openblas/2018.patch

28 lines
1013 B
Diff

From 69a97ca7b9d7bbbb9b9f018592586e3c17b51a57 Mon Sep 17 00:00:00 2001
From: Bart Oldeman <bart.oldeman@calculquebec.ca>
Date: Thu, 14 Feb 2019 16:19:41 +0000
Subject: [PATCH] dgemv_kernel_4x4(Haswell): add missing clobbers for
xmm0,xmm1,xmm2,xmm3
This fixes a crash in dblat2 when OpenBLAS is compiled using
-march=znver1 -ftree-vectorize -O2
See also:
https://github.com/easybuilders/easybuild-easyconfigs/issues/7180
---
kernel/x86_64/dgemv_n_microk_haswell-4.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/kernel/x86_64/dgemv_n_microk_haswell-4.c b/kernel/x86_64/dgemv_n_microk_haswell-4.c
index 584a6c6b5..da0fa2fff 100644
--- a/kernel/x86_64/dgemv_n_microk_haswell-4.c
+++ b/kernel/x86_64/dgemv_n_microk_haswell-4.c
@@ -104,6 +104,7 @@ static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT
"r" (ap[3]), // 7
"r" (alpha) // 8
: "cc",
+ "%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5",
"%xmm6", "%xmm7",
"%xmm8", "%xmm9",