From 5d26223f4a91e14ec711168f6e4a40f21729be38 Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Tue, 8 Dec 2020 20:59:56 +0100 Subject: [PATCH 1/3] remove extra/intermediate size step of min_jj from PR747 --- driver/level3/level3.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/driver/level3/level3.c b/driver/level3/level3.c index a38506585..9b44deb85 100644 --- a/driver/level3/level3.c +++ b/driver/level3/level3.c @@ -339,8 +339,10 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, #else if (min_jj >= 3*GEMM_UNROLL_N) min_jj = 3*GEMM_UNROLL_N; else - if (min_jj >= 2*GEMM_UNROLL_N) min_jj = 2*GEMM_UNROLL_N; +/* + if (min_jj >= 2*GEMM_UNROLL_N) min_jj = 2*GEMM_UNROLL_N; else +*/ if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; #endif From a5547124393a3ea7538998e98356cb052dc652d0 Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Tue, 8 Dec 2020 21:01:36 +0100 Subject: [PATCH 2/3] remove extra/intermediate size step for min_jj introduced in PR747 --- driver/level3/level3_thread.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/driver/level3/level3_thread.c b/driver/level3/level3_thread.c index 6e1fd9e99..2b33c9589 100644 --- a/driver/level3/level3_thread.c +++ b/driver/level3/level3_thread.c @@ -373,8 +373,10 @@ static int inner_thread(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, #else if (min_jj >= 3*GEMM_UNROLL_N) min_jj = 3*GEMM_UNROLL_N; else +/* if (min_jj >= 2*GEMM_UNROLL_N) min_jj = 2*GEMM_UNROLL_N; else +*/ if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N; #endif /* Copy part of local region of B into workspace */ From d71fe4ed4eff491a9e6aae87fbd46cf9d2914d9e Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Tue, 8 Dec 2020 21:07:57 +0100 Subject: [PATCH 3/3] Remove GEMM_DEFAULT_UNROLL_MN parameters for Haswell and ZEN (introduced in PR747) --- param.h | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/param.h b/param.h index a0d45c573..42f63b4b5 100644 --- a/param.h +++ b/param.h @@ -644,9 +644,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. #define CGEMM_DEFAULT_UNROLL_N 2 #define ZGEMM_DEFAULT_UNROLL_N 2 #define XGEMM_DEFAULT_UNROLL_N 1 - +/* #define SGEMM_DEFAULT_UNROLL_MN 32 #define DGEMM_DEFAULT_UNROLL_MN 32 +*/ #endif #ifdef ARCH_X86 @@ -1552,9 +1553,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. #define CGEMM_DEFAULT_UNROLL_N 2 #define ZGEMM_DEFAULT_UNROLL_N 2 #define XGEMM_DEFAULT_UNROLL_N 1 - +/* #define SGEMM_DEFAULT_UNROLL_MN 32 #define DGEMM_DEFAULT_UNROLL_MN 32 +*/ #endif #ifdef ARCH_X86