Correct range_n limiting
same bug as seen in #1388, somehow missed in corresponding PR #1389
This commit is contained in:
parent
69edc5bbe7
commit
78d9910236
|
@ -346,8 +346,8 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG incx, FLOAT *bu
|
||||||
|
|
||||||
range_m[MAX_CPU_NUMBER - num_cpu - 1] = range_m[MAX_CPU_NUMBER - num_cpu] - width;
|
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);
|
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].mode = mode;
|
||||||
queue[num_cpu].routine = trmv_kernel;
|
queue[num_cpu].routine = trmv_kernel;
|
||||||
queue[num_cpu].args = &args;
|
queue[num_cpu].args = &args;
|
||||||
|
@ -386,8 +386,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG incx, FLOAT *bu
|
||||||
|
|
||||||
range_m[num_cpu + 1] = range_m[num_cpu] + width;
|
range_m[num_cpu + 1] = range_m[num_cpu] + width;
|
||||||
range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16);
|
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].mode = mode;
|
||||||
queue[num_cpu].routine = trmv_kernel;
|
queue[num_cpu].routine = trmv_kernel;
|
||||||
queue[num_cpu].args = &args;
|
queue[num_cpu].args = &args;
|
||||||
|
|
Loading…
Reference in New Issue