On Thu, Jun 22, 2017 at 12:53:52PM -0300, James Almer wrote:
> 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().
Done in the following version of the patch.
Thanks.
--
Matthieu B.
>From 749b74d2146cd7ac4dd8e71bcf2a789b901590d8 Mon Sep 17 00:00:00 2001
From: Matthieu Bouron <matthieu.bou...@gmail.com>
Date: Fri, 9 Jun 2017 09:34:12 +0000
Subject: [PATCH 1/3] checkasm: add sbrdsp tests
---
tests/checkasm/Makefile | 3 +-
tests/checkasm/checkasm.c | 1 +
tests/checkasm/checkasm.h | 1 +
tests/checkasm/sbrdsp.c | 298 ++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 302 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..038318e021
--- /dev/null
+++ b/tests/checkasm/sbrdsp.c
@@ -0,0 +1,298 @@
+/*
+ * 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);
+ 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(const SBRDSPContext *sbrdsp)
+{
+ 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, j;
+
+ 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 < 4; i++) {
+ if (check_func(sbrdsp->hf_apply_noise[i], "hf_apply_noise_%d", i)) {
+ for (j = 0; j < 2; j++) {
+ memcpy(dst0, ref, 128 * 2 * sizeof(INTFLOAT));
+ memcpy(dst1, ref, 128 * 2 * sizeof(INTFLOAT));
+ call_ref(dst0, s_m, q_filt, noise, j, 128);
+ call_new(dst1, s_m, q_filt, noise, j, 128);
+ if (!float_near_abs_eps_array((INTFLOAT *)dst0, (INTFLOAT *)dst1, EPS, 128 * 2))
+ fail();
+ bench_new(dst1, s_m, q_filt, noise, j, 128);
+ }
+ }
+ }
+}
+
+void checkasm_check_sbrdsp(void)
+{
+ 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");
+
+ test_hf_apply_noise(&sbrdsp);
+ report("hf_apply_noise");
+}
--
2.13.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
