Home Original page

[PATCH] x86: Properly decode EVEX.W in vcvt[u]si2s[sd] in 32-bit

H.J. Lu hjl.tools@gmail.com
Mon Sep 17 12:33:00 GMT 2018 
On Mon, Sep 17, 2018 at 5:18 AM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> On 16.09.18 at 14:17, <hjl.tools@gmail.com> wrote:
>> On Sun, Sep 16, 2018 at 1:57 AM, Jan Beulich <jbeulich@suse.com> wrote:
>>>>>> "H.J. Lu" <hjl.tools@gmail.com> 09/14/18 7:47 PM >>>
>>>>--- /dev/null
>>>>+++ b/gas/testsuite/gas/i386/evex.d
>>>>@@ -0,0 +1,16 @@
>>>>+#objdump: -dw -Msuffix
>>>>+#name: i386 EVX insns
>>>>+
>>>>+.*: +file format .*
>>>>+
>>>>+
>>>>+Disassembly of section .text:
>>>>+
>>>>+0+ <_start>:
>>>>+ +[a-f0-9]+:  62 f1 d6 38 2a f0       vcvtsi2ssl %eax,\{rd-sae\},%xmm5,%xmm6
>>>>+ +[a-f0-9]+:  62 f1 d7 38 2a f0       vcvtsi2sdl %eax,\{rd-sae\},%xmm5,%xmm6
>>>>+ +[a-f0-9]+:  62 f1 d6 08 7b f0       vcvtusi2ssl %eax,%xmm5,%xmm6
>>>>+ +[a-f0-9]+:  62 f1 d7 08 7b f0       vcvtusi2sdl %eax,%xmm5,%xmm6
>>>>+ +[a-f0-9]+:  62 f1 d6 38 7b f0       vcvtusi2ssl %eax,\{rd-sae\},%xmm5,%xmm6
>>>>+ +[a-f0-9]+:  62 f1 d7 38 7b f0       vcvtusi2sdl %eax,\{rd-sae\},%xmm5,%xmm6
>>>
>>> Hmm, a new test demanding (according to what you've told me in earlier
>>> discussions) bad behavior: You've said that you don't want suffixes on newer
>>> insns when they're not needed. While these insns may indeed better have
>>> suffixes in 64-bit mode (they strictly need them only with memory operands),
>>> there's clearly nothing to disambiguate in 16- and 32-bit modes. May I ask
>>> for consistency please between what you demand for patches I submit and
>>> ones you commit, once again without even giving a little time for reviews?
>>>
>>
>> I didn't add any new instructions.  These testcases are written in .byte.
>> I just fixed the existing entries in disassembler.
>
> I didn't say "new instructions", but "new test": In a new test I don't think
> it is appropriate to record expectations (here: all of the l suffixes above)
> that are actually expected to not be there, but appear just because of
> brokenness. Since you touch the respective disassembler patterns
> anyway I don't really understand why you didn't make the bogus suffixes
> go away in one go. These instructions usefully have suffixes only in

I am fixing a different issue.  Please feel free to submit a separate
patch to address this particular issue.

> 64-bit mode, and earlier you've told me you don't want suffixes on newer
> insns when they're not needed.
>
> Jan
>
>



-- 
H.J.
More information about the Binutils mailing list