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

Adding performance patch for trmm, just like #2836

Browse Source
This commit is contained in:
fossum
2020-09-15 08:59:50 -05:00
parent 91c84e1c01
commit dfeca46098
2 changed files with 10 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
driver/level3/trmm_L.c
View File

@@ -139,7 +139,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -209,7 +209,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -304,7 +304,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -374,7 +374,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif

12
driver/level3/trmm_R.c
View File

@@ -126,7 +126,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -150,7 +150,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -207,7 +207,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -262,7 +262,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -287,7 +287,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif
@@ -348,7 +348,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
/* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */ /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N; if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
#else #else
if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3; if (min_jj >= GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
else else
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#endif #endif