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

fix sve ztrmm kernel

This commit is contained in:
Bine Brank 2022-01-04 14:42:07 +01:00
parent 68c414d3a6
commit 2e2c02b762
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
kernel/arm64/ztrmm_kernel_sve_v1x4.S
View File

@ -723,7 +723,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
add pB, pB, temp
#endif
#if defined(LEFT)
add tempOffset, tempOffset, #4
add tempOffset, tempOffset, lanes
#endif
prfm PLDL1KEEP, [pA]
@ -856,7 +856,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
add pB, pB, temp
#endif
#if defined(LEFT)
add tempOffset, tempOffset, #4
add tempOffset, tempOffset, lanes
#endif
.Lztrmm_kernel_L2_Mv1_END:
@ -923,7 +923,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
add tempK, tempOffset, #1
#endif
asr counterL , temp, #3 // counterL = counterL / 8
asr counterL , tempK, #3 // counterL = counterL / 8
cmp counterL , #0
ble .Lztrmm_kernel_L1_Mv1_40
.align 5
@ -972,7 +972,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
add pB, pB, temp
#endif
#if defined(LEFT)
add tempOffset, tempOffset, #4
add tempOffset, tempOffset, lanes
#endif
.Lztrmm_kernel_L1_Mv1_END: