Re: Request for code review - (ZEE patch : Redundant Zero extension elimination)

[Sriraman Tallam <tmsriram@google.com>]

Sorry, it is the other way around.

Total number of zero-extension instructions before  :  5814
Total number of zero-extension instructions after   :  1456

Thanks for pointing it.

On Wed, Sep 23, 2009 at 4:10 PM, Ramana Radhakrishnan
<ramana.r@gmail.com> wrote:
>>
>> GCC bootstrap :
>>
>> Total number of zero-extension instructions before  : 1456
>> Total number of zero-extension instructions after    :  5814
>> No impact on boot-strap time.
>
>
> You sure you have these numbers the right way around ? Shouldn't the
> number of zero-extension instructions after the patch be less than the
> number of zero-extension instructions before or is this a regression
> ?
>
> Thanks,
> Ramana
>
>>
>>
>> I have attached the latest patch :
>>
>>
>> On Sun, Aug 9, 2009 at 2:15 PM, Richard Guenther
>> <richard.guenther@gmail.com> wrote:
>>> On Sat, Aug 8, 2009 at 11:59 PM, Sriraman Tallam<tmsriram@google.com> wrote:
>>>> Hi,
>>>>
>>>>    Here is a patch to eliminate redundant zero-extension instructions
>>>> on x86_64.
>>>>
>>>> Tested: Ran the gcc regresssion testsuite on x86_64-linux and verified
>>>> that the results are the same with/without this patch.
>>>
>>> The patch misses testcases.  Why does zee run after register allocation?
>>> Your examples suggest that it will free hard registers so doing it before
>>> regalloc looks odd.
>>>
>>> What is the compile-time impact of your patch on say, gcc bootstrap?
>>> How many percent of instructions are removed as useless zero-extensions
>>> during gcc bootstrap?  How much do CSiBE numbers improve?
>>>
>>> Thanks,
>>> Richard.
>>>
>>>>
>>>> Problem Description :
>>>> ---------------------------------
>>>>
>>>> This pass is intended to be applicable only to targets that implicitly
>>>> zero-extend 64-bit registers after writing to their lower 32-bit half.
>>>> For instance, x86_64 zero-extends the upper bits of a register
>>>> implicitly whenever an instruction writes to its lower 32-bit half.
>>>> For example, the instruction *add edi,eax* also zero-extends the upper
>>>> 32-bits of rax after doing the addition.  These zero extensions come
>>>> for free and GCC does not always exploit this well.  That is, it has
>>>> been observed that there are plenty of cases where GCC explicitly
>>>> zero-extends registers for x86_64 that are actually useless because
>>>> these registers were already implicitly zero-extended in a prior
>>>> instruction.  This pass tries to eliminate such useless zero extension
>>>> instructions.
>>>>
>>>> Motivating Example I :
>>>> ----------------------------------
>>>> For this program :
>>>> **********************************************
>>>> bad_code.c
>>>>
>>>> int mask[1000];
>>>>
>>>> int foo(unsigned x)
>>>> {
>>>>  if (x < 10)
>>>>    x = x * 45;
>>>>  else
>>>>    x = x * 78;
>>>>  return mask[x];
>>>> }
>>>> **********************************************
>>>>
>>>> $ gcc -O2 bad_code.c
>>>>  ........
>>>>  400315:       b8 4e 00 00 00            mov    $0x4e,%eax
>>>>  40031a:       0f af f8                        imul   %eax,%edi
>>>>  40031d:       89 ff                             mov    %edi,%edi
>>>> ---> Useless zero extend.
>>>>  40031f:       8b 04 bd 60 19 40 00    mov    0x401960(,%rdi,4),%eax
>>>>  400326:       c3                                 retq
>>>>  ......
>>>>  400330:       ba 2d 00 00 00          mov    $0x2d,%edx
>>>>  400335:       0f af fa                      imul   %edx,%edi
>>>>  400338:       89 ff                           mov    %edi,%edi  --->
>>>> Useless zero extend.
>>>>  40033a:       8b 04 bd 60 19 40 00    mov    0x401960(,%rdi,4),%eax
>>>>  400341:       c3                      retq
>>>>
>>>> $ gcc -O2 -fzee bad_code.c
>>>>  ......
>>>>  400315:       6b ff 4e                imul   $0x4e,%edi,%edi
>>>>  400318:       8b 04 bd 40 19 40 00    mov    0x401940(,%rdi,4),%eax
>>>>  40031f:       c3                      retq
>>>>  400320:       6b ff 2d                imul   $0x2d,%edi,%edi
>>>>  400323:       8b 04 bd 40 19 40 00    mov    0x401940(,%rdi,4),%eax
>>>>  40032a:       c3                      retq
>>>>
>>>>
>>>>
>>>> Thanks,
>>>>
>>>> Sriraman M Tallam.
>>>> Google, Inc.
>>>> tmsriram@google.com
>>>>
>>>
>>
>