From 956be69e1dd0ab298f8f6f5d37119a266166349d Mon Sep 17 00:00:00 2001 From: Werner Saar Date: Tue, 17 May 2016 16:19:53 +0200 Subject: [PATCH] optimized getrf_single.c for POWER8 --- lapack/getrf/getrf_single.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/lapack/getrf/getrf_single.c b/lapack/getrf/getrf_single.c index e60a16c11..9f0f36b78 100644 --- a/lapack/getrf/getrf_single.c +++ b/lapack/getrf/getrf_single.c @@ -77,10 +77,17 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, blocking = (mn / 2 + GEMM_UNROLL_N - 1) & ~(GEMM_UNROLL_N - 1); if (blocking > GEMM_Q) blocking = GEMM_Q; +#ifdef POWER8 + if (blocking <= GEMM_UNROLL_N) { + info = GETF2(args, NULL, range_n, sa, sb, 0); + return info; + } +#else if (blocking <= GEMM_UNROLL_N * 2) { info = GETF2(args, NULL, range_n, sa, sb, 0); return info; } +#endif sbb = (FLOAT *)((((BLASULONG)(sb + blocking * blocking * COMPSIZE) + GEMM_ALIGN) & ~GEMM_ALIGN) + GEMM_OFFSET_B);