This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Remove incorrect register mov in floorf on x86_64
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: "Carlos O'Donell" <carlos at redhat dot com>
- Cc: Siddhesh Poyarekar <siddhesh at redhat dot com>, OndÅej BÃlka <neleai at seznam dot cz>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Fri, 14 Aug 2015 13:36:18 -0700
- Subject: Re: [PATCH] Remove incorrect register mov in floorf on x86_64
- Authentication-results: sourceware.org; auth=none
- References: <20150814103330 dot GA15748 at spoyarek dot pnq dot redhat dot com> <20150814111204 dot GA20519 at domone> <20150814111949 dot GK24813 at spoyarek dot pnq dot redhat dot com> <CAMe9rOqzxjfmGg1sJGPUtwmWWN58vQwUw711UfukNcRP_MUAKg at mail dot gmail dot com> <55CE4A20 dot 3060903 at redhat dot com>
On Fri, Aug 14, 2015 at 1:05 PM, Carlos O'Donell <carlos@redhat.com> wrote:
> On 08/14/2015 08:34 AM, H.J. Lu wrote:
>> On Fri, Aug 14, 2015 at 4:19 AM, Siddhesh Poyarekar <siddhesh@redhat.com> wrote:
>>> On Fri, Aug 14, 2015 at 01:12:04PM +0200, OndÅej BÃlka wrote:
>>>> It isn't incorrect, just dead move. Ok for me.
>>>
>>> No, it loads an arbitrary value (we don't know what is in %rax at that
>>> point) into %rdx and overwrites the struct address.
>>>
>>
>> Thanks for catching it. I checked in your patch together with the same
>> fix for sysdeps/x86_64/fpu/multiarch/s_nearbyint.S
>
> Why didn't testing catch this? Is it because this is a multiarch that
> didn't get tested?
RAX contains a valid address, but not the correct address. Since
my machine has SSE4.1, any version works. But machines without
SSE4.1 may fail.
--
H.J.