Fix unsigned integer overflow in sse computation

Basically port the fix from libaom:
https://aomedia-review.googlesource.com/c/aom/+/169361

Change-Id: Id06a5db91372037832399200ded75d514e096726
(cherry picked from commit a94cdd57ffd95ee7beb48d2794dae538f25da46c)

1 files changed, 24 insertions, 43 deletions

diff --git a/vpx_dsp/psnr.c b/vpx_dsp/psnr.c
index 48bac0450..f0d4e927a 100644
--- a/vpx_dsp/psnr.c
+++ b/vpx_dsp/psnr.c
@@ -26,57 +26,44 @@ double vpx_sse_to_psnr(double samples, double peak, double sse) {
 /* TODO(yaowu): The block_variance calls the unoptimized versions of variance()
  * and highbd_8_variance(). It should not.
  */
-static void encoder_variance(const uint8_t *a, int a_stride, const uint8_t *b,
-                             int b_stride, int w, int h, unsigned int *sse,
-                             int *sum) {
+static int64_t encoder_sse(const uint8_t *a, int a_stride, const uint8_t *b,
+                           int b_stride, int w, int h) {
   int i, j;
-
-  *sum = 0;
-  *sse = 0;
+  int64_t sse = 0;
 
   for (i = 0; i < h; i++) {
     for (j = 0; j < w; j++) {
       const int diff = a[j] - b[j];
-      *sum += diff;
-      *sse += diff * diff;
+      sse += diff * diff;
     }
 
     a += a_stride;
     b += b_stride;
   }
+
+  return sse;
 }
 
 #if CONFIG_VP9_HIGHBITDEPTH
-static void encoder_highbd_variance64(const uint8_t *a8, int a_stride,
-                                      const uint8_t *b8, int b_stride, int w,
-                                      int h, uint64_t *sse, int64_t *sum) {
+static int64_t encoder_highbd_8_sse(const uint8_t *a8, int a_stride,
+                                    const uint8_t *b8, int b_stride, int w,
+                                    int h) {
   int i, j;
+  int64_t sse = 0;
 
   uint16_t *a = CONVERT_TO_SHORTPTR(a8);
   uint16_t *b = CONVERT_TO_SHORTPTR(b8);
-  *sum = 0;
-  *sse = 0;
 
   for (i = 0; i < h; i++) {
     for (j = 0; j < w; j++) {
       const int diff = a[j] - b[j];
-      *sum += diff;
-      *sse += diff * diff;
+      sse += diff * diff;
     }
     a += a_stride;
     b += b_stride;
   }
-}
 
-static void encoder_highbd_8_variance(const uint8_t *a8, int a_stride,
-                                      const uint8_t *b8, int b_stride, int w,
-                                      int h, unsigned int *sse, int *sum) {
-  uint64_t sse_long = 0;
-  int64_t sum_long = 0;
-  encoder_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long,
-                            &sum_long);
-  *sse = (unsigned int)sse_long;
-  *sum = (int)sum_long;
+  return sse;
 }
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
@@ -85,26 +72,23 @@ static int64_t get_sse(const uint8_t *a, int a_stride, const uint8_t *b,
   const int dw = width % 16;
   const int dh = height % 16;
   int64_t total_sse = 0;
-  unsigned int sse = 0;
-  int sum = 0;
   int x, y;
 
   if (dw > 0) {
-    encoder_variance(&a[width - dw], a_stride, &b[width - dw], b_stride, dw,
-                     height, &sse, &sum);
-    total_sse += sse;
+    total_sse += encoder_sse(&a[width - dw], a_stride, &b[width - dw], b_stride,
+                             dw, height);
   }
 
   if (dh > 0) {
-    encoder_variance(&a[(height - dh) * a_stride], a_stride,
-                     &b[(height - dh) * b_stride], b_stride, width - dw, dh,
-                     &sse, &sum);
-    total_sse += sse;
+    total_sse +=
+        encoder_sse(&a[(height - dh) * a_stride], a_stride,
+                    &b[(height - dh) * b_stride], b_stride, width - dw, dh);
   }
 
   for (y = 0; y < height / 16; ++y) {
     const uint8_t *pa = a;
     const uint8_t *pb = b;
+    unsigned int sse;
     for (x = 0; x < width / 16; ++x) {
       vpx_mse16x16(pa, a_stride, pb, b_stride, &sse);
       total_sse += sse;
@@ -146,22 +130,19 @@ static int64_t highbd_get_sse(const uint8_t *a, int a_stride, const uint8_t *b,
   int x, y;
   const int dw = width % 16;
   const int dh = height % 16;
-  unsigned int sse = 0;
-  int sum = 0;
   if (dw > 0) {
-    encoder_highbd_8_variance(&a[width - dw], a_stride, &b[width - dw],
-                              b_stride, dw, height, &sse, &sum);
-    total_sse += sse;
+    total_sse += encoder_highbd_8_sse(&a[width - dw], a_stride, &b[width - dw],
+                                      b_stride, dw, height);
   }
   if (dh > 0) {
-    encoder_highbd_8_variance(&a[(height - dh) * a_stride], a_stride,
-                              &b[(height - dh) * b_stride], b_stride,
-                              width - dw, dh, &sse, &sum);
-    total_sse += sse;
+    total_sse += encoder_highbd_8_sse(&a[(height - dh) * a_stride], a_stride,
+                                      &b[(height - dh) * b_stride], b_stride,
+                                      width - dw, dh);
   }
   for (y = 0; y < height / 16; ++y) {
     const uint8_t *pa = a;
     const uint8_t *pb = b;
+    unsigned int sse;
     for (x = 0; x < width / 16; ++x) {
       vpx_highbd_8_mse16x16(pa, a_stride, pb, b_stride, &sse);
       total_sse += sse;