[FFmpeg-devel] libavcodec/blockdsp : add AVX version

[James Almer]

On 10/3/2017 4:47 PM, Martin Vignali wrote:
> Hello,
> 
> 
>> I used GCC 7.2. clear_blocks_mmx is slower than c for me as well, but
>> not the rest.
>> Your compiler seems to have done a much better job than mine. Is it
>> Clang? Does it somehow have vectorization enabled perhaps? Because
>> that's not supposed to happen.
>>
>>
> Yes it's Clang 8.1
> 
> I put the clear_blocks_c function, in a file and run
> clang -S -O1 test_asm_gen.c
> 
> the asm result is
>     .section    __TEXT,__text,regular,pure_instructions
>     .macosx_version_min 10, 12
>     .globl    _clear_blocks_c
>     .p2align    4, 0x90
> _clear_blocks_c:                        ## @clear_blocks_c
>     .cfi_startproc
> ## BB#0:
>     pushq    %rbp
> Ltmp0:
>     .cfi_def_cfa_offset 16
> Ltmp1:
>     .cfi_offset %rbp, -16
>     movq    %rsp, %rbp
> Ltmp2:
>     .cfi_def_cfa_register %rbp
>     movl    $768, %esi              ## imm = 0x300
>     callq    ___bzero
>     popq    %rbp
>     retq
>     .cfi_endproc
> 
> 
> .subsections_via_symbols
> 
> Seems like an optimized function is call for clear_blocks_c

Yeah, the c version uses memset. Guess clang's implementation is good.

Patch pushed. Thanks.