floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix ZHEMM copy for SVE 

Whilst disambiguating whilelt, I inadvertantly used the wrong datatype
for offsets, which can be negative. This rectifies that.
This commit is contained in:
Chris Sidebottom 2023-07-27 13:27:28 +01:00
parent 9ba9c8bdc0
commit 730ca04b48
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
kernel/arm64/zhemm_ltcopy_sve.c
View File

@ -80,7 +80,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
gat_ind = svadd_m(cmp, gat_ind, lda_vec);
gat_ind = svadd_m(svnot_z(pg, cmp) , gat_ind, 2);
if (offset <= 0) {
svbool_t off_g = svwhilelt_b64((uint64_t)offset, (uint64_t)0LL);
svbool_t off_g = svwhilelt_b64((int64_t)offset, (int64_t)0LL);
data_vec_imag = svneg_m(data_vec_imag, off_g, data_vec_imag);
}
@ -143,7 +143,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
gat_ind = svadd_m(cmp, gat_ind, lda_vec);
gat_ind = svadd_m(svnot_z(pg, cmp) , gat_ind, 2);
if (offset <= 0) {
svbool_t off_g = svwhilelt_b32((uint32_t)offset, (uint32_t)0);
svbool_t off_g = svwhilelt_b32((int32_t)offset, (int32_t)0);
data_vec_imag = svneg_m(data_vec_imag, off_g, data_vec_imag);
}

4
kernel/arm64/zhemm_utcopy_sve.c
View File

@ -81,7 +81,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
gat_ind = svadd_m(svnot_z(pg, cmp) , gat_ind, lda_vec);
data_vec_imag = svneg_z(pg, data_vec_imag);
if (offset <= 0) {
svbool_t off_g = svwhilelt_b64((uint64_t)offset, (uint64_t)0LL);
svbool_t off_g = svwhilelt_b64((int64_t)offset, (int64_t)0LL);
data_vec_imag = svneg_m(data_vec_imag, off_g, data_vec_imag);
}
@ -143,7 +143,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
gat_ind = svadd_m(svnot_z(pg, cmp) , gat_ind, lda_vec);
data_vec_imag = svneg_z(pg, data_vec_imag);
if (offset <= 0) {
svbool_t off_g = svwhilelt_b32((uint32_t)offset, (uint32_t)0);
svbool_t off_g = svwhilelt_b32((int32_t)offset, (int32_t)0);
data_vec_imag = svneg_m(data_vec_imag, off_g, data_vec_imag);
}