[FFmpeg-devel] [PATCH] avcodec/vp9: add vp9_idct_idct_4x4_add_ssse3

Ronald S. Bultje rsbultje at gmail.com
Tue Oct 29 12:19:11 CET 2013


Hi,

On Tue, Oct 29, 2013 at 6:36 AM, Ronald S. Bultje <rsbultje at gmail.com>wrote:

> Nice work overall. Some suggestions for testing:
>

Before I forget, I'm totally fine committing this as-is and working on XMM
possibilities later (after all SIMD has had a first iteration). Some basic
review points though:

> +pw_11585x2: times 4 dw 23170
> +pw_6270x2:  times 4 dw 12540
> +pw_15137x2: times 4 dw 30274
> +pw_t2_coef: dw -15137,  6270, -15137,  6270
> +pw_t3_coef: dw   6270, 15137,   6270, 15137
> +pd_round:   times 2 dd 1<<13
> +pw_2048:    times 4 dw 2048

This should probably just be pw_23170, pw_12540, pw_30274, pw_m15137_6270,
pw_6270_15137 and pd_8192. It makes it more likely that we can reuse the
constants when they are useful in another context (perhaps something that
has nothing to do with vp9's idct). Also feel free to make them xmm-sized
already, so times 8 dw or times 4 dd.

Ronald


More information about the ffmpeg-devel mailing list