diff --git a/kernel/x86/gemv_t_sse.S b/kernel/x86/gemv_t_sse.S index c72febe3d..42ed19998 100644 --- a/kernel/x86/gemv_t_sse.S +++ b/kernel/x86/gemv_t_sse.S @@ -236,6 +236,20 @@ jg .L06 ALIGN_4 +//Padding zero to prevent loading the dirty number from buffer. + movl M, I + movl $8, J + andl $7, I + xorps %xmm0, %xmm0 + subl I, J + ALIGN_2 +.L07: + movss %xmm0, 0 * SIZE(Y1) + addl $SIZE, Y1 + decl J + jg .L07 + ALIGN_4 + .L10: movl Y, Y1