Merge pull request easybuilders#19159 from Flamefire/20231106133537_n…

…ew_pr_OpenBLAS0320 fix OpenBLAS on newer Intel CPUs
CLIP-HPC · Nov 7, 2023 · c757d75 · c757d75
2 parents e8383fa + 9b85f42
commit c757d75
Show file tree

Hide file tree

Showing 7 changed files with 88 additions and 5 deletions.
diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-GCC-11.2.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-GCC-11.2.0.eb
@@ -17,6 +17,7 @@ patches = [
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
     'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch',
     'OpenBLAS-0.3.21_avoid-crash-in-zdot.patch',
@@ -28,6 +29,8 @@ checksums = [
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
     {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch':

diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-GCC-11.3.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-GCC-11.3.0.eb
@@ -16,26 +16,29 @@ patches = [
     ('large.tgz', '.'),
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
+    'OpenBLAS-0.3.20_fix-cpuid-neoverse-v1-n2.patch',
+    'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch',
     'OpenBLAS-0.3.21_avoid-crash-in-zdot.patch',
-    'OpenBLAS-0.3.20_fix-cpuid-neoverse-v1-n2.patch',
-    'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
 ]
 checksums = [
     {'v0.3.20.tar.gz': '8495c9affc536253648e942908e88e097f2ec7753ede55aca52e5dead3029e3c'},
     {'large.tgz': 'f328d88b7fa97722f271d7d0cfea1c220e0f8e5ed5ff01d8ef1eb51d6f4243a1'},
     {'timing.tgz': '999c65f8ea8bd4eac7f1c7f3463d4946917afd20a997807300fe35d70122f3af'},
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
+    {'OpenBLAS-0.3.20_fix-cpuid-neoverse-v1-n2.patch':
+     '1b495465f8dd1e151d74cf5aa4288120361d29164d6a377228a8d51c255b8a9e'},
+    {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch':
      'bd6836206a883208dc8bc997946f97e4c97d91d8e101fc54db414aaa56902fc3'},
     {'OpenBLAS-0.3.21_avoid-crash-in-zdot.patch': '3dac2c1ec896df574f1b37cde81a16f24550b7f1eb81fbfacb0c4449b0dc7894'},
-    {'OpenBLAS-0.3.20_fix-cpuid-neoverse-v1-n2.patch':
-     '1b495465f8dd1e151d74cf5aa4288120361d29164d6a377228a8d51c255b8a9e'},
-    {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
 ]
 
 builddependencies = [

diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-NVHPC-22.7-CUDA-11.7.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20-NVHPC-22.7-CUDA-11.7.0.eb
@@ -17,6 +17,7 @@ patches = [
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
     'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch',
     'OpenBLAS-0.3.21_avoid-crash-in-zdot.patch',
@@ -28,6 +29,8 @@ checksums = [
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
     {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.21_disable-fma-in-cscal-zscal.patch':

diff --git a/...ild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch b/...ild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch
@@ -0,0 +1,65 @@
+From 9019bc494514a74c2042152cdca0a36adea7b42f Mon Sep 17 00:00:00 2001
+From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
+Date: Sat, 4 Nov 2023 22:10:06 +0100
+Subject: [PATCH] Use SkylakeX ?ASUM microkernel for Cooperlake/Sapphirerapids
+ as well
+
+---
+ kernel/x86_64/casum.c | 2 +-
+ kernel/x86_64/dasum.c | 2 +-
+ kernel/x86_64/sasum.c | 2 +-
+ kernel/x86_64/zasum.c | 2 +-
+ 4 files changed, 4 insertions(+), 4 deletions(-)
+
+diff --git a/kernel/x86_64/casum.c b/kernel/x86_64/casum.c
+index 60feec0ceb..e4d0543114 100644
+--- a/kernel/x86_64/casum.c
++++ b/kernel/x86_64/casum.c
+@@ -4,7 +4,7 @@
+ #define ABS_K(a) ((a) > 0 ? (a) : (-(a)))
+ #endif
+
+-#if defined(SKYLAKEX)
++#if defined(SKYLAKEX) || defined(COOPERLAKE) || defined(SAPPHIRERAPIDS)
+ #include "casum_microk_skylakex-2.c"
+ #endif
+
+diff --git a/kernel/x86_64/dasum.c b/kernel/x86_64/dasum.c
+index a9c40f38f0..0147c6978a 100644
+--- a/kernel/x86_64/dasum.c
++++ b/kernel/x86_64/dasum.c
+@@ -4,7 +4,7 @@
+ #define ABS_K(a) ((a) > 0 ? (a) : (-(a)))
+ #endif
+
+-#if defined(SKYLAKEX)
++#if defined(SKYLAKEX) || defined(COOPERLAKE) || defined(SAPPHIRERAPIDS)
+ #include "dasum_microk_skylakex-2.c"
+ #elif defined(HASWELL) || defined(ZEN)
+ #include "dasum_microk_haswell-2.c"
+diff --git a/kernel/x86_64/sasum.c b/kernel/x86_64/sasum.c
+index 37a92468ff..3f22cb97a1 100644
+--- a/kernel/x86_64/sasum.c
++++ b/kernel/x86_64/sasum.c
+@@ -9,7 +9,7 @@
+
+ #endif
+
+-#if defined(SKYLAKEX)
++#if defined(SKYLAKEX) || defined(COOPERLAKE) || defined(SAPPHIRERAPIDS)
+ #include "sasum_microk_skylakex-2.c"
+ #elif defined(HASWELL) || defined(ZEN)
+ #include "sasum_microk_haswell-2.c"
+diff --git a/kernel/x86_64/zasum.c b/kernel/x86_64/zasum.c
+index 80e95a2c89..3f17ab1cfa 100644
+--- a/kernel/x86_64/zasum.c
++++ b/kernel/x86_64/zasum.c
+@@ -4,7 +4,7 @@
+ #define ABS_K(a) ((a) > 0 ? (a) : (-(a)))
+ #endif
+
+-#if defined(SKYLAKEX)
++#if defined(SKYLAKEX) || defined(COOPERLAKE) || defined(SAPPHIRERAPIDS)
+ #include "zasum_microk_skylakex-2.c"
+ #endif
+
diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.21-GCC-12.2.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.21-GCC-12.2.0.eb
@@ -17,6 +17,7 @@ patches = [
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
     'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.21-GCC-12.2.0_disable-fma-in-cscal-zscal.patch',
     'OpenBLAS-0.3.21_avoid-crash-in-zdot.patch',
@@ -28,6 +29,8 @@ checksums = [
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
     {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.21-GCC-12.2.0_disable-fma-in-cscal-zscal.patch':

diff --git a/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.23-GCC-12.3.0.eb b/easybuild/easyconfigs/o/OpenBLAS/OpenBLAS-0.3.23-GCC-12.3.0.eb
@@ -17,6 +17,7 @@ patches = [
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
     'OpenBLAS-0.3.20_fix-x86-cpuid.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     'OpenBLAS-0.3.21_fix-order-vectorization.patch',
     'OpenBLAS-0.3.23_fix-parallel-build.patch',
     'OpenBLAS-0.3.23_fix-lapack-test.patch',
@@ -30,6 +31,8 @@ checksums = [
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
     {'OpenBLAS-0.3.20_fix-x86-cpuid.patch': '57e8384404e136b9f0dafc26573adeb7dc69e60d84a7e189643b91d6299888fc'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.23_fix-parallel-build.patch': 'abe10ba3b0ca54772dbf235596e35325a5159018f6a60cfc88824c2c220d99d9'},

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
@@ -16,6 +16,7 @@ patches = [
     ('large.tgz', '.'),
     ('timing.tgz', '.'),
     'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch',
+    'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch',
     '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',
@@ -26,6 +27,8 @@ checksums = [
     {'timing.tgz': '999c65f8ea8bd4eac7f1c7f3463d4946917afd20a997807300fe35d70122f3af'},
     {'OpenBLAS-0.3.15_workaround-gcc-miscompilation.patch':
      'e6b326fb8c4a8a6fd07741d9983c37a72c55c9ff9a4f74a80e1352ce5f975971'},
+    {'OpenBLAS-0.3.20_use-xASUM-microkernels-on-new-intel-cpus.patch':
+     '1dbd0f9473963dbdd9131611b455d8a801f1e995eae82896186d3d3ffe6d5f03'},
     {'OpenBLAS-0.3.21_fix-order-vectorization.patch':
      '08af834e5d60441fd35c128758ed9c092ba6887c829e0471ecd489079539047d'},
     {'OpenBLAS-0.3.23_fix-lapack-test.patch': 'f6b3d81061f136e34aaf5359bb80fb9d2bba28825cc1dd26179b8dd01a9a0054'},