From 95dca2dccc15047d3cca317c8264e22025f6f51b Mon Sep 17 00:00:00 2001
From: Sebastian Achilles <s.achilles@fz-juelich.de>
Date: Tue, 28 Nov 2023 12:08:11 +0100
Subject: [PATCH] add patch to fix [cz]asums in OpenBLAS 0.3.24

---
 .../o/OpenBLAS/OpenBLAS-0.3.24-GCC-13.2.0.eb  |  3 +
 .../OpenBLAS/OpenBLAS-0.3.24_fix-czasum.patch | 76 +++++++++++++++++++
 2 files changed, 79 insertions(+)
 create mode 100644 easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24_fix-czasum.patch

diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24-GCC-13.2.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24-GCC-13.2.0.eb
index eb7737a76aa..be80c46e97c 100644
--- a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24-GCC-13.2.0.eb
+++ b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24-GCC-13.2.0.eb
@@ -20,6 +20,7 @@ patches = [
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.23_fix-lapack-test.patch',
     'OpenBLAS-0.3.23_disable-DDRGES3-LAPACK-test.patch',
+    'OpenBLAS-0.3.24_fix-czasum.patch',
 ]
 checksums = [
     {'v0.3.24.tar.gz': 'ceadc5065da97bd92404cac7254da66cc6eb192679cf1002098688978d4d5132'},
@@ -34,6 +35,8 @@ checksums = [
     {'OpenBLAS-0.3.23_fix-lapack-test.patch': 'f6b3d81061f136e34aaf5359bb80fb9d2bba28825cc1dd26179b8dd01a9a0054'},
     {'OpenBLAS-0.3.23_disable-DDRGES3-LAPACK-test.patch':
      '36a16b4d3b867897413b43b774f8b57d641ad9a1b452e9de33ced198ab25e461'},
+    {'OpenBLAS-0.3.24_fix-czasum.patch':
+     '8132b87c519fb08caa3bd7291fe8a1d0e1afe6fcb667d16f3020b46122afe20c'},
 ]
 
 builddependencies = [
diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24_fix-czasum.patch b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24_fix-czasum.patch
new file mode 100644
index 00000000000..ae926d20167
--- /dev/null
+++ b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.24_fix-czasum.patch
@@ -0,0 +1,76 @@
+From f8ad5344c210960fc399ca5b0ad8559ab5ca253e Mon Sep 17 00:00:00 2001
+From: Bart Oldeman <bart.oldeman@calculquebec.ca>
+Date: Fri, 17 Nov 2023 23:49:34 +0000
+Subject: [PATCH] Fix casum fallback kernel.
+
+This kernel is only used on Skylake+ if the kernel with AVX512
+intrinsics can't be used, but used the variable x1 incorrectly
+in the tail end of the loop, as it is still at the initial
+value instead of where x points to.
+
+This caused 55 "other error"s in the LAPACK tests
+(https://github.com/OpenMathLib/OpenBLAS/issues/4282)
+
+This change makes casum.c as similar as possible as zasum.c,
+because zasum.c does this correctly.
+---
+ kernel/x86_64/casum.c | 24 ++++++++++++------------
+ 1 file changed, 12 insertions(+), 12 deletions(-)
+
+diff --git a/kernel/x86_64/casum.c b/kernel/x86_64/casum.c
+index e4d0543114..28d78d2797 100644
+--- a/kernel/x86_64/casum.c
++++ b/kernel/x86_64/casum.c
+@@ -9,12 +9,12 @@
+ #endif
+ 
+ #ifndef HAVE_CASUM_KERNEL
+-static FLOAT casum_kernel(BLASLONG n, FLOAT *x1)
++static FLOAT casum_kernel(BLASLONG n, FLOAT *x)
+ {
+ 
+     BLASLONG i=0;
+     BLASLONG n_8 = n & -8;
+-    FLOAT *x = x1;
++    FLOAT *x1 = x;
+     FLOAT temp0, temp1, temp2, temp3;
+     FLOAT temp4, temp5, temp6, temp7;
+     FLOAT sum0 = 0.0;
+@@ -24,14 +24,14 @@ static FLOAT casum_kernel(BLASLONG n, FLOAT *x1)
+     FLOAT sum4 = 0.0;
+     
+     while (i < n_8) {
+-        temp0 = ABS_K(x[0]);
+-        temp1 = ABS_K(x[1]);
+-        temp2 = ABS_K(x[2]);
+-        temp3 = ABS_K(x[3]);
+-        temp4 = ABS_K(x[4]);
+-        temp5 = ABS_K(x[5]);
+-        temp6 = ABS_K(x[6]);
+-        temp7 = ABS_K(x[7]);
++        temp0 = ABS_K(x1[0]);
++        temp1 = ABS_K(x1[1]);
++        temp2 = ABS_K(x1[2]);
++        temp3 = ABS_K(x1[3]);
++        temp4 = ABS_K(x1[4]);
++        temp5 = ABS_K(x1[5]);
++        temp6 = ABS_K(x1[6]);
++        temp7 = ABS_K(x1[7]);
+         
+         sum0 += temp0;
+         sum1 += temp1;
+@@ -43,12 +43,12 @@ static FLOAT casum_kernel(BLASLONG n, FLOAT *x1)
+         sum2 += temp6;
+         sum3 += temp7;
+         
+-        x+=8;
++        x1+=8;
+         i+=4;
+     }
+ 
+      while (i < n) {
+-        sum4 += (ABS_K(x1[0]) + ABS_K(x1[1]));
++        sum4 += ABS_K(x1[0]) + ABS_K(x1[1]);
+         x1 += 2;
+         i++;
+      }