On 9/17/2015 5:43 AM, Henrik Gramner wrote:
> On Thu, Sep 17, 2015 at 12:02 AM, James Almer <jamr...@gmail.com> wrote:
>> +#define randomize_buffers() \
>> + do { \
>> + uint32_t r; \
>> + int i, j; \
>> + for (i = 0; i < BUF_SIZE; i += 4) { \
>> + for (j = 0; j < channels; j++) { \
>> + r = rnd() & (1 << (bits - 2)) - 1; \
>> + AV_WN32A(ref_src[j] + i, r); \
>> + AV_WN32A(new_src[j] + i, r); \
>> + } \
>> + } \
>> + } while (0)
>
> r could be declared and initialized at the same time.
Done.
>
>> + LOCAL_ALIGNED_16(uint8_t, ref_buf, [BUF_SIZE*MAX_CHANNELS]);
>> + LOCAL_ALIGNED_16(uint8_t, new_buf, [BUF_SIZE*MAX_CHANNELS]);
>> + uint8_t *ref_src[] = { &ref_buf[BUF_SIZE*0], &ref_buf[BUF_SIZE*1],
>> &ref_buf[BUF_SIZE*2], &ref_buf[BUF_SIZE*3],
>> + &ref_buf[BUF_SIZE*4], &ref_buf[BUF_SIZE*5],
>> &ref_buf[BUF_SIZE*6], &ref_buf[BUF_SIZE*7] };
>> + uint8_t *new_src[] = { &new_buf[BUF_SIZE*0], &new_buf[BUF_SIZE*1],
>> &new_buf[BUF_SIZE*2], &new_buf[BUF_SIZE*3],
>> + &new_buf[BUF_SIZE*4], &new_buf[BUF_SIZE*5],
>> &new_buf[BUF_SIZE*6], &new_buf[BUF_SIZE*7] };
>
> Those could potentially be int32_t*, then you could avoid the typecast
> later on. Not sure it that would actually makes things cleaner though,
> so not a big deal.
It looks better with the typecast inside check_decorrelate() IMO, so I'll
leave it as is.
>
> Otherwise seems OK.
Pushed, thanks.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
