On 19.05.2021 11:12, Takashi Yano via Cygwin wrote:
On Wed, 31 Mar 2021 21:03:03 +0200
Marco Atzeri wrote:
New releases 0.3.14-1 of


After upgrading libopenblas from 0.3.12-2 to 0.3.14-1,
octave 5.2.0 crashes on plot command in 64 bit environment.
In 32 bit cygwin (WoW64), this does not happen. Reverting
libopenblas to 0.3.12-1 resolve the issue.

octave:1> plot([])
fatal: caught signal Illegal instruction -- stopping myself...
Illegal instruction(core dumped)

My environment:
OS: Windows 10 20H2 (64 bit)
CPU: Inte Core i7 870 (Lynnfield)

[yano@Express5800-S70 ~]$ cygcheck -c cygwin octave libopenblas
Cygwin Package Information
Package              Version        Status
cygwin               3.2.0-1        OK
libopenblas          0.3.14-1       OK
octave               5.2.0-1        OK

By starting octave-cli in gdb, it stops at vmovaps instruction
in gemm_(). IIUC, vmovaps is an AVX instruction which is not
supported by Core i7 870.

In another PC with Core i7-4790 which supports AVX, the issue
does not occur.


Hi Takashi,

maybe is time to build the 0.3.15 ....
https://github.com/xianyi/OpenBLAS/releases

x86_64:

    added support for Intel Control-flow Enforcement Technology (CET)
    reverted the DOMATCOPY_RT code to the generic C version
    fixed a bug in the AVX512 SGEMM kernel introduced in 0.3.14
    fixed misapplication of -msse flag to non-SSE cpus in DYNAMIC_ARCH
    added support for compilation of the benchmarks on older OSX versions
    fixed propagation of the NO_AVX512 option in CMAKE builds
    fixed compilation of the AVX512 SGEMM kernel with clang-cl on Windows
fixed compilation of the CTESTs with INTERFACE64=1 (random faults on OSX) corrected the Haswell DROT kernel to require AVX2/FMA3 rather than AVX512


give me some time, openblas is a bit time consuming

Regards
Marco

--
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple

Reply via email to