From e90a147493f95d3d03acdf962ae335c128667ba6 Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Sat, 9 Dec 2017 18:44:34 +0100 Subject: [PATCH] Calculation of chunk range limits was ignoring num_cpu bug introduced by me in #1262 - should fix #1388 --- driver/level2/gbmv_thread.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/driver/level2/gbmv_thread.c b/driver/level2/gbmv_thread.c index c0cb947dc..4fce9744f 100644 --- a/driver/level2/gbmv_thread.c +++ b/driver/level2/gbmv_thread.c @@ -230,10 +230,10 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG ku, BLASLONG kl, FLOAT *alpha, FLOAT #ifndef TRANSA range_m[num_cpu] = num_cpu * ((m + 15) & ~15); - if (range_m[num_cpu] > m) range_m[num_cpu] = m; + if (range_m[num_cpu] > m * num_cpu) range_m[num_cpu] = m * num_cpu; #else range_m[num_cpu] = num_cpu * ((n + 15) & ~15); - if (range_m[num_cpu] > n) range_m[num_cpu] = n; + if (range_m[num_cpu] > n * num_cpu) range_m[num_cpu] = n * num_cpu; #endif queue[num_cpu].mode = mode;