A number of tests fail on PowerPC (simde backend): [ PASSED ] 2299 tests, [ FAILED ] 685 tests

[barracuda156]

This is on ppc (32-bit BE):

[----------] Global test environment tear-down
[==========] 2984 tests from 36 test cases ran. (3185408 ms total)
[  PASSED  ] 2299 tests.
[  FAILED  ] 685 tests, listed below:

Details:

tests.txt

The first question: has this been tested on Linux ppc64 (BE, of course) and/or s390x? (I am pretty sure no one tested on 32-bit ppc on any OS.) Debian and FreeBSD support ppc64, s390x is supported at least on Linux, so these should be feasible.

And could someone say if failing tests fit into some known pattern? Endianness, bitness, size of bool?

[markos]

hi @barracuda156, I have tested it and even started a ppc64 port as I have lots of Power hadware and even access to s390x VMs, however it will need substantial work to make it BE-friendly. It's possible that a s390x port will offer the best possibility for BE support and from there ppc64 support. For now I will move this to wishlist as there are no resources to dedicate to this effort.

[barracuda156]

Thank you! Fixing for s390x will be very helpful indeed, since endianness issues will be common across BE platforms, and s390x probably uses the same long double format as Darwin on ppc and AIX (I have no idea if it is relevant to the code here, but it may be).