prepared lapack/lauum for UNROLL values, that are not a power of two
This commit is contained in:
parent
1a39b92b1d
commit
209b63197e
|
@ -88,7 +88,7 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa,
|
|||
newarg.beta = NULL;
|
||||
newarg.nthreads = args -> nthreads;
|
||||
|
||||
blocking = (n / 2 + GEMM_UNROLL_N - 1) & ~(GEMM_UNROLL_N - 1);
|
||||
blocking = ((n / 2 + GEMM_UNROLL_N - 1)/GEMM_UNROLL_N) * GEMM_UNROLL_N;
|
||||
if (blocking > GEMM_Q) blocking = GEMM_Q;
|
||||
|
||||
for (i = 0; i < n; i += blocking) {
|
||||
|
|
|
@ -88,7 +88,7 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa,
|
|||
newarg.beta = NULL;
|
||||
newarg.nthreads = args -> nthreads;
|
||||
|
||||
blocking = (n / 2 + GEMM_UNROLL_N - 1) & ~(GEMM_UNROLL_N - 1);
|
||||
blocking = ((n / 2 + GEMM_UNROLL_N - 1)/GEMM_UNROLL_N) * GEMM_UNROLL_N;
|
||||
if (blocking > GEMM_Q) blocking = GEMM_Q;
|
||||
|
||||
for (i = 0; i < n; i += blocking) {
|
||||
|
|
Loading…
Reference in New Issue