mirror of
https://code.videolan.org/videolan/dav1d
synced 2026-06-11 04:03:05 +00:00
Niklas Haas
3a2a874994
There are a number of benefits tied to the upstream / third-party checkasm version, including: - Improved long-term maintainability, code reuse with other projects, etc. - Vastly improved overall performance / runtime for benchmarking, due primarily to the ability to scale the runtime of each test to that test's complexity. - Much more robust statistical analysis of benchmarking results; including robust outlier rejection, an estimation of the histogram, and the ability to report the variance / stddev in addition to the (trimmed) mean. - Interactive HTML and JSON output formats in addition to CSV/TSV. - More readable and user-friendly output across the board, especially for failures and data dumps (e.g. also showing errors inside padding bytes). - Better cross-platform support, including dynamic fallback of timer implementations on ARM platforms, a better RISC-V harness, and more. There are multiple approaches to how we can solve the problem of integrating this third party checkasm into dav1d, but I think the hybrid approach of loading it as an external dependency, falling back to a meson wrap file, provides the best overall compromise. This avoids the messiness of git e.g. git submodules, while still allowing us to pin individual tags.
307 lines
12 KiB
C
307 lines
12 KiB
C
/*
|
|
* Copyright © 2018, VideoLAN and dav1d authors
|
|
* Copyright © 2018, Two Orioles, LLC
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions are met:
|
|
*
|
|
* 1. Redistributions of source code must retain the above copyright notice, this
|
|
* list of conditions and the following disclaimer.
|
|
*
|
|
* 2. Redistributions in binary form must reproduce the above copyright notice,
|
|
* this list of conditions and the following disclaimer in the documentation
|
|
* and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
|
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
|
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
|
|
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
|
|
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
|
|
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
|
|
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#include "tests/checkasm/internal.h"
|
|
#include "src/ipred.h"
|
|
#include "src/levels.h"
|
|
|
|
#include <stdio.h>
|
|
|
|
static const char *const intra_pred_mode_names[N_IMPL_INTRA_PRED_MODES] = {
|
|
[DC_PRED] = "dc",
|
|
[DC_128_PRED] = "dc_128",
|
|
[TOP_DC_PRED] = "dc_top",
|
|
[LEFT_DC_PRED] = "dc_left",
|
|
[HOR_PRED] = "h",
|
|
[VERT_PRED] = "v",
|
|
[PAETH_PRED] = "paeth",
|
|
[SMOOTH_PRED] = "smooth",
|
|
[SMOOTH_V_PRED] = "smooth_v",
|
|
[SMOOTH_H_PRED] = "smooth_h",
|
|
[Z1_PRED] = "z1",
|
|
[Z2_PRED] = "z2",
|
|
[Z3_PRED] = "z3",
|
|
[FILTER_PRED] = "filter"
|
|
};
|
|
|
|
static const char *const cfl_ac_names[3] = { "420", "422", "444" };
|
|
|
|
static const char *const cfl_pred_mode_names[DC_128_PRED + 1] = {
|
|
[DC_PRED] = "cfl",
|
|
[DC_128_PRED] = "cfl_128",
|
|
[TOP_DC_PRED] = "cfl_top",
|
|
[LEFT_DC_PRED] = "cfl_left",
|
|
};
|
|
|
|
static const uint8_t z_angles[27] = {
|
|
3, 6, 9,
|
|
14, 17, 20, 23, 26, 29, 32,
|
|
36, 39, 42, 45, 48, 51, 54,
|
|
58, 61, 64, 67, 70, 73, 76,
|
|
81, 84, 87
|
|
};
|
|
|
|
/* Generate max_width/max_height values that covers all edge cases */
|
|
static int gen_z2_max_wh(const int sz) {
|
|
const int n = rnd();
|
|
if (n & (1 << 17)) /* edge block */
|
|
return (n & (sz - 1)) + 1;
|
|
if (n & (1 << 16)) /* max size, exceeds uint16_t */
|
|
return 65536;
|
|
return (n & 65535) + 1;
|
|
}
|
|
|
|
static void check_intra_pred(Dav1dIntraPredDSPContext *const c) {
|
|
PIXEL_RECT(c_dst, 64, 64);
|
|
PIXEL_RECT(a_dst, 64, 64);
|
|
ALIGN_STK_64(pixel, topleft_buf, 257,);
|
|
pixel *const topleft = topleft_buf + 128;
|
|
|
|
declare_func(void, pixel *dst, ptrdiff_t stride, const pixel *topleft,
|
|
int width, int height, int angle, int max_width, int max_height
|
|
HIGHBD_DECL_SUFFIX);
|
|
|
|
for (int mode = 0; mode < N_IMPL_INTRA_PRED_MODES; mode++) {
|
|
int bpc_min = BITDEPTH, bpc_max = BITDEPTH;
|
|
if (mode == FILTER_PRED && BITDEPTH == 16) {
|
|
bpc_min = 10;
|
|
bpc_max = 12;
|
|
}
|
|
for (int bpc = bpc_min; bpc <= bpc_max; bpc += 2)
|
|
for (int w = 4; w <= (mode == FILTER_PRED ? 32 : 64); w <<= 1)
|
|
if (check_func(c->intra_pred[mode], "intra_pred_%s_w%d_%dbpc",
|
|
intra_pred_mode_names[mode], w, bpc))
|
|
{
|
|
for (int h = imax(w / 4, 4); h <= imin(w * 4,
|
|
(mode == FILTER_PRED ? 32 : 64)); h <<= 1)
|
|
{
|
|
const ptrdiff_t stride = c_dst_stride;
|
|
int nb_iters = (mode >= Z1_PRED && mode <= Z3_PRED) ? 5 : 1;
|
|
|
|
for (int iter = 0; iter < nb_iters; iter++) {
|
|
int a = 0, maxw = 0, maxh = 0;
|
|
if (mode >= Z1_PRED && mode <= Z3_PRED) { /* angle */
|
|
a = (90 * (mode - Z1_PRED) + z_angles[rnd() % 27]) |
|
|
(rnd() & 0x600);
|
|
if (mode == Z2_PRED) {
|
|
maxw = gen_z2_max_wh(w);
|
|
maxh = gen_z2_max_wh(h);
|
|
}
|
|
} else if (mode == FILTER_PRED) /* filter_idx */
|
|
a = (rnd() % 5) | (rnd() & ~511);
|
|
|
|
int bitdepth_max;
|
|
if (bpc == 16)
|
|
bitdepth_max = rnd() & 1 ? 0x3ff : 0xfff;
|
|
else
|
|
bitdepth_max = (1 << bpc) - 1;
|
|
|
|
for (int i = -h * 2; i <= w * 2; i++)
|
|
topleft[i] = rnd() & bitdepth_max;
|
|
|
|
CLEAR_PIXEL_RECT(c_dst);
|
|
CLEAR_PIXEL_RECT(a_dst);
|
|
call_ref(c_dst, stride, topleft, w, h, a, maxw, maxh
|
|
HIGHBD_TAIL_SUFFIX);
|
|
call_new(a_dst, stride, topleft, w, h, a, maxw, maxh
|
|
HIGHBD_TAIL_SUFFIX);
|
|
if (checkasm_check_pixel_padded(c_dst, stride,
|
|
a_dst, stride,
|
|
w, h, "dst"))
|
|
{
|
|
if (mode == Z1_PRED || mode == Z3_PRED)
|
|
fprintf(stderr, "angle = %d (0x%03x)\n",
|
|
a & 0x1ff, a & 0x600);
|
|
else if (mode == Z2_PRED)
|
|
fprintf(stderr, "angle = %d (0x%03x), "
|
|
"max_width = %d, max_height = %d\n",
|
|
a & 0x1ff, a & 0x600, maxw, maxh);
|
|
else if (mode == FILTER_PRED)
|
|
fprintf(stderr, "filter_idx = %d\n", a & 0x1ff);
|
|
break;
|
|
}
|
|
|
|
bench_new(a_dst, stride, topleft, w, h, a, 128, 128
|
|
HIGHBD_TAIL_SUFFIX);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
report("intra_pred");
|
|
}
|
|
|
|
static void check_cfl_ac(Dav1dIntraPredDSPContext *const c) {
|
|
ALIGN_STK_64(int16_t, c_dst, 32 * 32,);
|
|
ALIGN_STK_64(int16_t, a_dst, 32 * 32,);
|
|
ALIGN_STK_64(pixel, luma, 32 * 32,);
|
|
|
|
declare_func(void, int16_t *ac, const pixel *y, ptrdiff_t stride,
|
|
int w_pad, int h_pad, int cw, int ch);
|
|
|
|
for (int layout = 1; layout <= DAV1D_PIXEL_LAYOUT_I444; layout++) {
|
|
const int ss_ver = layout == DAV1D_PIXEL_LAYOUT_I420;
|
|
const int ss_hor = layout != DAV1D_PIXEL_LAYOUT_I444;
|
|
const int h_step = 2 >> ss_hor, v_step = 2 >> ss_ver;
|
|
for (int w = 4; w <= (32 >> ss_hor); w <<= 1)
|
|
if (check_func(c->cfl_ac[layout - 1], "cfl_ac_%s_w%d_%dbpc",
|
|
cfl_ac_names[layout - 1], w, BITDEPTH))
|
|
{
|
|
for (int h = imax(w / 4, 4);
|
|
h <= imin(w * 4, (32 >> ss_ver)); h <<= 1)
|
|
{
|
|
const ptrdiff_t stride = 32 * sizeof(pixel);
|
|
for (int w_pad = imax((w >> 2) - h_step, 0);
|
|
w_pad >= 0; w_pad -= h_step)
|
|
{
|
|
for (int h_pad = imax((h >> 2) - v_step, 0);
|
|
h_pad >= 0; h_pad -= v_step)
|
|
{
|
|
#if BITDEPTH == 16
|
|
const int bitdepth_max = rnd() & 1 ? 0x3ff : 0xfff;
|
|
#else
|
|
const int bitdepth_max = 0xff;
|
|
#endif
|
|
for (int y = 0; y < (h << ss_ver); y++)
|
|
for (int x = 0; x < (w << ss_hor); x++)
|
|
luma[y * 32 + x] = rnd() & bitdepth_max;
|
|
|
|
call_ref(c_dst, luma, stride, w_pad, h_pad, w, h);
|
|
call_new(a_dst, luma, stride, w_pad, h_pad, w, h);
|
|
checkasm_check(int16_t, c_dst, w * sizeof(*c_dst),
|
|
a_dst, w * sizeof(*a_dst),
|
|
w, h, "dst");
|
|
}
|
|
}
|
|
|
|
bench_new(a_dst, luma, stride, 0, 0, w, h);
|
|
}
|
|
}
|
|
}
|
|
report("cfl_ac");
|
|
}
|
|
|
|
static void check_cfl_pred(Dav1dIntraPredDSPContext *const c) {
|
|
PIXEL_RECT(c_dst, 32, 32);
|
|
PIXEL_RECT(a_dst, 32, 32);
|
|
ALIGN_STK_64(int16_t, ac, 32 * 32,);
|
|
ALIGN_STK_64(pixel, topleft_buf, 257,);
|
|
pixel *const topleft = topleft_buf + 128;
|
|
|
|
declare_func(void, pixel *dst, ptrdiff_t stride, const pixel *topleft,
|
|
int width, int height, const int16_t *ac, int alpha
|
|
HIGHBD_DECL_SUFFIX);
|
|
|
|
for (int mode = 0; mode <= DC_128_PRED; mode += 1 + 2 * !mode)
|
|
for (int w = 4; w <= 32; w <<= 1)
|
|
if (check_func(c->cfl_pred[mode], "cfl_pred_%s_w%d_%dbpc",
|
|
cfl_pred_mode_names[mode], w, BITDEPTH))
|
|
{
|
|
for (int h = imax(w / 4, 4); h <= imin(w * 4, 32); h <<= 1)
|
|
{
|
|
#if BITDEPTH == 16
|
|
const int bitdepth_max = rnd() & 1 ? 0x3ff : 0xfff;
|
|
#else
|
|
const int bitdepth_max = 0xff;
|
|
#endif
|
|
|
|
int alpha = ((rnd() & 15) + 1) * (1 - (rnd() & 2));
|
|
|
|
for (int i = -h * 2; i <= w * 2; i++)
|
|
topleft[i] = rnd() & bitdepth_max;
|
|
|
|
int luma_avg = w * h >> 1;
|
|
for (int i = 0; i < w * h; i++)
|
|
luma_avg += ac[i] = rnd() & (bitdepth_max << 3);
|
|
luma_avg /= w * h;
|
|
for (int i = 0; i < w * h; i++)
|
|
ac[i] -= luma_avg;
|
|
|
|
CLEAR_PIXEL_RECT(c_dst);
|
|
CLEAR_PIXEL_RECT(a_dst);
|
|
|
|
call_ref(c_dst, c_dst_stride, topleft, w, h, ac, alpha
|
|
HIGHBD_TAIL_SUFFIX);
|
|
call_new(a_dst, a_dst_stride, topleft, w, h, ac, alpha
|
|
HIGHBD_TAIL_SUFFIX);
|
|
checkasm_check_pixel_padded(c_dst, c_dst_stride, a_dst, a_dst_stride,
|
|
w, h, "dst");
|
|
|
|
bench_new(a_dst, a_dst_stride, topleft, w, h, ac, alpha
|
|
HIGHBD_TAIL_SUFFIX);
|
|
}
|
|
}
|
|
report("cfl_pred");
|
|
}
|
|
|
|
static void check_pal_pred(Dav1dIntraPredDSPContext *const c) {
|
|
PIXEL_RECT(c_dst, 64, 64);
|
|
PIXEL_RECT(a_dst, 64, 64);
|
|
ALIGN_STK_64(uint8_t, idx, 32 * 64,);
|
|
ALIGN_STK_16(pixel, pal, 8,);
|
|
|
|
declare_func(void, pixel *dst, ptrdiff_t stride, const pixel *pal,
|
|
const uint8_t *idx, int w, int h);
|
|
|
|
for (int w = 4; w <= 64; w <<= 1)
|
|
if (check_func(c->pal_pred, "pal_pred_w%d_%dbpc", w, BITDEPTH))
|
|
for (int h = imax(w / 4, 4); h <= imin(w * 4, 64); h <<= 1)
|
|
{
|
|
#if BITDEPTH == 16
|
|
const int bitdepth_max = rnd() & 1 ? 0x3ff : 0xfff;
|
|
#else
|
|
const int bitdepth_max = 0xff;
|
|
#endif
|
|
|
|
for (int i = 0; i < 8; i++)
|
|
pal[i] = rnd() & bitdepth_max;
|
|
|
|
for (int i = 0; i < w * h / 2; i++)
|
|
idx[i] = rnd() & 0x77;
|
|
|
|
CLEAR_PIXEL_RECT(c_dst);
|
|
CLEAR_PIXEL_RECT(a_dst);
|
|
|
|
call_ref(c_dst, c_dst_stride, pal, idx, w, h);
|
|
call_new(a_dst, a_dst_stride, pal, idx, w, h);
|
|
checkasm_check_pixel_padded(c_dst, c_dst_stride,
|
|
a_dst, a_dst_stride, w, h, "dst");
|
|
|
|
bench_new(a_dst, a_dst_stride, pal, idx, w, h);
|
|
}
|
|
report("pal_pred");
|
|
}
|
|
|
|
void bitfn(checkasm_check_ipred)(void) {
|
|
Dav1dIntraPredDSPContext c;
|
|
bitfn(dav1d_intra_pred_dsp_init)(&c);
|
|
|
|
check_intra_pred(&c);
|
|
check_cfl_ac(&c);
|
|
check_cfl_pred(&c);
|
|
check_pal_pred(&c);
|
|
}
|