Added a 32-bit version of the test, and constexpr.
Added warnings for missing or failed lit-cpuid. Not that there any failure conditions in the program right now ;-).
This test is pretty much a reimplementation of TestYMMRegisters dotest test. I like this one more for several reasons, but if we're going to check this in, we should also delete the other one.
Fixed unnecessary array in ymm_t decls.
Fixed unnecessary array in xmm_t decls.
I can confirm that it fixes our issue but I don't feel confident reviewing the code itself. Copying reviewers from the original diff.
Thu, Apr 18
Wed, Apr 17
Updated per request.
Tue, Apr 16
What's the value in checking in this xfail'ed test without an actual fix for the problem?
This change apparently introduced a regression: https://bugs.llvm.org/show_bug.cgi?id=41027
Mon, Apr 15
Sat, Apr 13
Thu, Apr 11
Will do, thanks.
Wed, Apr 10
Tue, Apr 9
Updated 'REQUIRES'. I've looked through other LLVM tests, and noticed there's already 'native' concept (when host==target), and some 'target*' features for systems. I've chosen to combine the former with new 'target-x86'. I'll link the other patch in a minute.
Mon, Apr 8
Rewritten the test to C++ with inline assembly, and switched to CHECK-DAG. Still need to figure out how to filter it for correct platform.
Thanks. I'll try converting it to C now.
Thanks, looks good. Do you have commit access or do you need me to commit it for you?
Fri, Apr 5
Just add it to the regex above.
Thu, Apr 4
Wed, Apr 3
Mon, Apr 1
Sun, Mar 31
Reordered as requested.
Sat, Mar 30
Fri, Mar 29
WFM. We've been stripping it in Gentoo since the inception.
Tue, Mar 26
Is there a reason you can't reuse/extend DynamicLibrary from LLVMSupport?
Mar 26 2019
Is there really any good case to build LLVM without zlib? Maybe it'd be better to just require zlib unconditionally (or mark building it really unsupported) rather than bundling extra algorithms that will be used only conditionally (until someone mistakenly uses it unconditionally, and LLVM would start using different CRC32 impls in different places) like this.
Mar 25 2019
Mar 21 2019
Now, the curious thing is that so far I have 2 more test failures with this. I'll investigate closer, maybe it's just flakiness.
Applied requested changes.
Thanks for the review. Addressed comments.
Covered more callsites.
Mar 20 2019
Got rid of close() and fclose() wrapping, for now.