From 6f01cd2c548e209f83d26a2fbb139ba7cc4a9c31 Mon Sep 17 00:00:00 2001 From: Nallani Bhaskar Date: Mon, 1 Jun 2020 17:04:59 +0530 Subject: [PATCH] Fix for sblat3.x failure in make check Details: Using of ymm registers storing 8 float values than 4 floats values Changed register from ymm to xmm in required places. This can be found only when leading dimension is greater than the actual dimension. Change-Id: I39f04eac18c4fa3a8c93048c977d6a83aa92b800 --- kernels/zen/3/sup/bli_gemmsup_rd_zen_asm_s6x16.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/kernels/zen/3/sup/bli_gemmsup_rd_zen_asm_s6x16.c b/kernels/zen/3/sup/bli_gemmsup_rd_zen_asm_s6x16.c index ffcb507ad..96bc92749 100644 --- a/kernels/zen/3/sup/bli_gemmsup_rd_zen_asm_s6x16.c +++ b/kernels/zen/3/sup/bli_gemmsup_rd_zen_asm_s6x16.c @@ -290,9 +290,9 @@ void bli_sgemmsup_rd_zen_asm_2x16 label(.SBETAZERO) label(.SROWSTORBZ) - vmovups(ymm4, mem(rcx)) + vmovups(xmm4, mem(rcx)) add(rdi, rcx) - vmovups(ymm5, mem(rcx)) + vmovups(xmm5, mem(rcx)) label(.SDONE) add(imm(4), r15) // jj += 4; @@ -817,9 +817,9 @@ void bli_sgemmsup_rd_zen_asm_2x8 label(.SBETAZERO) label(.SROWSTORBZ) - vmovups(ymm4, mem(rcx)) + vmovups(xmm4, mem(rcx)) add(rdi, rcx) - vmovups(ymm5, mem(rcx)) + vmovups(xmm5, mem(rcx)) label(.SDONE) @@ -1051,7 +1051,7 @@ void bli_sgemmsup_rd_zen_asm_1x8 label(.SBETAZERO) label(.SROWSTORBZ) - vmovups(ymm4, mem(rcx)) + vmovups(xmm4, mem(rcx)) label(.SDONE) add(imm(4), r15) // jj += 4; @@ -1323,9 +1323,9 @@ void bli_sgemmsup_rd_zen_asm_2x4 label(.SBETAZERO) label(.SROWSTORBZ) - vmovups(ymm4, mem(rcx)) + vmovups(xmm4, mem(rcx)) add(rdi, rcx) - vmovups(ymm5, mem(rcx)) + vmovups(xmm5, mem(rcx)) label(.SDONE) end_asm( : // output operands (none) @@ -1536,7 +1536,7 @@ void bli_sgemmsup_rd_zen_asm_1x4 label(.SBETAZERO) label(.SROWSTORBZ) - vmovups(ymm4, mem(rcx)) + vmovups(xmm4, mem(rcx)) label(.SDONE)