Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-23 Thread Luca Barbato
On 9/22/11 8:27 PM, Justin Ruggles wrote: ping. I tested on my laptop measuring the whole encoding and apparently the baseline is faster (2.66 vs 2.75 to encode 30m of audio) at least on macosx+gcc-4.5.2 I'm wondering what is upsetting it... ___ l

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-23 Thread Justin Ruggles
On 09/22/2011 02:27 PM, Justin Ruggles wrote: > On 09/08/2011 12:56 PM, Justin Ruggles wrote: > >> On 09/08/2011 11:49 AM, Måns Rullgård wrote: >> >>> Justin Ruggles writes: >>> On 09/08/2011 07:46 AM, Luca Barbato wrote: > On 9/8/11 4:58 AM, Justin Ruggles wrote: >> From: Mich

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-22 Thread Justin Ruggles
On 09/08/2011 12:56 PM, Justin Ruggles wrote: > On 09/08/2011 11:49 AM, Måns Rullgård wrote: > >> Justin Ruggles writes: >> >>> On 09/08/2011 07:46 AM, Luca Barbato wrote: >>> On 9/8/11 4:58 AM, Justin Ruggles wrote: > From: Michael Niedermayer What is the performance gain in

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-08 Thread Justin Ruggles
On 09/08/2011 11:49 AM, Måns Rullgård wrote: > Justin Ruggles writes: > >> On 09/08/2011 07:46 AM, Luca Barbato wrote: >> >>> On 9/8/11 4:58 AM, Justin Ruggles wrote: From: Michael Niedermayer >>> >>> What is the performance gain in hand unrolling this loop? >> >> On Athlon64 it goes from 6

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-08 Thread Måns Rullgård
Justin Ruggles writes: > On 09/08/2011 07:46 AM, Luca Barbato wrote: > >> On 9/8/11 4:58 AM, Justin Ruggles wrote: >>> From: Michael Niedermayer >> >> What is the performance gain in hand unrolling this loop? > > On Athlon64 it goes from 60 cycles to 54 cycles. Which compiler? What about sligh

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-08 Thread Justin Ruggles
On 09/08/2011 07:46 AM, Luca Barbato wrote: > On 9/8/11 4:58 AM, Justin Ruggles wrote: >> From: Michael Niedermayer > > What is the performance gain in hand unrolling this loop? On Athlon64 it goes from 60 cycles to 54 cycles. -Justin ___ libav-devel

Re: [libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-08 Thread Luca Barbato
On 9/8/11 4:58 AM, Justin Ruggles wrote: From: Michael Niedermayer What is the performance gain in hand unrolling this loop? lu ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel

[libav-devel] [PATCH 5/6] adpcmenc:Optimize adpcm_ima_qt_compress_sample()

2011-09-07 Thread Justin Ruggles
From: Michael Niedermayer Signed-off-by: Michael Niedermayer --- libavcodec/adpcmenc.c | 33 ++--- 1 files changed, 18 insertions(+), 15 deletions(-) diff --git a/libavcodec/adpcmenc.c b/libavcodec/adpcmenc.c index 51334e3..56587da 100644 --- a/libavcodec/adpcmenc