* [PATCH] Fix AVX check
@ 2011-07-21 14:04 Andreas Schwab
0 siblings, 0 replies; only message in thread
From: Andreas Schwab @ 2011-07-21 14:04 UTC (permalink / raw)
To: libc-hacker
2011-07-21 Andreas Schwab <schwab@redhat.com>
* sysdeps/x86_64/dl-trampoline.S (_dl_runtime_profile): Fix last
change.
(_dl_x86_64_save_sse): Use correct AVX check.
---
sysdeps/x86_64/dl-trampoline.S | 21 +++++++++++++++------
1 files changed, 15 insertions(+), 6 deletions(-)
diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 1b97929..ad7479f 100644
--- a/sysdeps/x86_64/dl-trampoline.S
+++ b/sysdeps/x86_64/dl-trampoline.S
@@ -150,7 +150,8 @@ L(have_avx):
cmpl $0x6, %eax
// Nonzero if SSE and AVX state saving is enabled.
sete %al
-2: movl %eax, L(have_avx)(%rip)
+2: leal -1(%eax,%eax), %eax
+ movl %eax, L(have_avx)(%rip)
cmpl $0, %eax
1: js L(no_avx)
@@ -183,11 +184,19 @@ _dl_x86_64_save_sse:
movl $1, %eax
cpuid
movq %r11,%rbx # Restore rbx
- movl $1, %eax
- testl $(1 << 28), %ecx
- jne 2f
- negl %eax
-2: movl %eax, L(have_avx)(%rip)
+ xorl %eax, %eax
+ // AVX and XSAVE supported?
+ testl $((1 << 28) | (1 << 27)), %ecx
+ je 2f
+ xorl %ecx, %ecx
+ // Get XFEATURE_ENABLED_MASK
+ xgetbv
+ andl $0x6, %eax
+ cmpl $0x6, %eax
+ // Nonzero if SSE and AVX state saving is enabled.
+ sete %al
+2: leal -1(%eax,%eax), %eax
+ movl %eax, L(have_avx)(%rip)
cmpl $0, %eax
1: js L(no_avx5)
--
1.7.6
--
Andreas Schwab, schwab@redhat.com
GPG Key fingerprint = D4E8 DBE3 3813 BB5D FA84 5EC7 45C6 250E 6F00 984E
"And now for something completely different."
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2011-07-21 14:04 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-07-21 14:04 [PATCH] Fix AVX check Andreas Schwab
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).