From a445c355569d19123397bc684e5f8994bb883e5e Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Sat, 9 Dec 2017 18:55:23 +0100 Subject: [PATCH] Calculation of chunk range limits was ignoring num_cpu bug introduced by me in #1262 --- driver/level2/tpmv_thread.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/driver/level2/tpmv_thread.c b/driver/level2/tpmv_thread.c index b7eb10571..a077591a5 100644 --- a/driver/level2/tpmv_thread.c +++ b/driver/level2/tpmv_thread.c @@ -303,7 +303,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *x, BLASLONG incx, FLOAT *buffer, int nthr range_m[MAX_CPU_NUMBER - num_cpu - 1] = range_m[MAX_CPU_NUMBER - num_cpu] - width; range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16); - if (range_n[num_cpu] > m) range_n[num_cpu] = m; + if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu; queue[num_cpu].mode = mode; queue[num_cpu].routine = tpmv_kernel; @@ -343,7 +343,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *x, BLASLONG incx, FLOAT *buffer, int nthr range_m[num_cpu + 1] = range_m[num_cpu] + width; range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16); - if (range_n[num_cpu] > m) range_n[num_cpu] = m; + if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu; queue[num_cpu].mode = mode; queue[num_cpu].routine = tpmv_kernel;