Re: Shark suggests SSE on code that uses .... SSE
Re: Shark suggests SSE on code that uses .... SSE
- Subject: Re: Shark suggests SSE on code that uses .... SSE
- From: Benson Margulies <email@hidden>
- Date: Tue, 14 Apr 2009 17:11:46 -0400
The wording, if you ask, me, could be a hair clearer, but I guess I'm
splitting a hair.
On Tue, Apr 14, 2009 at 4:43 PM, Rick Altherr <email@hidden> wrote:
>
> On Apr 14, 2009, at 1:08 PM, Benson Margulies wrote:
>
>> 5.1%   0x100410dec                   movq   (%rdx,
>> %rcx, 8), %rdi  !    Loop start[6],
>> SSE Â Â bt_ape_decoder.cpp:167
>> 0.0%   0x100410df0                   movq   (%rdi),
>> %rax          bt_ape_decoder.cpp:168
>> 0.0%   0x100410df3                   movl   %r14d,
>> %r8d           bt_ape_decoder.cpp:168
>>     0x100410df6                   movl   %r12d,
>> ìx           bt_ape_decoder.cpp:168
>> 5.0%   0x100410df9                   movl   %r13d,
>> íx           bt_ape_decoder.cpp:168
>>     0x100410dfc                   movq
>> -168(%rbp), %rsi            bt_ape_decoder.cpp:168
>> 0.0%   0x100410e03                   call   +32(%rax)
>> Â Â Â Â Â Â Â Â Â Â Â bt_ape_decoder.cpp:168
>> 5.1%   0x100410e06                   movss   %xmm0,
>> -68(%rbp) Â Â Â Â Â Â Â Â Â Â Â bt_ape_decoder.cpp:168
>> 0.0%   0x100410e0b                   movss
>> Â -68(%rbp), %xmm0 Â Â Â Â Â Â Â Â Â Â Â bt_ape_decoder.cpp:168
>> 6.9% Â Â 0x100410e10 Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â addss
>> Â -56(%rbp), %xmm0 Â Â Â Â Â Â Â Â Â Â Â bt_ape_decoder.cpp:168
>> 10.2%  0x100410e15                   movss   %xmm0,
>> -56(%rbp) Â Â Â Â Â Â Â Â Â Â Â bt_ape_decoder.cpp:168
>> 6.1%   0x100410e1a                   movl   ëx,
>> ìx           stl_vector.h:166
>> 0.0%   0x100410e1c                   movq   +8(%r15),
>> %rax         stl_vector.h:166
>> 0.0%   0x100410e20                   movq   (%rax),
>> %rdx          stl_vector.h:343
>> 0.0%   0x100410e23                   movq   +8(%rax),
>> %rax         stl_vector.h:361
>> 6.0%   0x100410e27                   incq   %rbx
>> Â Â Â Â Â Â Â bt_ape_decoder.cpp:403
>> 0.0%   0x100410e2a                   subq   %rdx,
>> %rax           bt_ape_matrix.h:166
>> 0.0%   0x100410e2d                   sarq   $3, %rax
>> Â Â Â Â Â Â Â Â Â Â Â bt_ape_matrix.h:166
>> 0.0%   0x100410e31                   cmpq   %rax,
>> %rcx           bt_ape_matrix.h:166
>> 6.1% Â Â 0x100410e34 Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â jb
>> 0x0000000100410dec
>> Â <BT_ApeDecoder::forwardViterbi(BT_ApeDecodeState&)+478> Â Â Â Â Â Loop
>> end[6] Â bt_ape_matrix.h:166
>> _______________________________________________
>> Do not post admin requests to the list. They will be ignored.
>> Xcode-users mailing list    (email@hidden)
>> Help/Unsubscribe/Update your Subscription:
>>
>> This email sent to email@hidden
>
>
> SSE is being used for scalar operations here. Â Rather than use x87, the
> compiler defaults to using SSE scalar operations for floating point. Â Shark
> is suggesting you use vector SSE operations.
>
> --
> Rick Altherr
> Performance Tools
> Development Technologies
> email@hidden
>
>
>
>
>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Xcode-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden