1 files changed, 144 insertions, 0 deletions
diff --git a/third_party/aom/av1/encoder/dwt.c b/third_party/aom/av1/encoder/dwt.c
new file mode 100644
index 000000000..0a57ebcfb
--- /dev/null
+++ b/third_party/aom/av1/encoder/dwt.c
@@ -0,0 +1,144 @@
+#include <assert.h>
+#include <stdlib.h>
+#include <math.h>
+
+#include "config/av1_rtcd.h"
+#include "av1/encoder/dwt.h"
+
+// Note: block length must be even for this implementation
+static void analysis_53_row(int length, tran_low_t *x, tran_low_t *lowpass,
+                            tran_low_t *highpass) {
+  int n;
+  tran_low_t r, *a, *b;
+
+  n = length >> 1;
+  b = highpass;
+  a = lowpass;
+  while (--n) {
+    *a++ = (r = *x++) * 2;
+    *b++ = *x - ((r + x[1] + 1) >> 1);
+    x++;
+  }
+  *a = (r = *x++) * 2;
+  *b = *x - r;
+
+  n = length >> 1;
+  b = highpass;
+  a = lowpass;
+  r = *highpass;
+  while (n--) {
+    *a++ += (r + (*b) + 1) >> 1;
+    r = *b++;
+  }
+}
+
+static void analysis_53_col(int length, tran_low_t *x, tran_low_t *lowpass,
+                            tran_low_t *highpass) {
+  int n;
+  tran_low_t r, *a, *b;
+
+  n = length >> 1;
+  b = highpass;
+  a = lowpass;
+  while (--n) {
+    *a++ = (r = *x++);
+    *b++ = (((*x) * 2) - (r + x[1]) + 2) >> 2;
+    x++;
+  }
+  *a = (r = *x++);
+  *b = (*x - r + 1) >> 1;
+
+  n = length >> 1;
+  b = highpass;
+  a = lowpass;
+  r = *highpass;
+  while (n--) {
+    *a++ += (r + (*b) + 1) >> 1;
+    r = *b++;
+  }
+}
+
+static void dyadic_analyze_53_uint8_input(int levels, int width, int height,
+                                          uint8_t *x, int pitch_x,
+                                          tran_low_t *c, int pitch_c,
+                                          int dwt_scale_bits, int hbd) {
+  int lv, i, j, nh, nw, hh = height, hw = width;
+  tran_low_t buffer[2 * DWT_MAX_LENGTH];
+
+  if (hbd) {
+    uint16_t *x16 = CONVERT_TO_SHORTPTR(x);
+    for (i = 0; i < height; i++) {
+      for (j = 0; j < width; j++) {
+        c[i * pitch_c + j] = x16[i * pitch_x + j] << dwt_scale_bits;
+      }
+    }
+  } else {
+    for (i = 0; i < height; i++) {
+      for (j = 0; j < width; j++) {
+        c[i * pitch_c + j] = x[i * pitch_x + j] << dwt_scale_bits;
+      }
+    }
+  }
+
+  for (lv = 0; lv < levels; lv++) {
+    nh = hh;
+    hh = (hh + 1) >> 1;
+    nw = hw;
+    hw = (hw + 1) >> 1;
+    if ((nh < 2) || (nw < 2)) return;
+    for (i = 0; i < nh; i++) {
+      memcpy(buffer, &c[i * pitch_c], nw * sizeof(tran_low_t));
+      analysis_53_row(nw, buffer, &c[i * pitch_c], &c[i * pitch_c] + hw);
+    }
+    for (j = 0; j < nw; j++) {
+      for (i = 0; i < nh; i++) buffer[i + nh] = c[i * pitch_c + j];
+      analysis_53_col(nh, buffer + nh, buffer, buffer + hh);
+      for (i = 0; i < nh; i++) c[i * pitch_c + j] = buffer[i];
+    }
+  }
+}
+
+void av1_fdwt8x8_uint8_input_c(uint8_t *input, tran_low_t *output, int stride,
+                               int hbd) {
+  dyadic_analyze_53_uint8_input(4, 8, 8, input, stride, output, 8, 2, hbd);
+}
+
+int av1_haar_ac_sad(tran_low_t *output, int bw, int bh, int stride) {
+  int acsad = 0;
+
+  for (int r = 0; r < bh; ++r)
+    for (int c = 0; c < bw; ++c) {
+      if (r >= bh / 2 || c >= bw / 2) acsad += abs(output[r * stride + c]);
+    }
+  return acsad;
+}
+
+uint64_t av1_dct_ac_sad(tran_low_t *output, int bw, int bh, int stride) {
+  uint64_t acsad = 0;
+
+  for (int r = 0; r < bh; ++r)
+    for (int c = 0; c < bw; ++c) {
+      if (r > 0 || c > 0) acsad += abs(output[r * stride + c]);
+    }
+
+  return acsad;
+}
+
+uint32_t av1_variance(uint8_t *input, int bw, int bh, int stride) {
+  int sum = 0;
+  uint32_t sse = 0;
+
+  for (int r = 0; r < bh; ++r)
+    for (int c = 0; c < bw; ++c) {
+      sum += input[r * stride + c];
+      sse += input[r * stride + c] * input[r * stride + c];
+    }
+  return sse - (uint32_t)(((int64_t)sum * sum) / (bw * bh));
+}
+
+int av1_haar_ac_sad_8x8_uint8_input(uint8_t *input, int stride, int hbd) {
+  tran_low_t output[64];
+
+  av1_fdwt8x8_uint8_input_c(input, output, stride, hbd);
+  return av1_haar_ac_sad(output, 8, 8, 8);
+}