<div dir="ltr"><div class="gmail_extra"> <div class="gmail_quote">On Tue, Jan 10, 2017 at 4:21 PM, Mark Thompson <<a href="mailto:sw@jkqxz.net" target="_blank">sw@jkqxz.net</a>> wrote: <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">On 10/01/17 22:02, Sean V Kelley wrote: > From: "Xiang, Haihao" <<a href="mailto:haihao.xiang@intel.com">haihao.xiang@intel.com</a>> > > Currently only one temporal layer is supported > > Signed-off-by: Xiang, Haihao <<a href="mailto:haihao.xiang@intel.com">haihao.xiang@intel.com</a>> > Reviewed-by: Sean V Kelley <<a href="mailto:seanvk@posteo.de">seanvk@posteo.de</a>> > --- > src/Makefile.am | 3 + > src/gen8_encoder_vp8.c | 140 + > src/gen8_mfc.c | 8 +- > src/gen8_vme.c | 5 + > src/i965_defines.h | 10 + > src/i965_encoder.c | 2 + > src/i965_encoder_vp8.c | 6697 ++++++++++++++++++++++++++++++++++++++++++++++++ > src/i965_encoder_vp8.h | 2643 +++++++++++++++++++ > 8 files changed, 9507 insertions(+), 1 deletion(-) I had a go with this on Kaby Lake. In general, big win - looks like it can be under half the bitrate at comparable quality (though it was pretty terrible before...). However, the rate control seems to do odd things at low bitrate relative to the frame size? I can get GPU hangs and wildly varying output bitrate with it, though it seems ok at high bitrate. </blockquote><div> </div><div>That's a concern. Please report the If it really is a GPU hang, I need the error report for the DRM card0 log.</div><div> </div><div>cat /sys/class/drm/card0/error</div><div class="gmail_extra" style="font-size:12.800000190734863px"> </div><div class="gmail_extra" style="font-size:12.800000190734863px">Please rerun and capture the DRM (i915) card0 error log.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"> I had a look around the rate control and found two minor issues in the RC configuration, though I don't think either of them are relevant to my problem (see below). I can try to make a reproducer if this is not already known? </blockquote><div>Please do attempt to reproduce. That's why I've put the patches out here to test.</div><div> </div><div>Thanks,</div><div> </div><div>Sean</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"> Thanks, - Mark > ... > + > +static void > +i965_encoder_vp8_get_misc_parameters(VADriverContextP ctx, > + struct encode_state *encode_state, > + struct intel_encoder_context *encoder_context) > +{ > + struct i965_encoder_vp8_context *vp8_context = encoder_context->vme_context; > + > + if (vp8_context->internal_rate_mode == I965_BRC_CQP) { > + vp8_context->init_vbv_buffer_fullness_in_bit = 0; > + vp8_context->vbv_buffer_size_in_bit = 0; > + vp8_context->target_bit_rate = 0; > + vp8_context->max_bit_rate = 0; > + vp8_context->min_bit_rate = 0; > + vp8_context->brc_need_reset = 0; > + } else { > + vp8_context->gop_size = encoder_context->brc.gop_size; > + > + if (encoder_context->brc.need_reset) { > + vp8_context->framerate = encoder_context->brc.framerate[0]; > + vp8_context->vbv_buffer_size_in_bit = encoder_context->brc.hrd_buffer_size; > + vp8_context->init_vbv_buffer_fullness_in_bit = encoder_context->brc.hrd_initial_buffer_fullness; > + vp8_context->max_bit_rate = encoder_context->brc.bits_per_second[0]; // currently only one layer is supported > + vp8_context->brc_need_reset = (vp8_context->brc_initted && encoder_context->brc.need_reset); > + > + if (vp8_context->internal_rate_mode == I965_BRC_CBR) { > + vp8_context->min_bit_rate = vp8_context->max_bit_rate; > + vp8_context->target_bit_rate = vp8_context->max_bit_rate; > + } else { > + assert(vp8_context->internal_rate_mode == I965_BRC_VBR); > + vp8_context->min_bit_rate = vp8_context->max_bit_rate * (2 * encoder_context->brc.target_percentage[0] - 100) / 100; If target percentage is < 50 then (2 * encoder_context->brc.target_percentage[0] - 100) is negative. Since it's unsigned, you end up with a garbage number in min_bit_rate. </blockquote><div> </div><div>That's a concern, also we may need to reconcile this with our handling for VP9 encode.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"> > + vp8_context->target_bit_rate = vp8_context->max_bit_rate * encoder_context->brc.target_percentage[0] / 100; > + } > + } > + } > + > + if (encoder_context->quality_level == ENCODER_LOW_QUALITY) > + vp8_context->hme_16x_supported = 0; > +} > + > ... > + > +static void > +i965_encoder_vp8_vme_brc_init_reset_set_curbe(VADriverContextP ctx, > + struct encode_state *encode_state, > + struct intel_encoder_context *encoder_context, > + struct i965_gpe_context *gpe_context) > +{ > + struct i965_encoder_vp8_context *vp8_context = encoder_context->vme_context; > + VAEncPictureParameterBufferVP8 *pic_param = (VAEncPictureParameterBufferVP8 *)encode_state->pic_param_ext->buffer; > + struct vp8_brc_init_reset_curbe_data *pcmd = i965_gpe_context_map_curbe(gpe_context); > + double input_bits_per_frame, bps_ratio; > + > + memset(pcmd, 0, sizeof(*pcmd)); > + > + pcmd->dw0.profile_level_max_frame = vp8_context->frame_width * vp8_context->frame_height; > + pcmd->dw1.init_buf_full_in_bits = vp8_context->init_vbv_buffer_fullness_in_bit; > + pcmd->dw2.buf_size_in_bits = vp8_context->vbv_buffer_size_in_bit; > + pcmd->dw3.average_bitrate = ALIGN(vp8_context->target_bit_rate, VP8_BRC_KBPS) / VP8_BRC_KBPS * VP8_BRC_KBPS; > + pcmd->dw4.max_bitrate = ALIGN(vp8_context->max_bit_rate, VP8_BRC_KBPS) / VP8_BRC_KBPS * VP8_BRC_KBPS; VP8_BRC_KBPS is 1000 which is not a power of two, so the ALIGN macro isn't doing anything sensible here. </blockquote><div> </div><div>Agree...</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"> > + pcmd->dw6.frame_rate_m = vp8_context->framerate.num; > + pcmd->dw7.frame_rate_d = vp8_context->framerate.den; > + pcmd->dw8.brc_flag = 0; > + pcmd->dw8.gop_minus1 = vp8_context->gop_size - 1; > + > + if (vp8_context->internal_rate_mode == I965_BRC_CBR) { > + pcmd->dw4.max_bitrate = pcmd->dw3.average_bitrate; > + > + pcmd->dw8.brc_flag = pcmd->dw8.brc_flag | BRC_KERNEL_CBR; > + } else if (vp8_context->internal_rate_mode == I965_BRC_VBR) { > + if (pcmd->dw4.max_bitrate < pcmd->dw3.average_bitrate) { > + pcmd->dw4.max_bitrate = 2 * pcmd->dw3.average_bitrate; > + } > + > + pcmd->dw8.brc_flag = pcmd->dw8.brc_flag | BRC_KERNEL_VBR; > + } > + > + input_bits_per_frame = > + ((double)(pcmd->dw4.max_bitrate) * (double)(pcmd->dw7.frame_rate_d) / > + (double)(pcmd->dw6.frame_rate_m)); > + > + if (pcmd->dw2.buf_size_in_bits < (unsigned int)input_bits_per_frame * 4) { > + pcmd->dw2.buf_size_in_bits = (unsigned int)input_bits_per_frame * 4; > + } > + > + if (pcmd->dw1.init_buf_full_in_bits == 0) { > + pcmd->dw1.init_buf_full_in_bits = 7 * pcmd->dw2.buf_size_in_bits / 8; > + } > + > + if (pcmd->dw1.init_buf_full_in_bits < (unsigned int)(input_bits_per_frame * 2)) { > + pcmd->dw1.init_buf_full_in_bits = (unsigned int)(input_bits_per_frame * 2); > + } > + > + if (pcmd->dw1.init_buf_full_in_bits > pcmd->dw2.buf_size_in_bits) { > + pcmd->dw1.init_buf_full_in_bits = pcmd->dw2.buf_size_in_bits; > + } > + > + bps_ratio = input_bits_per_frame / ((double)(pcmd->dw2.buf_size_in_bits) / 30); > + bps_ratio = (bps_ratio < 0.1) ? 0.1 : (bps_ratio > 3.5) ? 3.5 : bps_ratio; > + > + pcmd->dw9.frame_width_in_bytes = vp8_context->frame_width; > + pcmd->dw10.frame_height_in_bytes = vp8_context->frame_height; > + pcmd->dw10.avbr_accuracy = 30; > + pcmd->dw11.avbr_convergence = 150; > + pcmd->dw11.min_qp = pic_param->clamp_qindex_low; > + pcmd->dw12.max_qp = pic_param->clamp_qindex_high; > + pcmd->dw12.level_qp = 60; > + > + // DW13 default 100 > + pcmd->dw13.max_section_pct = 100; > + pcmd->dw13.under_shoot_cbr_pct = 115; > + > + // DW14 default 100 > + pcmd->dw14.min_section_pct = 100; > + pcmd->dw14.vbr_bias_pct = 100; > + pcmd->dw15.instant_rate_threshold_0_for_p = 30; > + pcmd->dw15.instant_rate_threshold_1_for_p = 50; > + pcmd->dw15.instant_rate_threshold_2_for_p = 70; > + pcmd->dw15.instant_rate_threshold_3_for_p = 120; > + > + pcmd->dw17.instant_rate_threshold_0_for_i = 30; > + pcmd->dw17.instant_rate_threshold_1_for_i = 50; > + pcmd->dw17.instant_rate_threshold_2_for_i = 90; > + pcmd->dw17.instant_rate_threshold_3_for_i = 115; > + pcmd->dw18.deviation_threshold_0_for_p = (unsigned int)(-50 * pow(0.9, bps_ratio)); > + pcmd->dw18.deviation_threshold_1_for_p = (unsigned int)(-50 * pow(0.66, bps_ratio)); > + pcmd->dw18.deviation_threshold_2_for_p = (unsigned int)(-50 * pow(0.46, bps_ratio)); > + pcmd->dw18.deviation_threshold_3_for_p = (unsigned int)(-50 * pow(0.3, bps_ratio)); > + pcmd->dw19.deviation_threshold_4_for_p = (unsigned int)(50 * pow(0.3, bps_ratio)); > + pcmd->dw19.deviation_threshold_5_for_p = (unsigned int)(50 * pow(0.46, bps_ratio)); > + pcmd->dw19.deviation_threshold_6_for_p = (unsigned int)(50 * pow(0.7, bps_ratio)); > + pcmd->dw19.deviation_threshold_7_for_p = (unsigned int)(50 * pow(0.9, bps_ratio)); > + pcmd->dw20.deviation_threshold_0_for_vbr = (unsigned int)(-50 * pow(0.9, bps_ratio)); > + pcmd->dw20.deviation_threshold_1_for_vbr = (unsigned int)(-50 * pow(0.7, bps_ratio)); > + pcmd->dw20.deviation_threshold_2_for_vbr = (unsigned int)(-50 * pow(0.5, bps_ratio)); > + pcmd->dw20.deviation_threshold_3_for_vbr = (unsigned int)(-50 * pow(0.3, bps_ratio)); > + pcmd->dw21.deviation_threshold_4_for_vbr = (unsigned int)(100 * pow(0.4, bps_ratio)); > + pcmd->dw21.deviation_threshold_5_for_vbr = (unsigned int)(100 * pow(0.5, bps_ratio)); > + pcmd->dw21.deviation_threshold_6_for_vbr = (unsigned int)(100 * pow(0.75, bps_ratio)); > + pcmd->dw21.deviation_threshold_7_for_vbr = (unsigned int)(100 * pow(0.9, bps_ratio)); > + pcmd->dw22.deviation_threshold_0_for_i = (unsigned int)(-50 * pow(0.8, bps_ratio)); > + pcmd->dw22.deviation_threshold_1_for_i = (unsigned int)(-50 * pow(0.6, bps_ratio)); > + pcmd->dw22.deviation_threshold_2_for_i = (unsigned int)(-50 * pow(0.34, bps_ratio)); > + pcmd->dw22.deviation_threshold_3_for_i = (unsigned int)(-50 * pow(0.2, bps_ratio)); > + pcmd->dw23.deviation_threshold_4_for_i = (unsigned int)(50 * pow(0.2, bps_ratio)); > + pcmd->dw23.deviation_threshold_5_for_i = (unsigned int)(50 * pow(0.4, bps_ratio)); > + pcmd->dw23.deviation_threshold_6_for_i = (unsigned int)(50 * pow(0.66, bps_ratio)); > + pcmd->dw23.deviation_threshold_7_for_i = (unsigned int)(50 * pow(0.9, bps_ratio)); > + > + // Default: 1 > + pcmd->dw24.num_t_levels = 1; > + > + if (!vp8_context->brc_initted) { > + vp8_context->brc_init_current_target_buf_full_in_bits = pcmd->dw1.init_buf_full_in_bits; > + } > + > + vp8_context->brc_init_reset_buf_size_in_bits = pcmd->dw2.buf_size_in_bits; > + vp8_context->brc_init_reset_input_bits_per_frame = input_bits_per_frame; > + > + pcmd->dw26.history_buffer_bti = VP8_BTI_BRC_INIT_RESET_HISTORY; > + pcmd->dw27.distortion_buffer_bti = VP8_BTI_BRC_INIT_RESET_DISTORTION; > + > + i965_gpe_context_unmap_curbe(gpe_context); > +} > + > ... _______________________________________________ Libva mailing list <a href="mailto:Libva@lists.freedesktop.org">Libva@lists.freedesktop.org</a> <a href="https://lists.freedesktop.org/mailman/listinfo/libva" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/libva</a> </blockquote></div> <div> </div>-- <div class="gmail_signature">Sean V. Kelley <<a href="mailto:sean.v.kelley@intel.com" target="_blank">sean.v.kelley@intel.com</a>> Open Source Technology Center / SSG Intel Corp. </div> </div></div>