On 6/22/2017 9:56 AM, Matthieu Bouron wrote:
> ---
> 
> The following patchset applies on top of Clément's aacpsdsp patchset.
> 
> ---
>  tests/checkasm/Makefile   |   3 +-
>  tests/checkasm/checkasm.c |   1 +
>  tests/checkasm/checkasm.h |   1 +
>  tests/checkasm/sbrdsp.c   | 297 
> ++++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 301 insertions(+), 1 deletion(-)
>  create mode 100644 tests/checkasm/sbrdsp.c
> 
> diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> index 638e811931..60e80ab738 100644
> --- a/tests/checkasm/Makefile
> +++ b/tests/checkasm/Makefile
> @@ -13,7 +13,8 @@ AVCODECOBJS-$(CONFIG_VP8DSP)            += vp8dsp.o
>  AVCODECOBJS-$(CONFIG_VIDEODSP)          += videodsp.o
>  
>  # decoders/encoders
> -AVCODECOBJS-$(CONFIG_AAC_DECODER)       += aacpsdsp.o
> +AVCODECOBJS-$(CONFIG_AAC_DECODER)       += aacpsdsp.o \
> +                                           sbrdsp.o
>  AVCODECOBJS-$(CONFIG_ALAC_DECODER)      += alacdsp.o
>  AVCODECOBJS-$(CONFIG_DCA_DECODER)       += synth_filter.o
>  AVCODECOBJS-$(CONFIG_JPEG2000_DECODER)  += jpeg2000dsp.o
> diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> index e66744b162..29f201b1b3 100644
> --- a/tests/checkasm/checkasm.c
> +++ b/tests/checkasm/checkasm.c
> @@ -67,6 +67,7 @@ static const struct {
>  #if CONFIG_AVCODEC
>      #if CONFIG_AAC_DECODER
>          { "aacpsdsp", checkasm_check_aacpsdsp },
> +        { "sbrdsp",   checkasm_check_sbrdsp },
>      #endif
>      #if CONFIG_ALAC_DECODER
>          { "alacdsp", checkasm_check_alacdsp },
> diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> index dfb0ce561c..fa51e71e4b 100644
> --- a/tests/checkasm/checkasm.h
> +++ b/tests/checkasm/checkasm.h
> @@ -50,6 +50,7 @@ void checkasm_check_hevc_idct(void);
>  void checkasm_check_jpeg2000dsp(void);
>  void checkasm_check_llviddsp(void);
>  void checkasm_check_pixblockdsp(void);
> +void checkasm_check_sbrdsp(void);
>  void checkasm_check_synth_filter(void);
>  void checkasm_check_v210enc(void);
>  void checkasm_check_vp8dsp(void);
> diff --git a/tests/checkasm/sbrdsp.c b/tests/checkasm/sbrdsp.c
> new file mode 100644
> index 0000000000..8333510c6b
> --- /dev/null
> +++ b/tests/checkasm/sbrdsp.c
> @@ -0,0 +1,297 @@
> +/*
> + * 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/sbrdsp.h"
> +
> +#include "checkasm.h"
> +
> +#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 0.0001
> +
> +static void test_sum64x5(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [64 + 256]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [64 + 256]);
> +
> +    declare_func(void, INTFLOAT *z);
> +
> +    randomize((INTFLOAT *)dst0, 64 + 256);
> +    memcpy(dst1, dst0, (64 + 256) * sizeof(INTFLOAT));
> +    call_ref(dst0);
> +    call_new(dst1);
> +    if (!float_near_abs_eps_array(dst0, dst1, EPS, 64 + 256))
> +        fail();
> +    bench_new(dst1);
> +}
> +
> +static void test_sum_square(void)
> +{
> +    INTFLOAT res0;
> +    INTFLOAT res1;
> +    LOCAL_ALIGNED_16(INTFLOAT, src, [256], [2]);
> +
> +    declare_func(INTFLOAT, INTFLOAT (*x)[2], int n);
> +
> +    randomize((INTFLOAT *)src, 256 * 2);
> +    res0 = call_ref(src, 256);
> +    res1 = call_new(src, 256);

This one is failing on x86 because the second argument has garbage in
the upper 32 bits of the grp.

The solution is to either change n from int to ptrdiff_t, or zero the
upper bits of n in ff_sbr_sum_square_sse(), which can be done implicitly
in the first instruction.

> +    if (!float_near_abs_eps(res0, res1, EPS))
> +        fail();
> +    bench_new(src, 256);
> +}
> +
> +static void test_neg_odd_64(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [64]);
> +
> +    declare_func(void, INTFLOAT *x);
> +
> +    randomize((INTFLOAT *)dst0, 64);
> +    memcpy(dst1, dst0, (64) * sizeof(INTFLOAT));
> +    call_ref(dst0);
> +    call_new(dst1);
> +    if (!float_near_abs_eps_array(dst0, dst1, EPS, 64))
> +        fail();
> +    bench_new(dst1);
> +}
> +
> +static void test_qmf_pre_shuffle(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [128]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [128]);
> +
> +    declare_func(void, INTFLOAT *z);
> +
> +    randomize((INTFLOAT *)dst0, 128);
> +    memcpy(dst1, dst0, (128) * sizeof(INTFLOAT));
> +    call_ref(dst0);
> +    call_new(dst1);
> +    if (!float_near_abs_eps_array(dst0, dst1, EPS, 128))
> +        fail();
> +    bench_new(dst1);
> +}
> +
> +static void test_qmf_post_shuffle(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, src, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [32], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [32], [2]);
> +
> +    declare_func(void, INTFLOAT W[32][2], const INTFLOAT *z);
> +
> +    randomize((INTFLOAT *)src, 64);
> +    call_ref(dst0, src);
> +    call_new(dst1, src);
> +    if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, EPS, 
> 64))
> +        fail();
> +    bench_new(dst1, src);
> +}
> +
> +static void test_qmf_deint_neg(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, src, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [64]);
> +
> +    declare_func(void, INTFLOAT *v, const INTFLOAT *src);
> +
> +    randomize((INTFLOAT *)src, 64);
> +    call_ref(dst0, src);
> +    call_new(dst1, src);
> +    if (!float_near_abs_eps_array(dst0, dst1, EPS, 64))
> +        fail();
> +    bench_new(dst1, src);
> +}
> +
> +static void test_qmf_deint_bfly(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, src0, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, src1, [64]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [128]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [128]);
> +
> +    declare_func(void, INTFLOAT *v, const INTFLOAT *src0, const INTFLOAT 
> *src1);
> +
> +    memset(dst0, 0, 128 * sizeof(INTFLOAT));
> +    memset(dst1, 0, 128 * sizeof(INTFLOAT));
> +
> +    randomize((INTFLOAT *)src0, 64);
> +    randomize((INTFLOAT *)src1, 64);
> +    call_ref(dst0, src0, src1);
> +    call_new(dst1, src0, src1);
> +    if (!float_near_abs_eps_array(dst0, dst1, EPS, 128))
> +        fail();
> +    bench_new(dst1, src0, src1);
> +}
> +
> +static void test_autocorrelate(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, src, [40], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [3], [2][2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [3], [2][2]);
> +
> +    declare_func(void, const INTFLOAT x[40][2], INTFLOAT phi[3][2][2]);
> +
> +    memset(dst0, 0, 3 * 2 * 2 * sizeof(INTFLOAT));
> +    memset(dst1, 0, 3 * 2 * 2 * sizeof(INTFLOAT));
> +
> +    randomize((INTFLOAT *)src, 80);
> +    call_ref(src, dst0);
> +    call_new(src, dst1);
> +    if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, EPS, 3 
> * 2 * 2))
> +        fail();
> +    bench_new(src, dst1);
> +}
> +
> +static void test_hf_gen(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, low, [128], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, alpha0, [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, alpha1, [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [128], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [128], [2]);
> +    INTFLOAT bw = (INTFLOAT)rnd() / UINT_MAX;
> +    int i;
> +
> +    declare_func(void, INTFLOAT (*X_high)[2], const INTFLOAT (*X_low)[2],
> +                       const INTFLOAT alpha0[2], const INTFLOAT alpha1[2],
> +                       INTFLOAT bw, int start, int end);
> +
> +    randomize((INTFLOAT *)low, 128 * 2);
> +    randomize((INTFLOAT *)alpha0, 2);
> +    randomize((INTFLOAT *)alpha1, 2);
> +    for (i = 2; i < 64; i += 2) {
> +        memset(dst0, 0, 128 * 2 * sizeof(INTFLOAT));
> +        memset(dst1, 0, 128 * 2 * sizeof(INTFLOAT));
> +        call_ref(dst0, low, alpha0, alpha1, 0.0, i, 128);
> +        call_new(dst1, low, alpha0, alpha1, 0.0, i, 128);
> +        if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, 
> EPS, 128 * 2))
> +            fail();
> +        bench_new(dst1, low, alpha0, alpha1, bw, i, 128);
> +    }
> +}
> +
> +static void test_hf_g_filt(void)
> +{
> +    LOCAL_ALIGNED_16(INTFLOAT, high, [128], [40][2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, g_filt, [128]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [128], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [128], [2]);
> +
> +    declare_func(void, INTFLOAT (*Y)[2], const INTFLOAT (*X_high)[40][2],
> +                       const INTFLOAT *g_filt, int m_max, intptr_t ixh);
> +
> +    randomize((INTFLOAT *)high, 128 * 40 * 2);
> +    randomize((INTFLOAT *)g_filt, 128);
> +
> +    call_ref(dst0, high, g_filt, 128, 20);
> +    call_new(dst1, high, g_filt, 128, 20);
> +    if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, EPS, 
> 128 * 2))
> +        fail();
> +    bench_new(dst1, high, g_filt, 128, 20);
> +}
> +
> +static void test_hf_apply_noise(void)
> +{
> +    LOCAL_ALIGNED_16(AAC_FLOAT, s_m, [128]);
> +    LOCAL_ALIGNED_16(AAC_FLOAT, q_filt, [128]);
> +    LOCAL_ALIGNED_16(INTFLOAT, ref, [128], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst0, [128], [2]);
> +    LOCAL_ALIGNED_16(INTFLOAT, dst1, [128], [2]);
> +    int noise = 0x2a;
> +    int i;
> +
> +    declare_func(void, INTFLOAT (*Y)[2], const AAC_FLOAT *s_m,
> +                       const AAC_FLOAT *q_filt, int noise,
> +                       int kx, int m_max);
> +
> +    randomize((INTFLOAT *)ref, 128 * 2);
> +    randomize((INTFLOAT *)s_m, 128);
> +    randomize((INTFLOAT *)q_filt, 128);
> +    for (i = 0; i < 2; i++) {
> +        memcpy(dst0, ref, 128 * 2 * sizeof(INTFLOAT));
> +        memcpy(dst1, ref, 128 * 2 * sizeof(INTFLOAT));
> +        call_ref(dst0, s_m, q_filt, noise, i, 128);
> +        call_new(dst1, s_m, q_filt, noise, i, 128);
> +        if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, 
> EPS, 128 * 2))
> +            fail();
> +        bench_new(dst1, s_m, q_filt, noise, i, 128);
> +    }
> +}
> +
> +void checkasm_check_sbrdsp(void)
> +{
> +    int i;
> +    SBRDSPContext sbrdsp;
> +
> +    ff_sbrdsp_init(&sbrdsp);
> +
> +    if (check_func(sbrdsp.sum64x5, "sum64x5"))
> +        test_sum64x5();
> +    report("sum64x5");
> +
> +    if (check_func(sbrdsp.sum_square, "sum_square"))
> +        test_sum_square();
> +    report("sum_square");
> +
> +    if (check_func(sbrdsp.neg_odd_64, "neg_odd_64"))
> +        test_neg_odd_64();
> +    report("neg_odd_64");
> +
> +    if (check_func(sbrdsp.qmf_pre_shuffle, "qmf_pre_shuffle"))
> +        test_qmf_pre_shuffle();
> +    report("qmf_pre_shuffle");
> +
> +    if (check_func(sbrdsp.qmf_post_shuffle, "qmf_post_shuffle"))
> +        test_qmf_post_shuffle();
> +    report("qmf_post_shuffle");
> +
> +    if (check_func(sbrdsp.qmf_deint_neg, "qmf_deint_neg"))
> +        test_qmf_deint_neg();
> +    report("qmf_deint_neg");
> +
> +    if (check_func(sbrdsp.qmf_deint_bfly, "qmf_deint_bfly"))
> +        test_qmf_deint_bfly();
> +    report("qmf_deint_bfly");
> +
> +    if (check_func(sbrdsp.autocorrelate, "autocorrelate"))
> +        test_autocorrelate();
> +    report("autocorrelate");
> +
> +    if (check_func(sbrdsp.hf_gen, "hf_gen"))
> +        test_hf_gen();
> +    report("hf_gen");
> +
> +    if (check_func(sbrdsp.hf_g_filt, "hf_g_filt"))
> +        test_hf_g_filt();
> +    report("hf_g_filt");
> +
> +    for (i = 0; i < 4; i++) {

You could instead add this loop inside test_hf_apply_noise(). It would
save you a bunch of unnecessary calls to randomize().

> +        if (check_func(sbrdsp.hf_apply_noise[i], "hf_apply_noise_%d", i))
> +            test_hf_apply_noise();
> +    }
> +    report("hf_apply_noise");
> +}
> 

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to