-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
[Please CC me, I'm not subscribed]
Hello!
I noticed that gcc mixes x87 with sse code on x86_64. Consider the
following:
- -------------------------------------------------------------------
$ gcc -v
Reading specs from /usr/lib/gcc/x86_64-redhat-linux/3.4.2/specs
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man
- --infodir=/usr/share/info --enable-shared -
Thread model: posix
gcc version 3.4.2 20041017 (Red Hat 3.4.2-6.fc3)
$ uname -a
Linux amd64-linux1.sourceforge.net 2.6.9-1.667smp #1 SMP Tue Nov 2
15:09:11 EST 2004 x86_64 x86_64 x86_64 GNU/Linux
$ cat test.cc
void blub(double *out, double in)
{
float a, b;
b = in;
a = b+b;
*out = a;
}
$ gcc -S -O2 -o test.S test.cc
$ cat test.S
.file "test.cc"
.text
.align 2
.p2align 4,,15
.globl _Z4blubPdd
.type _Z4blubPdd, @function
_Z4blubPdd:
.LFB2:
cvtsd2ss %xmm0, %xmm0
addss %xmm0, %xmm0
movss %xmm0, -12(%rsp)
flds -12(%rsp)
fstpl (%rdi)
ret
.LFE2:
[..]
- -------------------------------------------------------------------
IMHO the flds/fstpl pair could easily replaced by cvtsd2ss. Is this
intentional or already fixed in later versions? Unfortunately I have no
other gcc to try out than the sourceforge compile farm.
(Additionally it looks rather odd that it's using a stack slot without
allocating it, but I'm not that familiar with the x86_64 platform/ABI)
Sebastian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iQEVAwUBRCKyjv81M8QtvOSJAQIHmAgAjTqNBrxCx70r7/2sa5tP/zqOmWTsO5fx
W/QLAori7tlhTuvtZXoQRuNvBavaIySv+qCyVYrfz6PwYCws2k132ED6CtyBTIZg
Gm9/9O29tLxywZXoo2oYFIjClIe5flfGFE7g0PdpHOoHaKrB4kZuhS+fXWFLeNBY
CkTK26BAzt4kk2pLS4hWHeXfsGNmWPSeY4grMP+3TlEhwaFyHeGv4Ylh/5Z2j2zq
IxONCTBxSBjmd9g1T7SsQkDmonwzzcdV3zdaDZNEFtq1O3UBpvFxTcjPaHXlU7kJ
/2uwc38mb45qe3wH+qcoWFv2mcQA9U7xz0VcrcMGhNL7ugnh8D0oHA==
=1KRj
-----END PGP SIGNATURE-----
