va_dec_av1: Add max_frame_width/height_minus_1

[nowrep]

AMD VCN needs this to correctly decode av1-1-b8-22-svc-L2T1 and av1-1-b8-22-svc-L2T2.

[XinfengZhang]

not understand , this structure is frame level, each frame already has current width and height. from my understanding , max_frame_width_minus_1 max_frame_height_minus_1. should in seq level to determine the maximum resolution. why we need it for one frame ?

[nowrep]

Yes, these values are in sequence header, but there is only VADecPictureParameterBufferAV1 for both sequence and frame parameters.

should in seq level to determine the maximum resolution. why we need it for one frame ?

It's not maximum resolution according to seq level, it's maximum resolution used in the sequence and this can differ from frame resolution with multiple spatial layers.

For example av1-1-b8-22-svc-L2T1.ivf: max resolution in sequence header is 1280x720 and first frame resolution is 640x360.

[XinfengZhang]

For example av1-1-b8-22-svc-L2T1.ivf: max resolution in sequence header is 1280x720 and first frame resolution is 640x360.

yes, there could be multiple resolution in one sequence, but I am wondering why HW need this, because you set parameters per picture/frame. so each frame has it's own resolution. it should be enough.
why backend driver need know the maximum resolution?