ int NAME(char *UPLO, char *DIAG, blasint *N, FLOAT *a, blasint *ldA, blasint *In

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(4);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/sbmv.c View File

@@ -198,7 +198,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMPTEST
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/scal.c View File

@@ -79,7 +79,7 @@ void CNAME(blasint n, FLOAT alpha, FLOAT *x, blasint incx){
if (n <= 1048576 )
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();


if (nthreads == 1) {


+ 1
- 1
interface/spmv.c View File

@@ -182,7 +182,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMPTEST
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/spr.c View File

@@ -172,7 +172,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/spr2.c View File

@@ -178,7 +178,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/symm.c View File

@@ -375,7 +375,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo,

#ifdef SMP
args.common = NULL;
args.nthreads = num_cpu_avail(3);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/symv.c View File

@@ -180,7 +180,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/syr.c View File

@@ -174,7 +174,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/syr2.c View File

@@ -178,7 +178,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/syr2k.c View File

@@ -368,7 +368,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Tr
mode |= (uplo << BLAS_UPLO_SHIFT);

args.common = NULL;
args.nthreads = num_cpu_avail(3);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/syrk.c View File

@@ -354,7 +354,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Tr
#endif

args.common = NULL;
args.nthreads = num_cpu_avail(3);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {
#endif


+ 1
- 1
interface/tbmv.c View File

@@ -223,7 +223,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/tpmv.c View File

@@ -221,7 +221,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/trmv.c View File

@@ -218,7 +218,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
/* nthreads = num_cpu_avail(2);
/* nthreads = num_cpu_avail();

FIXME trmv_thread was found to be broken, see issue 1332 */
nthreads = 1;


+ 1
- 1
interface/trsm.c View File

@@ -372,7 +372,7 @@ void CNAME(enum CBLAS_ORDER order,
if ( args.n < 2*GEMM_MULTITHREAD_THRESHOLD )
args.nthreads = 1;
else
args.nthreads = num_cpu_avail(3);
args.nthreads = num_cpu_avail();

if (args.nthreads == 1) {


+ 1
- 1
interface/zaxpy.c View File

@@ -98,7 +98,7 @@ void CNAME(blasint n, FLOAT *ALPHA, FLOAT *x, blasint incx, FLOAT *y, blasint in
if (incx == 0 || incy == 0 || n <= MULTI_THREAD_MINIMAL)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zgbmv.c View File

@@ -251,7 +251,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zgemv.c View File

@@ -255,7 +255,7 @@ void CNAME(enum CBLAS_ORDER order,
if ( 1L * m * n < 1024L * GEMM_MULTITHREAD_THRESHOLD )
nthreads = 1;
else
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zger.c View File

@@ -220,7 +220,7 @@ void CNAME(enum CBLAS_ORDER order,
#ifdef SMPTEST
// Threshold chosen so that speed-up is > 1 on a Xeon E5-2630
if(1L * m * n > 36L * sizeof(FLOAT) * sizeof(FLOAT) * GEMM_MULTITHREAD_THRESHOLD)
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();
else
nthreads = 1;



+ 1
- 1
interface/zhbmv.c View File

@@ -204,7 +204,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMPBUG
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zhemv.c View File

@@ -195,7 +195,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, void *VALPHA
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zher.c View File

@@ -177,7 +177,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zher2.c View File

@@ -186,7 +186,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, void *VALPHA
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zhpmv.c View File

@@ -194,7 +194,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zhpr.c View File

@@ -175,7 +175,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zhpr2.c View File

@@ -187,7 +187,7 @@ void CNAME(enum CBLAS_ORDER order,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zsbmv.c View File

@@ -140,7 +140,7 @@ void NAME(char *UPLO, blasint *N, blasint *K, FLOAT *ALPHA, FLOAT *a, blasint *
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMPTEST
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zscal.c View File

@@ -93,7 +93,7 @@ void CNAME(blasint n, FLOAT alpha_r, void *vx, blasint incx){
if ( n <= 1048576 )
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zspmv.c View File

@@ -128,7 +128,7 @@ void NAME(char *UPLO, blasint *N, FLOAT *ALPHA, FLOAT *a,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMPTEST
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zspr.c View File

@@ -121,7 +121,7 @@ void NAME(char *UPLO, blasint *N, FLOAT *ALPHA,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zspr2.c View File

@@ -124,7 +124,7 @@ void NAME(char *UPLO, blasint *N, FLOAT *ALPHA,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zswap.c View File

@@ -84,7 +84,7 @@ FLOAT *y = (FLOAT*)vy;
if (incx == 0 || incy == 0)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zsymv.c View File

@@ -118,7 +118,7 @@ void NAME(char *UPLO, blasint *N, FLOAT *ALPHA, FLOAT *a, blasint *LDA,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zsyr.c View File

@@ -181,7 +181,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, int n, FLOAT alpha, FLO
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/zsyr2.c View File

@@ -126,7 +126,7 @@ void NAME(char *UPLO, blasint *N, FLOAT *ALPHA,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/ztbmv.c View File

@@ -238,7 +238,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/ztpmv.c View File

@@ -229,7 +229,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
buffer = (FLOAT *)blas_memory_alloc(1);

#ifdef SMP
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();

if (nthreads == 1) {
#endif


+ 1
- 1
interface/ztrmv.c View File

@@ -233,7 +233,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
#ifdef SMP
// Calibrated on a Xeon E5-2630
if(1L * n * n > 36L * sizeof(FLOAT) * sizeof(FLOAT) * GEMM_MULTITHREAD_THRESHOLD) {
nthreads = num_cpu_avail(2);
nthreads = num_cpu_avail();
if(nthreads > 2 && 1L * n * n < 64L * sizeof(FLOAT) * sizeof(FLOAT) * GEMM_MULTITHREAD_THRESHOLD)
nthreads = 2;
} else


+ 1
- 1
kernel/arm64/casum_thunderx2t99.c View File

@@ -236,7 +236,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
asum = casum_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/copy_thunderx2t99.c View File

@@ -186,7 +186,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
do_copy(n, x, inc_x, y, inc_y);


+ 1
- 1
kernel/arm64/dasum_thunderx2t99.c View File

@@ -231,7 +231,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
asum = dasum_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/dot_thunderx2t99.c View File

@@ -387,7 +387,7 @@ RETURN_TYPE CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y
if (inc_x == 0 || inc_y == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
dot = dot_compute(n, x, inc_x, y, inc_y);


+ 1
- 1
kernel/arm64/dznrm2_thunderx2t99.c View File

@@ -331,7 +331,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
nrm2_compute(n, x, inc_x, &ssq, &scale);


+ 1
- 1
kernel/arm64/dznrm2_thunderx2t99_fast.c View File

@@ -238,7 +238,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
nrm2 = nrm2_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/iamax_thunderx2t99.c View File

@@ -324,7 +324,7 @@ BLASLONG CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
max_index = iamax_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/izamax_thunderx2t99.c View File

@@ -333,7 +333,7 @@ BLASLONG CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
max_index = izamax_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/sasum_thunderx2t99.c View File

@@ -233,7 +233,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
asum = sasum_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/scnrm2_thunderx2t99.c View File

@@ -321,7 +321,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
nrm2_double = nrm2_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/zasum_thunderx2t99.c View File

@@ -233,7 +233,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
if (inc_x == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
asum = zasum_compute(n, x, inc_x);


+ 1
- 1
kernel/arm64/zdot_thunderx2t99.c View File

@@ -320,7 +320,7 @@ OPENBLAS_COMPLEX_FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLA
if (inc_x == 0 || inc_y == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
zdot_compute(n, x, inc_x, y, inc_y, &zdot);


+ 1
- 1
kernel/x86_64/ddot.c View File

@@ -174,7 +174,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
if (inc_x == 0 || inc_y == 0 || n <= 10000)
nthreads = 1;
else
nthreads = num_cpu_avail(1);
nthreads = num_cpu_avail();

if (nthreads == 1) {
dot = dot_compute(n, x, inc_x, y, inc_y);


Write Preview
Loading…
Cancel
Save