[llvm-commits] [llvm] r77940 - in /llvm/trunk: lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86InstrMMX.td test/CodeGen/X86/2009-08-02-mmx-scalar-to-vector.ll test/CodeGen/X86/mmx-bitcast-to-i64.ll

[Eli Friedman]

On Sun, Aug 2, 2009 at 7:53 PM, Rafael Espindola<espindola at google.com> wrote:
>> AFAIK, MMX_MOVD64from64rr was originally written as "movd" for
>> compatibility reasons; why are you changing it?
>
> Because it is supposed to move 64 bits. movd is 32 bits, right?
>
> I checked with gcc 4.3 and it uses movq.

Really?  I get two movd's for the following testcase:

#include <mmintrin.h>
long long a(__m64 x) { return _mm_cvtm64_si64(x); }
__m64 b(long long x) { return _mm_cvtsi64_m64(x); }

-Eli