Hi, > 1. might I suggest using the {} form of quoting these regexes - it makes > them much more readable.
Sure. > 2. I think there are some syntax errors in the regexes for these tests > (because it’s very hard to check them when using the “” quotes). > > "(?:movdqu|movups)\[ \\t\]+\[^\n\]*%xmm1,\[^\n\r]*16" > ………………………………………………………^ missing \ (in several places) Sorry, I will change them. > 3. are you intending to update the tests? Yes, so could you tell me what does missing “_” means? I have some trouble building darwin target for now. > ==== As for the comments on the asm output. > > 1) it would seem that both comments can’t be correct (since they contradict!) > 2) AFAICT, None of the assemblers I use has any issue with either order > 3) perhaps there’s no assembler in use that cares any more > 4) clang produces symbol+offset for that case on Darwin (i.e. the same as > final.c). > That means the i386.c part should align with final.c, but I can't make the decision, and I'm not sure if there is more failure in x86 tests with this change. Iain Sandoe <i...@sandoe.co.uk> 于2020年11月4日周三 下午4:27写道: > > Hi, > > Hongyu Wang via Gcc <gcc@gcc.gnu.org> wrote: > > >> Maybe those scan-asm regexp are too strict and should be relaxed a > >> bit. > > > > I agree with this, since with -fPIC the code produced would be different, > > just use symbol + constant may be too strict. > > > > I think the scan-assembler could be reduced to > > /* { dg-final { scan-assembler "(?:movdqu|movups)\[ > > \\t\]+\[^\n\]*%xmm1,\[^\n\r]*16" } } */ > > 1. might I suggest using the {} form of quoting these regexes - it makes > them much more readable. > > … e.g. more like this: > {(?:movdqu|movups)[\t]+[^\n]*%xmm1,[^\n\r]*16} > > 2. I think there are some syntax errors in the regexes for these tests > (because it’s very hard to check them when using the “” quotes). > > "(?:movdqu|movups)\[ \\t\]+\[^\n\]*%xmm1,\[^\n\r]*16" > ………………………………………………………^ missing \ (in several places) > > 3. are you intending to update the tests? > > > ==== As for the comments on the asm output. > > 1) it would seem that both comments can’t be correct (since they contradict!) > 2) AFAICT, None of the assemblers I use has any issue with either order > 3) perhaps there’s no assembler in use that cares any more > 4) clang produces symbol+offset for that case on Darwin (i.e. the same as > final.c). > > thanks > Iain > -- Regards, Hongyu, Wang