https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
--- Comment #6 from Petr ---
Ok, that's fair enough. I didn't know GCC needs an additional option to switch
to fully compatible Intel syntax. The code that I posted works fine in clang,
so sorry about that.
And yes, the instruction will #UD,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
--- Comment #5 from Jakub Jelinek ---
Ah, if you want to avoid the %xmm4, %r13, %xmm3, etc., you need .intel_syntax
noprefix.
And, the insn is invalid anyway, because you can't have the same mask and
destination.
In any case, this has nothing to
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
--- Comment #4 from Petr ---
In this case, DWORD PTR is redundant, nasm and yasm is fine with the syntax I
posted as well.
It's a simplified test just to show that it won't pass.
Try:
__asm(".intel_syntax\n"
"vpgatherdd xmm4, dword
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
--- Comment #3 from Jakub Jelinek ---
Perhaps something to report to llvm that they accept invalid, that is a bug.
Also, why do you use Intel syntax when you don't know how should it be used?
In AT syntax you wouldn't need to specify this extra
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79880
Andrew Pinski changed:
What|Removed |Added
Status|UNCONFIRMED |RESOLVED
Resolution|---