From 934e601e934f5cf930382dfd9d7e92b937d1d2ed Mon Sep 17 00:00:00 2001 From: wjc404 <52632443+wjc404@users.noreply.github.com> Date: Thu, 28 Nov 2019 19:56:35 +0800 Subject: [PATCH] Update dgemm_kernel_4x8_skylakex_2.c --- kernel/x86_64/dgemm_kernel_4x8_skylakex_2.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/kernel/x86_64/dgemm_kernel_4x8_skylakex_2.c b/kernel/x86_64/dgemm_kernel_4x8_skylakex_2.c index 51b0b94..90a4c2b 100644 --- a/kernel/x86_64/dgemm_kernel_4x8_skylakex_2.c +++ b/kernel/x86_64/dgemm_kernel_4x8_skylakex_2.c @@ -97,15 +97,17 @@ /* %10 for prefetch of C elements before storage; %4 = ldc(in bytes),%11 for prefetch of next B block */ #define INNER_KERNELm8(nn) \ - "movq %3,%10;cmpq $16,%2;jb "#nn"001f;"\ + "movq %3,%10;cmpq $18,%2;jb "#nn"001f;"\ #nn"008:\n\t"\ INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ + INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ "prefetcht1 (%10); prefetcht1 63(%10); addq %4,%10;"\ INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ - "prefetcht1 (%11); addq $16,%11;"\ - "subq $4,%2;cmpq $16,%2;jnb "#nn"008b;"\ + INNER_KERNEL_k1m8n##nn "addq $64,%1;"\ + "prefetcht1 (%11); addq $32,%11;"\ + "subq $6,%2;cmpq $18,%2;jnb "#nn"008b;"\ "movq %3,%10;"\ #nn"001:\n\t"\ "cmpq $1,%2;jb "#nn"000f;"\ -- 2.7.4