Closed boegel closed 1 year ago
Does this patch fix it on aarch64? It's similar to what fixes it on Haswell, except for single precision instead of double precision.
--- OpenBLAS-0.3.23/lapack-netlib/TESTING/sgd.in.orig 2023-06-06 11:01:50.512947527 +0000
+++ OpenBLAS-0.3.23/lapack-netlib/TESTING/sgd.in 2023-06-06 11:02:05.318078733 +0000
@@ -1,6 +1,6 @@
SGS Data for the Real Nonsymmetric Schur Form Driver
5 Number of matrix dimensions
-2 6 10 12 20 30 Matrix dimensions
+6 2 10 12 20 30 Matrix dimensions
1 1 1 2 1 Parameters NB, NBMIN, NXOVER, NS, NBCOL
10 Threshold for test ratios
.TRUE. Put T to test the error exits
@bartoldeman This patch works like a charm on aarch64
:+1:
--> LAPACK TESTING SUMMARY <--
SUMMARY nb test run numerical error other error
================ =========== ================= ================
REAL 1328283 0 (0.000%) 0 (0.000%)
DOUBLE PRECISION 1327377 14 (0.001%) 0 (0.000%)
COMPLEX 788035 0 (0.000%) 0 (0.000%)
COMPLEX16 789102 0 (0.000%) 0 (0.000%)
--> ALL PRECISIONS 4232797 14 (0.000%) 0 (0.000%)
(and it doesn't cause trouble on Intel Haswell, although I am seeing 25 numerical errors there, 4 complex + 21 complex16)
Fixed in 0a838b38e5c2f10d9c1b4a15f5f8918e597d72d9
It looks like we have to reopen this issue as with https://github.com/easybuilders/easybuild-easyconfigs/commit/0a838b38e5c2f10d9c1b4a15f5f8918e597d72d9 the same issue (but with single precision) happens on PPC now: https://github.com/OpenMathLib/OpenBLAS/issues/4415
Fixing in #19495
cfr. https://github.com/easybuilders/easybuild-easyconfigs/pull/18012#issuecomment-1574982631