checkasm: add tests for dcadsp

tests/checkasm/Makefile

 # libavcodec tests
 AVCODECOBJS-$(CONFIG_BSWAPDSP) += bswapdsp.o
+AVCODECOBJS-$(CONFIG_DCA_DECODER) += dcadsp.o
 AVCODECOBJS-$(CONFIG_H264PRED) += h264pred.o
 AVCODECOBJS-$(CONFIG_H264QPEL) += h264qpel.o
 AVCODECOBJS-$(CONFIG_HEVC_DECODER) += hevc_mc.o

tests/checkasm/checkasm.c

@@ -61,6 +61,9 @@ static const struct {
 #if CONFIG_BSWAPDSP
     { "bswapdsp", checkasm_check_bswapdsp },
 #endif
+#if CONFIG_DCA_DECODER
+    { "dcadsp", checkasm_check_dcadsp },
+#endif
 #if CONFIG_H264PRED
     { "h264pred", checkasm_check_h264pred },
 #endif

tests/checkasm/checkasm.h

@@ -31,6 +31,7 @@
 #include "libavutil/timer.h"
 
 void checkasm_check_bswapdsp(void);
+void checkasm_check_dcadsp(void);
 void checkasm_check_h264pred(void);
 void checkasm_check_h264qpel(void);
 void checkasm_check_hevc_mc(void);

tests/checkasm/dcadsp.c

+/*
+ * Copyright (c) 2015 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#include <math.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include "libavutil/internal.h"
+#include "libavutil/intfloat.h"
+#include "libavcodec/dca.h"
+#include "libavcodec/dcadsp.h"
+#include "libavcodec/dcadata.h"
+
+#include "checkasm.h"
+
+#define randomize_lfe_fir(size)                                 \
+    do {                                                        \
+        int i;                                                  \
+        for (i = 0; i < size; i++) {                            \
+            float f = (float)rnd() / (UINT_MAX >> 1) - 1.0f;    \
+            in[i] = f;                                          \
+        }                                                       \
+        for (i = 0; i < 256; i++) {                             \
+            float f = (float)rnd() / (UINT_MAX >> 1) - 1.0f;    \
+            coeffs[i] = f;                                      \
+        }                                                       \
+    } while (0)
+
+#define check_lfe_fir(decifactor, eps)                                  \
+    do {                                                                \
+        LOCAL_ALIGNED_16(float, in,     [256 / decifactor]);            \
+        LOCAL_ALIGNED_16(float, out0,   [decifactor * 2]);              \
+        LOCAL_ALIGNED_16(float, out1,   [decifactor * 2]);              \
+        LOCAL_ALIGNED_16(float, coeffs, [256]);                         \
+        int i;                                                          \
+        const float * in_ptr = in + (256 / decifactor) - 1;             \
+        declare_func(void, float *out, const float *in, const float *coeffs); \
+        /* repeat the test several times */                             \
+        for (i = 0; i < 32; i++) {                                      \
+            int j;                                                      \
+            memset(out0,    0, sizeof(*out0) * 2 * decifactor);         \
+            memset(out1, 0xFF, sizeof(*out1) * 2 * decifactor);         \
+            randomize_lfe_fir(256 / decifactor);                        \
+            call_ref(out0, in_ptr, coeffs);                             \
+            call_new(out1, in_ptr, coeffs);                             \
+            for (j = 0; j < 2 * decifactor; j++) {                      \
+                if (!float_near_abs_eps(out0[j], out1[j], eps)) {       \
+                    if (0) {                                            \
+                        union av_intfloat32 x, y; x.f = out0[j]; y.f = out1[j]; \
+                        fprintf(stderr, "%3d: %11g (0x%08x); %11g (0x%08x)\n", \
+                                j, x.f, x.i, y.f, y.i);                 \
+                    }                                                   \
+                    fail();                                             \
+                    break;                                              \
+                }                                                       \
+            }                                                           \
+            bench_new(out1, in_ptr, coeffs);                            \
+        }                                                               \
+    } while (0)
+
+#define randomize_decode_hf()                                   \
+    do {                                                        \
+        int i;                                                  \
+        for (i = 0; i < DCA_SUBBANDS; i++) {                    \
+            vq_num[i]   = rnd() >> 22;                          \
+            scale[i][0] = rnd() >> 26;                          \
+            scale[i][1] = INT32_MIN;                            \
+        }                                                       \
+    } while (0)
+
+void checkasm_check_dcadsp(void)
+{
+    DCADSPContext c;
+
+    ff_dcadsp_init(&c);
+
+    /* values are limited to {-8, 8} so absolute epsilon is good enough */
+    if (check_func(c.lfe_fir[0], "dca_lfe_fir0"))
+        check_lfe_fir(32, 1.0e-6f);
+
+    if (check_func(c.lfe_fir[1], "dca_lfe_fir1"))
+        check_lfe_fir(64, 1.0e-6f);
+
+    if (check_func(c.decode_hf,  "dca_decode_hf")) {
+        LOCAL_ALIGNED_16(float,   dst0,   [DCA_SUBBANDS], [8]);
+        LOCAL_ALIGNED_16(float,   dst1,   [DCA_SUBBANDS], [8]);
+        LOCAL_ALIGNED_16(int32_t, scale,  [DCA_SUBBANDS], [2]);
+        LOCAL_ALIGNED_16(int32_t, vq_num, [DCA_SUBBANDS]);
+        intptr_t start, end = 32, offset;
+
+        declare_func(void, float[DCA_SUBBANDS][8], const int32_t[DCA_SUBBANDS],
+                     const int8_t[1024][DCA_SUBBANDS], intptr_t, int32_t[DCA_SUBBANDS][2],
+                     intptr_t, intptr_t);
+
+        for (start = 0; start < 32; start++) {
+            for (offset = 0; offset < 32; offset += 8) {
+                int j;
+                for (j = 0; j < DCA_SUBBANDS; j++) {
+                    memset(dst0[j], 0, sizeof(*(dst0[j])) * 8);
+                    memset(dst1[j], 0, sizeof(*(dst1[j])) * 8);
+                }
+                randomize_decode_hf();
+
+                call_ref(dst0, vq_num, ff_dca_high_freq_vq, offset, scale, start, end);
+                call_new(dst1, vq_num, ff_dca_high_freq_vq, offset, scale, start, end);
+
+                for (j = 0; j < 8 * DCA_SUBBANDS; j++) {
+                    if (!float_near_ulp(dst0[j>>3][j&7], dst1[j>>3][j&7], 1)) {
+                        fail();
+                        break;
+                    }
+                }
+
+                bench_new(dst1, vq_num, ff_dca_high_freq_vq, offset, scale, start, end);
+            }
+        }
+    }
+
+    report("dcadsp");
+}