https://bugs.kde.org/show_bug.cgi?id=438640
--- Comment #22 from Mark Wielaard <[email protected]> --- (In reply to mcermak from comment #20) > Created attachment 193287 [details] > updated patch > > (In reply to Mark Wielaard from comment #19) > > This looks good, but could the new crc32 testcases be added to the crc32.c > > testcase instead of the sse4 one? The sse4-x86 is already pretty big and > > matches > > the sse4-64 one (mostly). Best (imho) to keep the crc32 ones separate. > > Yup, makes sense! Please check the updated patch. Note that I've been > trying to > reuse none/tests/sse4-common.h in my new version of none/tests/x86/crc32.c. > But there are some problems like test_PEXTRD() depends on > DO_imm_r_to_mandrscalar > macro defined elsewhere. This can be guarded by #if > defined(DO_imm_r_to_mandrscalar). > But then there are things like zeroes, fives etc, which would become unused > in crc32.c ... > ... and ... I've just concluded that a conservative solution of duplicating > randUChar(), > randULong() and randUInt() is ... less messy. Yeah, maybe we should split out some of the helper functions in "common" to "more-common". But just copying the rand functions into this testcase seems the simplest solution. > Please review. Looks good. Please commit. -- You are receiving this mail because: You are watching all bug changes.
