Actually, movq is recognized by the assembler on Leopard. I've fixed this: http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of- Mon-20071210/056334.html Please verify. Thanks. Evan On Dec 14, 2007, at 10:12 AM, Evan Cheng wrote:> > On Dec 12, 2007, at 7:42 PM, David A. Greene wrote: > >> On Wednesday 12 December 2007 06:16:25 pm Evan Cheng wrote: >> >>> Hold on, changing to movq will break it on Mac OS X. I'll need to >>> find a proper solution. >>> >>> Thanks, >> >> The proper solution is to fix the assembler, but in the meantime, >> the mr and rm variants at least should be changed to movq. They >> are broken on MacOS X right now anyway. From what you've said, >> one can't do a 32-bit memory mov to/from an xmm on MacOS X. >> You still won't be able to do a 32-bit memory operation but at least >> systems that use the latest binutils will work properly. > > I should get an answer today. The right workaround now is to disable > this for Darwin. > > Thanks! > > Evan > >> >> -Dave >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
On Friday 14 December 2007 13:55, Evan Cheng wrote:> Actually, movq is recognized by the assembler on Leopard. I've fixed > this: > http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of- > Mon-20071210/056334.html > > Please verify. Thanks.The patch is similar to what I did. I'll have to create an llvm testcase to verify it as this is a Fortran code. It would also be good to have a regression test for this anyway. Does the Leopard assembler support movq for rr instructions? The rep64 movd sequence is ugly and esoteric. It's much more readable to say movq. -Dave
On Dec 17, 2007, at 10:04 AM, David Greene wrote:> On Friday 14 December 2007 13:55, Evan Cheng wrote: >> Actually, movq is recognized by the assembler on Leopard. I've fixed >> this: >> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of- >> Mon-20071210/056334.html >> >> Please verify. Thanks. > > The patch is similar to what I did. I'll have to create an llvm > testcase to > verify it as this is a Fortran code. It would also be good to have a > regression test for this anyway. > > Does the Leopard assembler support movq for rr instructions? The > rep64 movd sequence is ugly and esoteric. It's much more readable > to say movq.I agree. But unfortunately the assembler still does not recognize movq for rr instructions. If this really bug you I'd suggest adding a comment to the emitted assembly string. Evan> > > -Dave > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev