• mclysenk mtu.edu (4/4) Aug 10 2005 The movss instruction doesn't seem to be assembling correctly
• mclysenk mtu.edu (13/17) Aug 10 2005 A correction, it's not a bug in movss, but in sqrtps
• Walter (3/7) Aug 11 2005 Shalzbat! This affects obj2asm as well, since it shares code with the co...

mclysenk mtu.edu writes:

The movss instruction doesn't seem to be assembling correctly

movss [ECX], XMM0

Will be compile just fine, but the outputs

movups [ECX], XMM0

mclysenk mtu.edu writes:

In article <ddecd6$1p46$1 digitaldaemon.com>, mclysenk mtu.edu says...
The movss instruction doesn't seem to be assembling correctly

movss [ECX], XMM0

Will be compile just fine, but the outputs

movups [ECX], XMM0

A correction, it's not a bug in movss, but in sqrtps
Here was an excerpt of from the code

sqrtps XMM2, XMM1;
movss [EBX], XMM0;


Which emitted the following assembly

0f 51 f3 0f 11 03
^
Missing ModRM byte

This translates into the following in correct code

sqrtps XMM6, XMM3;
movups [EBX], XMM0;

The problem also extends to the sqrtss instruction.

"Walter" <newshound digitalmars.com> writes:

<mclysenk mtu.edu> wrote in message news:ddecd6$1p46$1 digitaldaemon.com...
 The movss instruction doesn't seem to be assembling correctly

 movss [ECX], XMM0

 Will be compile just fine, but the outputs

 movups [ECX], XMM0

Shalzbat! This affects obj2asm as well, since it shares code with the code
generator. Anyhow, it'll get fixed in the next update.