--- /dev/null
+#include "x264_speed_control.h"
+
+#include <dlfcn.h>
+#include <math.h>
+#include <stdio.h>
+#include <x264.h>
+#include <algorithm>
+#include <chrono>
+#include <cmath>
+#include <ratio>
+#include <type_traits>
+
+#include "flags.h"
+#include "metrics.h"
+
+using namespace std;
+using namespace std::chrono;
+
+#define SC_PRESETS 23
+
+X264SpeedControl::X264SpeedControl(x264_t *x264, float f_speed, int i_buffer_size, float f_buffer_init)
+ : dyn(load_x264_for_bit_depth(global_flags.x264_bit_depth)),
+ x264(x264), f_speed(f_speed)
+{
+ x264_param_t param;
+ dyn.x264_encoder_parameters(x264, ¶m);
+
+ float fps = (float)param.i_fps_num / param.i_fps_den;
+ uspf = 1e6 / fps;
+ set_buffer_size(i_buffer_size);
+ buffer_fill = buffer_size * f_buffer_init;
+ buffer_fill = max<int64_t>(buffer_fill, uspf);
+ buffer_fill = min(buffer_fill, buffer_size);
+ timestamp = steady_clock::now();
+ preset = -1;
+ cplx_num = 3e3; //FIXME estimate initial complexity
+ cplx_den = .1;
+ stat.min_buffer = buffer_size;
+ stat.max_buffer = 0;
+ stat.avg_preset = 0.0;
+ stat.den = 0;
+
+ metric_x264_speedcontrol_buffer_available_seconds = buffer_fill * 1e-6;
+ metric_x264_speedcontrol_buffer_size_seconds = buffer_size * 1e-6;
+ metric_x264_speedcontrol_preset_used_frames.init_uniform(SC_PRESETS);
+ global_metrics.add("x264_speedcontrol_preset_used_frames", &metric_x264_speedcontrol_preset_used_frames);
+ global_metrics.add("x264_speedcontrol_buffer_available_seconds", &metric_x264_speedcontrol_buffer_available_seconds, Metrics::TYPE_GAUGE);
+ global_metrics.add("x264_speedcontrol_buffer_size_seconds", &metric_x264_speedcontrol_buffer_size_seconds, Metrics::TYPE_GAUGE);
+ global_metrics.add("x264_speedcontrol_idle_frames", &metric_x264_speedcontrol_idle_frames);
+ global_metrics.add("x264_speedcontrol_late_frames", &metric_x264_speedcontrol_late_frames);
+}
+
+X264SpeedControl::~X264SpeedControl()
+{
+ fprintf(stderr, "speedcontrol: avg preset=%.3f buffer min=%.3f max=%.3f\n",
+ stat.avg_preset / stat.den,
+ (float)stat.min_buffer / buffer_size,
+ (float)stat.max_buffer / buffer_size );
+ // x264_log( x264, X264_LOG_INFO, "speedcontrol: avg cplx=%.5f\n", cplx_num / cplx_den );
+ if (dyn.handle) {
+ dlclose(dyn.handle);
+ }
+}
+
+typedef struct
+{
+ float time; // relative encoding time, compared to the other presets
+ int subme;
+ int me;
+ int refs;
+ int mix;
+ int trellis;
+ int partitions;
+ int direct;
+ int merange;
+} sc_preset_t;
+
+// The actual presets, including the equivalent commandline options. Note that
+// all presets are benchmarked with --weightp 1 --mbtree --rc-lookahead 20
+// --b-adapt 1 --bframes 3 on top of the given settings (equivalent settings to
+// the "faster" preset). Timings and SSIM measurements were done on a four cores
+// of a 6-core Coffee Lake i5 2.8 GHz on the first 1000 frames of “Elephants
+// Dream” in 1080p. See experiments/measure-x264.pl for a way to reproduce.
+//
+// Note that the two first and the two last are also used for extrapolation
+// should the desired time be outside the range. Thus, it is disadvantageous if
+// they are chosen so that the timings are too close to each other.
+static const sc_preset_t presets[SC_PRESETS] = {
+#define I4 X264_ANALYSE_I4x4
+#define I8 X264_ANALYSE_I8x8
+#define P4 X264_ANALYSE_PSUB8x8
+#define P8 X264_ANALYSE_PSUB16x16
+#define B8 X264_ANALYSE_BSUB16x16
+
+ // Preset 0: 17.386db, --preset superfast
+ { .time= 1.000, .subme=1, .me=X264_ME_DIA, .refs=1, .mix=0, .trellis=0, .partitions=I8|I4, .direct=1, .merange=16 },
+
+ // Preset 1: 17.919db, --preset superfast --subme 2
+ { .time= 1.707, .subme=2, .me=X264_ME_DIA, .refs=1, .mix=0, .trellis=0, .partitions=I8|I4, .direct=1, .merange=16 },
+
+ // Preset 2: 18.051db, --preset veryfast
+ { .time= 1.832, .subme=2, .me=X264_ME_HEX, .refs=1, .mix=0, .trellis=0, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 3: 18.422db, --preset veryfast --subme 3
+ { .time= 1.853, .subme=3, .me=X264_ME_HEX, .refs=1, .mix=0, .trellis=0, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 4: 18.514db, --preset veryfast --subme 3 --ref 2
+ { .time= 1.925, .subme=3, .me=X264_ME_HEX, .refs=2, .mix=0, .trellis=0, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 5: 18.564db, --preset veryfast --subme 4 --ref 2
+ { .time= 2.111, .subme=4, .me=X264_ME_HEX, .refs=2, .mix=0, .trellis=0, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 6: 18.411db, --preset faster
+ { .time= 2.240, .subme=4, .me=X264_ME_HEX, .refs=2, .mix=0, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 7: 18.429db, --preset faster --mixed-refs
+ { .time= 2.414, .subme=4, .me=X264_ME_HEX, .refs=2, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 8: 18.454db, --preset faster --mixed-refs --subme 5
+ { .time= 2.888, .subme=5, .me=X264_ME_HEX, .refs=2, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 9: 18.528db, --preset fast
+ { .time= 3.570, .subme=6, .me=X264_ME_HEX, .refs=2, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 10: 18.762db, --preset fast --subme 7
+ { .time= 3.698, .subme=7, .me=X264_ME_HEX, .refs=2, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 11: 18.819db, --preset medium
+ { .time= 4.174, .subme=7, .me=X264_ME_HEX, .refs=3, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 12: 18.889db, --preset medium --subme 8
+ { .time= 5.155, .subme=8, .me=X264_ME_HEX, .refs=3, .mix=1, .trellis=1, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 13: 19.127db, --preset medium --subme 8 --trellis 2
+ { .time= 7.237, .subme=8, .me=X264_ME_HEX, .refs=3, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=1, .merange=16 },
+
+ // Preset 14: 19.118db, --preset medium --subme 8 --trellis 2 --direct auto
+ { .time= 7.240, .subme=8, .me=X264_ME_HEX, .refs=3, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 15: 19.172db, --preset slow
+ { .time= 7.910, .subme=8, .me=X264_ME_HEX, .refs=5, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 16: 19.208db, --preset slow --subme 9
+ { .time= 8.091, .subme=9, .me=X264_ME_HEX, .refs=5, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 17: 19.216db, --preset slow --subme 9 --me umh
+ { .time= 9.539, .subme=9, .me=X264_ME_UMH, .refs=5, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 18: 19.253db, --preset slow --subme 9 --me umh --ref 6
+ { .time=10.521, .subme=9, .me=X264_ME_UMH, .refs=6, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 19: 19.275db, --preset slow --subme 9 --me umh --ref 7
+ { .time=11.461, .subme=9, .me=X264_ME_UMH, .refs=7, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8, .direct=3, .merange=16 },
+
+ // Preset 20: 19.314db, --preset slower
+ { .time=13.145, .subme=9, .me=X264_ME_UMH, .refs=8, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8|P4, .direct=3, .merange=16 },
+
+ // Preset 21: 19.407db, --preset slower --subme 10
+ { .time=16.386, .subme=10, .me=X264_ME_UMH, .refs=8, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8|P4, .direct=3, .merange=16 },
+
+ // Preset 22: 19.483db, --preset veryslow
+ { .time=26.861, .subme=10, .me=X264_ME_UMH, .refs=16, .mix=1, .trellis=2, .partitions=I8|I4|P8|B8|P4, .direct=3, .merange=24 },
+
+#undef I4
+#undef I8
+#undef P4
+#undef P8
+#undef B8
+};
+
+void X264SpeedControl::before_frame(float new_buffer_fill, int new_buffer_size, float new_uspf)
+{
+ if (new_uspf > 0.0) {
+ uspf = new_uspf;
+ }
+ if (new_buffer_size) {
+ set_buffer_size(new_buffer_size);
+ }
+ buffer_fill = buffer_size * new_buffer_fill;
+ metric_x264_speedcontrol_buffer_available_seconds = buffer_fill * 1e-6;
+
+ steady_clock::time_point t;
+
+ // update buffer state after encoding and outputting the previous frame(s)
+ if (first) {
+ t = timestamp = steady_clock::now();
+ first = false;
+ } else {
+ t = steady_clock::now();
+ }
+
+ auto delta_t = t - timestamp;
+ timestamp = t;
+
+ // update the time predictor
+ if (preset >= 0) {
+ int cpu_time = duration_cast<microseconds>(cpu_time_last_frame).count();
+ cplx_num *= cplx_decay;
+ cplx_den *= cplx_decay;
+ cplx_num += cpu_time / presets[preset].time;
+ ++cplx_den;
+
+ stat.avg_preset += preset;
+ ++stat.den;
+ }
+
+ stat.min_buffer = min(buffer_fill, stat.min_buffer);
+ stat.max_buffer = max(buffer_fill, stat.max_buffer);
+
+ if (buffer_fill >= buffer_size) { // oops, cpu was idle
+ // not really an error, but we'll warn for debugging purposes
+ static int64_t idle_t = 0;
+ static steady_clock::time_point print_interval;
+ static bool first = false;
+ idle_t += buffer_fill - buffer_size;
+ if (first || duration<double>(t - print_interval).count() > 0.1) {
+ //fprintf(stderr, "speedcontrol idle (%.6f sec)\n", idle_t/1e6);
+ print_interval = t;
+ idle_t = 0;
+ first = false;
+ }
+ buffer_fill = buffer_size;
+ metric_x264_speedcontrol_buffer_available_seconds = buffer_fill * 1e-6;
+ ++metric_x264_speedcontrol_idle_frames;
+ } else if (buffer_fill <= 0) { // oops, we're late
+ // fprintf(stderr, "speedcontrol underflow (%.6f sec)\n", buffer_fill/1e6);
+ ++metric_x264_speedcontrol_late_frames;
+ }
+
+ {
+ // Pick the preset that should return the buffer to 3/4-full within a time
+ // specified by compensation_period.
+ //
+ // NOTE: This doesn't actually do that, at least assuming the same target is
+ // chosen for every frame; exactly what it does is unclear to me. It seems
+ // to consistently undershoot a bit, so it needs to be saved by the second
+ // predictor below. However, fixing the formula seems to yield somewhat less
+ // stable results in practice; in particular, once the buffer is half-full
+ // or so, it would give us a negative target. Perhaps increasing
+ // compensation_period would be a good idea, but initial (very brief) tests
+ // did not yield good results.
+ float target = uspf / f_speed
+ * (buffer_fill + compensation_period)
+ / (buffer_size*3/4 + compensation_period);
+ float cplx = cplx_num / cplx_den;
+ float set, t0, t1;
+ float filled = (float) buffer_fill / buffer_size;
+ int i;
+ t0 = presets[0].time * cplx;
+ for (i = 1; ; i++) {
+ t1 = presets[i].time * cplx;
+ if (t1 >= target || i == SC_PRESETS - 1)
+ break;
+ t0 = t1;
+ }
+ // exponential interpolation between states
+ set = i-1 + (log(target) - log(t0)) / (log(t1) - log(t0));
+ set = max<float>(set, -5);
+ set = min<float>(set, (SC_PRESETS-1) + 5);
+ // Even if our time estimations in the SC_PRESETS array are off
+ // this will push us towards our target fullness
+ float s1 = set;
+ set += (40 * (filled-0.75));
+ float s2 = (40 * (filled-0.75));
+ set = min<float>(max<float>(set, 0), SC_PRESETS - 1);
+ apply_preset(dither_preset(set));
+
+ if (global_flags.x264_speedcontrol_verbose) {
+ static float cpu, wall, tgt, den;
+ const float decay = 1-1/100.;
+ cpu = cpu*decay + duration_cast<microseconds>(cpu_time_last_frame).count();
+ wall = wall*decay + duration_cast<microseconds>(delta_t).count();
+ tgt = tgt*decay + target;
+ den = den*decay + 1;
+ fprintf(stderr, "speed: %.2f+%.2f %d[%.5f] (t/c/w: %6.0f/%6.0f/%6.0f = %.4f) fps=%.2f\r",
+ s1, s2, preset, (float)buffer_fill / buffer_size,
+ tgt/den, cpu/den, wall/den, cpu/wall, 1e6*den/wall );
+ }
+ }
+
+}
+
+void X264SpeedControl::after_frame()
+{
+ cpu_time_last_frame = steady_clock::now() - timestamp;
+}
+
+void X264SpeedControl::set_buffer_size(int new_buffer_size)
+{
+ new_buffer_size = max(3, new_buffer_size);
+ buffer_size = new_buffer_size * uspf;
+ cplx_decay = 1 - 1./new_buffer_size;
+ compensation_period = buffer_size/4;
+ metric_x264_speedcontrol_buffer_size_seconds = buffer_size * 1e-6;
+}
+
+int X264SpeedControl::dither_preset(float f)
+{
+ int i = f;
+ if (f < 0) {
+ i--;
+ }
+ dither += f - i;
+ if (dither >= 1.0) {
+ dither--;
+ i++;
+ }
+ return i;
+}
+
+void X264SpeedControl::apply_preset(int new_preset)
+{
+ new_preset = max(new_preset, 0);
+ new_preset = min(new_preset, SC_PRESETS - 1);
+
+ const sc_preset_t *s = &presets[new_preset];
+ x264_param_t p;
+ dyn.x264_encoder_parameters(x264, &p);
+
+ p.i_frame_reference = s->refs;
+ p.analyse.inter = s->partitions;
+ p.analyse.i_subpel_refine = s->subme;
+ p.analyse.i_me_method = s->me;
+ p.analyse.i_trellis = s->trellis;
+ p.analyse.b_mixed_references = s->mix;
+ p.analyse.i_direct_mv_pred = s->direct;
+ p.analyse.i_me_range = s->merange;
+ if (override_func) {
+ override_func(&p);
+ }
+ dyn.x264_encoder_reconfig(x264, &p);
+ preset = new_preset;
+
+ metric_x264_speedcontrol_preset_used_frames.count_event(new_preset);
+}
projects / nageru / blobdiff