mirror of
https://git.ffmpeg.org/ffmpeg.git
synced 2024-12-22 23:33:07 +00:00
aad0e26f93
The stereo_interpolate functions add h_step to the values h BUF_SIZE times. Within the stereo_interpolate C functions, the values h (h0-h3, h00-h13) are declared as local float variables, but the compiler is free to keep them in a register with extra precision. If the accumulation is rounded to 32 bit float precision after each step, the less significant bits of h_step end up ignored and the sum can deviate, affecting the end result more than the currently set EPS. By clearing the log2(BUF_SIZE) lower bits of h_step, we make sure that the accumulation shouldn't differ significantly, regardless of any extra precision in the accmulating register/variable. This fixes the aacpsdsp checkasm test when built with clang for mingw/x86_32. Signed-off-by: Martin Storsjö <martin@martin.st>
262 lines
9.0 KiB
C
262 lines
9.0 KiB
C
/*
|
|
* This file is part of FFmpeg.
|
|
*
|
|
* FFmpeg is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* FFmpeg is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License along
|
|
* with FFmpeg; if not, write to the Free Software Foundation, Inc.,
|
|
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
*/
|
|
|
|
#include "libavcodec/aacpsdsp.h"
|
|
#include "libavutil/intfloat.h"
|
|
|
|
#include "checkasm.h"
|
|
|
|
#define N 32
|
|
#define STRIDE 128
|
|
#define BUF_SIZE (N * STRIDE)
|
|
|
|
#define randomize(buf, len) do { \
|
|
int i; \
|
|
for (i = 0; i < len; i++) { \
|
|
const INTFLOAT f = (INTFLOAT)rnd() / UINT_MAX; \
|
|
(buf)[i] = f; \
|
|
} \
|
|
} while (0)
|
|
|
|
#define EPS 0.005
|
|
|
|
static void clear_less_significant_bits(INTFLOAT *buf, int len, int bits)
|
|
{
|
|
int i;
|
|
for (i = 0; i < len; i++) {
|
|
union av_intfloat32 u = { .f = buf[i] };
|
|
u.i &= (0xffffffff << bits);
|
|
buf[i] = u.f;
|
|
}
|
|
}
|
|
|
|
static void test_add_squares(void)
|
|
{
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst0, [BUF_SIZE]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst1, [BUF_SIZE]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, src, [BUF_SIZE], [2]);
|
|
|
|
declare_func(void, INTFLOAT *dst,
|
|
const INTFLOAT (*src)[2], int n);
|
|
|
|
randomize((INTFLOAT *)src, BUF_SIZE * 2);
|
|
randomize(dst0, BUF_SIZE);
|
|
memcpy(dst1, dst0, BUF_SIZE * sizeof(INTFLOAT));
|
|
call_ref(dst0, src, BUF_SIZE);
|
|
call_new(dst1, src, BUF_SIZE);
|
|
if (!float_near_abs_eps_array(dst0, dst1, EPS, BUF_SIZE))
|
|
fail();
|
|
bench_new(dst1, src, BUF_SIZE);
|
|
}
|
|
|
|
static void test_mul_pair_single(void)
|
|
{
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst0, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst1, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, src0, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, src1, [BUF_SIZE]);
|
|
|
|
declare_func(void, INTFLOAT (*dst)[2],
|
|
INTFLOAT (*src0)[2], INTFLOAT *src1, int n);
|
|
|
|
randomize((INTFLOAT *)src0, BUF_SIZE * 2);
|
|
randomize(src1, BUF_SIZE);
|
|
call_ref(dst0, src0, src1, BUF_SIZE);
|
|
call_new(dst1, src0, src1, BUF_SIZE);
|
|
if (!float_near_abs_eps_array((float *)dst0, (float *)dst1, EPS, BUF_SIZE * 2))
|
|
fail();
|
|
bench_new(dst1, src0, src1, BUF_SIZE);
|
|
}
|
|
|
|
static void test_hybrid_analysis(void)
|
|
{
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst0, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, dst1, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, in, [13], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, filter, [N], [8][2]);
|
|
|
|
declare_func(void, INTFLOAT (*out)[2], INTFLOAT (*in)[2],
|
|
const INTFLOAT (*filter)[8][2],
|
|
ptrdiff_t stride, int n);
|
|
|
|
randomize((INTFLOAT *)in, 13 * 2);
|
|
randomize((INTFLOAT *)filter, N * 8 * 2);
|
|
|
|
randomize((INTFLOAT *)dst0, BUF_SIZE * 2);
|
|
memcpy(dst1, dst0, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
|
|
call_ref(dst0, in, filter, STRIDE, N);
|
|
call_new(dst1, in, filter, STRIDE, N);
|
|
|
|
if (!float_near_abs_eps_array((float *)dst0, (float *)dst1, EPS, BUF_SIZE * 2))
|
|
fail();
|
|
bench_new(dst1, in, filter, STRIDE, N);
|
|
}
|
|
|
|
static void test_hybrid_analysis_ileave(void)
|
|
{
|
|
LOCAL_ALIGNED_16(INTFLOAT, in, [2], [38][64]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, out0, [91], [32][2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, out1, [91], [32][2]);
|
|
|
|
declare_func(void, INTFLOAT (*out)[32][2], INTFLOAT L[2][38][64],
|
|
int i, int len);
|
|
|
|
randomize((INTFLOAT *)out0, 91 * 32 * 2);
|
|
randomize((INTFLOAT *)in, 2 * 38 * 64);
|
|
memcpy(out1, out0, 91 * 32 * 2 * sizeof(INTFLOAT));
|
|
|
|
/* len is hardcoded to 32 as that's the only value used in
|
|
libavcodec. asm functions are likely to be optimized
|
|
hardcoding this value in their loops and could fail with
|
|
anything else.
|
|
i is hardcoded to the two values currently used by the
|
|
aac decoder because the arm neon implementation is
|
|
micro-optimized for them and will fail for almost every
|
|
other value. */
|
|
call_ref(out0, in, 3, 32);
|
|
call_new(out1, in, 3, 32);
|
|
|
|
/* the function just moves data around, so memcmp is enough */
|
|
if (memcmp(out0, out1, 91 * 32 * 2 * sizeof(INTFLOAT)))
|
|
fail();
|
|
|
|
call_ref(out0, in, 5, 32);
|
|
call_new(out1, in, 5, 32);
|
|
|
|
if (memcmp(out0, out1, 91 * 32 * 2 * sizeof(INTFLOAT)))
|
|
fail();
|
|
|
|
bench_new(out1, in, 3, 32);
|
|
}
|
|
|
|
static void test_hybrid_synthesis_deint(void)
|
|
{
|
|
LOCAL_ALIGNED_16(INTFLOAT, out0, [2], [38][64]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, out1, [2], [38][64]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, in, [91], [32][2]);
|
|
|
|
declare_func(void, INTFLOAT out[2][38][64], INTFLOAT (*in)[32][2],
|
|
int i, int len);
|
|
|
|
randomize((INTFLOAT *)in, 91 * 32 * 2);
|
|
randomize((INTFLOAT *)out0, 2 * 38 * 64);
|
|
memcpy(out1, out0, 2 * 38 * 64 * sizeof(INTFLOAT));
|
|
|
|
/* len is hardcoded to 32 as that's the only value used in
|
|
libavcodec. asm functions are likely to be optimized
|
|
hardcoding this value in their loops and could fail with
|
|
anything else.
|
|
i is hardcoded to the two values currently used by the
|
|
aac decoder because the arm neon implementation is
|
|
micro-optimized for them and will fail for almost every
|
|
other value. */
|
|
call_ref(out0, in, 3, 32);
|
|
call_new(out1, in, 3, 32);
|
|
|
|
/* the function just moves data around, so memcmp is enough */
|
|
if (memcmp(out0, out1, 2 * 38 * 64 * sizeof(INTFLOAT)))
|
|
fail();
|
|
|
|
call_ref(out0, in, 5, 32);
|
|
call_new(out1, in, 5, 32);
|
|
|
|
if (memcmp(out0, out1, 2 * 38 * 64 * sizeof(INTFLOAT)))
|
|
fail();
|
|
|
|
bench_new(out1, in, 3, 32);
|
|
}
|
|
|
|
static void test_stereo_interpolate(PSDSPContext *psdsp)
|
|
{
|
|
int i;
|
|
LOCAL_ALIGNED_16(INTFLOAT, l, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, r, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, l0, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, r0, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, l1, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, r1, [BUF_SIZE], [2]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, h, [2], [4]);
|
|
LOCAL_ALIGNED_16(INTFLOAT, h_step, [2], [4]);
|
|
|
|
declare_func(void, INTFLOAT (*l)[2], INTFLOAT (*r)[2],
|
|
INTFLOAT h[2][4], INTFLOAT h_step[2][4], int len);
|
|
|
|
randomize((INTFLOAT *)l, BUF_SIZE * 2);
|
|
randomize((INTFLOAT *)r, BUF_SIZE * 2);
|
|
|
|
for (i = 0; i < 2; i++) {
|
|
if (check_func(psdsp->stereo_interpolate[i], "ps_stereo_interpolate%s", i ? "_ipdopd" : "")) {
|
|
memcpy(l0, l, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
memcpy(l1, l, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
memcpy(r0, r, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
memcpy(r1, r, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
|
|
randomize((INTFLOAT *)h, 2 * 4);
|
|
randomize((INTFLOAT *)h_step, 2 * 4);
|
|
// Clear the least significant 14 bits of h_step, to avoid
|
|
// divergence when accumulating h_step BUF_SIZE times into
|
|
// a float variable which may or may not have extra intermediate
|
|
// precision. Therefore clear roughly log2(BUF_SIZE) less
|
|
// significant bits, to get the same result regardless of any
|
|
// extra precision in the accumulator.
|
|
clear_less_significant_bits((INTFLOAT *)h_step, 2 * 4, 14);
|
|
|
|
call_ref(l0, r0, h, h_step, BUF_SIZE);
|
|
call_new(l1, r1, h, h_step, BUF_SIZE);
|
|
if (!float_near_abs_eps_array((float *)l0, (float *)l1, EPS, BUF_SIZE * 2) ||
|
|
!float_near_abs_eps_array((float *)r0, (float *)r1, EPS, BUF_SIZE * 2))
|
|
fail();
|
|
|
|
memcpy(l1, l, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
memcpy(r1, r, BUF_SIZE * 2 * sizeof(INTFLOAT));
|
|
bench_new(l1, r1, h, h_step, BUF_SIZE);
|
|
}
|
|
}
|
|
}
|
|
|
|
void checkasm_check_aacpsdsp(void)
|
|
{
|
|
PSDSPContext psdsp;
|
|
|
|
ff_psdsp_init(&psdsp);
|
|
|
|
if (check_func(psdsp.add_squares, "ps_add_squares"))
|
|
test_add_squares();
|
|
report("add_squares");
|
|
|
|
if (check_func(psdsp.mul_pair_single, "ps_mul_pair_single"))
|
|
test_mul_pair_single();
|
|
report("mul_pair_single");
|
|
|
|
if (check_func(psdsp.hybrid_analysis, "ps_hybrid_analysis"))
|
|
test_hybrid_analysis();
|
|
report("hybrid_analysis");
|
|
|
|
if (check_func(psdsp.hybrid_analysis_ileave, "ps_hybrid_analysis_ileave"))
|
|
test_hybrid_analysis_ileave();
|
|
report("hybrid_analysis_ileave");
|
|
|
|
if (check_func(psdsp.hybrid_synthesis_deint, "ps_hybrid_synthesis_deint"))
|
|
test_hybrid_synthesis_deint();
|
|
report("hybrid_synthesis_deint");
|
|
|
|
test_stereo_interpolate(&psdsp);
|
|
report("stereo_interpolate");
|
|
}
|