libvpx: Pull from upstream

source/libvpx/vp9/encoder/x86/vp9_variance_avx2.c

Index: source/libvpx/vp9/encoder/x86/vp9_variance_avx2.c
===================================================================
--- source/libvpx/vp9/encoder/x86/vp9_variance_avx2.c	(revision 291857)
+++ source/libvpx/vp9/encoder/x86/vp9_variance_avx2.c	(working copy)
@@ -12,67 +12,39 @@
 #include "vp9/encoder/vp9_variance.h"
 #include "vpx_ports/mem.h"
 
-typedef void (*get_var_avx2) (
-  const unsigned char *src_ptr,
-  int source_stride,
-  const unsigned char *ref_ptr,
-  int recon_stride,
-  unsigned int *SSE,
-  int *Sum
-);
+typedef void (*get_var_avx2)(const uint8_t *src, int src_stride,
+                             const uint8_t *ref, int ref_stride,
+                             unsigned int *sse, int *sum);
 
-void vp9_get16x16var_avx2
-(
-  const unsigned char *src_ptr,
-  int source_stride,
-  const unsigned char *ref_ptr,
-  int recon_stride,
-  unsigned int *SSE,
-  int *Sum
-);
+void vp9_get16x16var_avx2(const uint8_t *src, int src_stride,
+                          const uint8_t *ref, int ref_stride,
+                          unsigned int *sse, int *sum);
 
-void vp9_get32x32var_avx2
-(
-  const unsigned char *src_ptr,
-  int source_stride,
-  const unsigned char *ref_ptr,
-  int recon_stride,
-  unsigned int *SSE,
-  int *Sum
-);
+void vp9_get32x32var_avx2(const uint8_t *src, int src_stride,
+                          const uint8_t *ref, int ref_stride,
+                          unsigned int *sse, int *sum);
 
-unsigned int vp9_sub_pixel_variance32xh_avx2
-(
-  const uint8_t *src,
-  int src_stride,
-  int x_offset,
-  int y_offset,
-  const uint8_t *dst,
-  int dst_stride,
-  int height,
-  unsigned int *sse
-);
+unsigned int vp9_sub_pixel_variance32xh_avx2(const uint8_t *src, int src_stride,
+                                             int x_offset, int y_offset,
+                                             const uint8_t *dst, int dst_stride,
+                                             int height,
+                                             unsigned int *sse);
 
-unsigned int vp9_sub_pixel_avg_variance32xh_avx2
-(
-  const uint8_t *src,
-  int src_stride,
-  int x_offset,
-  int y_offset,
-  const uint8_t *dst,
-  int dst_stride,
-  const uint8_t *sec,
-  int sec_stride,
-  int height,
-  unsigned int *sseptr
-);
+unsigned int vp9_sub_pixel_avg_variance32xh_avx2(const uint8_t *src,
+                                                 int src_stride,
+                                                 int x_offset,
+                                                 int y_offset,
+                                                 const uint8_t *dst,
+                                                 int dst_stride,
+                                                 const uint8_t *sec,
+                                                 int sec_stride,
+                                                 int height,
+                                                 unsigned int *sseptr);
 
-static void variance_avx2(const unsigned char *src_ptr, int  source_stride,
-                        const unsigned char *ref_ptr, int  recon_stride,
-                        int  w, int  h, unsigned int *sse, int *sum,
-                        get_var_avx2 var_fn, int block_size) {
-  unsigned int sse0;
-  int sum0;
+static void variance_avx2(const uint8_t *src, int src_stride,
+                          const uint8_t *ref, int  ref_stride,
+                          int w, int h, unsigned int *sse, int *sum,
+                          get_var_avx2 var_fn, int block_size) {
   int i, j;
 
   *sse = 0;
@@ -80,105 +52,68 @@
 
   for (i = 0; i < h; i += 16) {
     for (j = 0; j < w; j += block_size) {
-      // processing 16 rows horizontally each call
-      var_fn(src_ptr + source_stride * i + j, source_stride,
-             ref_ptr + recon_stride * i + j, recon_stride, &sse0, &sum0);
+      unsigned int sse0;
+      int sum0;
+      var_fn(&src[src_stride * i + j], src_stride,
+             &ref[ref_stride * i + j], ref_stride, &sse0, &sum0);
       *sse += sse0;
       *sum += sum0;
     }
   }
 }
 
-unsigned int vp9_variance16x16_avx2
-(
-  const unsigned char *src_ptr,
-  int  source_stride,
-  const unsigned char *ref_ptr,
-  int  recon_stride,
-  unsigned int *sse) {
-  unsigned int var;
-  int avg;
 
-  variance_avx2(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16,
-                &var, &avg, vp9_get16x16var_avx2, 16);
-  *sse = var;
-  return (var - (((unsigned int)avg * avg) >> 8));
+unsigned int vp9_variance16x16_avx2(const uint8_t *src, int src_stride,
+                                    const uint8_t *ref, int ref_stride,
+                                    unsigned int *sse) {
+  int sum;
+  variance_avx2(src, src_stride, ref, ref_stride, 16, 16,
+                sse, &sum, vp9_get16x16var_avx2, 16);
+  return *sse - (((unsigned int)sum * sum) >> 8);
 }
 
-unsigned int vp9_mse16x16_avx2(
-  const unsigned char *src_ptr,
-  int  source_stride,
-  const unsigned char *ref_ptr,
-  int  recon_stride,
-  unsigned int *sse) {
-  unsigned int sse0;
-  int sum0;
-  vp9_get16x16var_avx2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
-                       &sum0);
-  *sse = sse0;
-  return sse0;
+unsigned int vp9_mse16x16_avx2(const uint8_t *src, int src_stride,
+                               const uint8_t *ref, int ref_stride,
+                               unsigned int *sse) {
+  int sum;
+  vp9_get16x16var_avx2(src, src_stride, ref, ref_stride, sse, &sum);
+  return *sse;
 }
 
-unsigned int vp9_variance32x32_avx2(const uint8_t *src_ptr,
-                                    int  source_stride,
-                                    const uint8_t *ref_ptr,
-                                    int  recon_stride,
+unsigned int vp9_variance32x16_avx2(const uint8_t *src, int src_stride,
+                                    const uint8_t *ref, int ref_stride,
                                     unsigned int *sse) {
-  unsigned int var;
-  int avg;
-
-  // processing 32 elements vertically in parallel
-  variance_avx2(src_ptr, source_stride, ref_ptr, recon_stride, 32, 32,
-                &var, &avg, vp9_get32x32var_avx2, 32);
-  *sse = var;
-  return (var - (((int64_t)avg * avg) >> 10));
+  int sum;
+  variance_avx2(src, src_stride, ref, ref_stride, 32, 16,
+                sse, &sum, vp9_get32x32var_avx2, 32);
+  return *sse - (((int64_t)sum * sum) >> 9);
 }
 
-unsigned int vp9_variance32x16_avx2(const uint8_t *src_ptr,
-                                    int  source_stride,
-                                    const uint8_t *ref_ptr,
-                                    int  recon_stride,
+unsigned int vp9_variance32x32_avx2(const uint8_t *src, int src_stride,
+                                    const uint8_t *ref, int ref_stride,
                                     unsigned int *sse) {
-  unsigned int var;
-  int avg;
-
-  // processing 32 elements vertically in parallel
-  variance_avx2(src_ptr, source_stride, ref_ptr, recon_stride, 32, 16,
-                &var, &avg, vp9_get32x32var_avx2, 32);
-  *sse = var;
-  return (var - (((int64_t)avg * avg) >> 9));
+  int sum;
+  variance_avx2(src, src_stride, ref, ref_stride, 32, 32,
+                sse, &sum, vp9_get32x32var_avx2, 32);
+  return *sse - (((int64_t)sum * sum) >> 10);
 }
 
-
-unsigned int vp9_variance64x64_avx2(const uint8_t *src_ptr,
-                                    int  source_stride,
-                                    const uint8_t *ref_ptr,
-                                    int  recon_stride,
+unsigned int vp9_variance64x64_avx2(const uint8_t *src, int src_stride,
+                                    const uint8_t *ref, int ref_stride,
                                     unsigned int *sse) {
-  unsigned int var;
-  int avg;
-
-  // processing 32 elements vertically in parallel
-  variance_avx2(src_ptr, source_stride, ref_ptr, recon_stride, 64, 64,
-                &var, &avg, vp9_get32x32var_avx2, 32);
-  *sse = var;
-  return (var - (((int64_t)avg * avg) >> 12));
+  int sum;
+  variance_avx2(src, src_stride, ref, ref_stride, 64, 64,
+                sse, &sum, vp9_get32x32var_avx2, 32);
+  return *sse - (((int64_t)sum * sum) >> 12);
 }
 
-unsigned int vp9_variance64x32_avx2(const uint8_t *src_ptr,
-                                    int  source_stride,
-                                    const uint8_t *ref_ptr,
-                                    int  recon_stride,
+unsigned int vp9_variance64x32_avx2(const uint8_t *src, int src_stride,
+                                    const uint8_t *ref, int ref_stride,
                                     unsigned int *sse) {
-  unsigned int var;
-  int avg;
-
-  // processing 32 elements vertically in parallel
-  variance_avx2(src_ptr, source_stride, ref_ptr, recon_stride, 64, 32,
-                &var, &avg, vp9_get32x32var_avx2, 32);
-
-  *sse = var;
-  return (var - (((int64_t)avg * avg) >> 11));
+  int sum;
+  variance_avx2(src, src_stride, ref, ref_stride, 64, 32,
+                sse, &sum, vp9_get32x32var_avx2, 32);
+  return *sse - (((int64_t)sum * sum) >> 11);
 }
 
 unsigned int vp9_sub_pixel_variance64x64_avx2(const uint8_t *src,
@@ -187,22 +122,19 @@
                                               int y_offset,
                                               const uint8_t *dst,
                                               int dst_stride,
-                                              unsigned int *sse_ptr) {
-  // processing 32 elements in parallel
-  unsigned int sse;
-  int se = vp9_sub_pixel_variance32xh_avx2(src, src_stride, x_offset,
-                                           y_offset, dst, dst_stride,
-                                           64, &sse);
-  // processing the next 32 elements in parallel
+                                              unsigned int *sse) {
+  unsigned int sse1;
+  const int se1 = vp9_sub_pixel_variance32xh_avx2(src, src_stride, x_offset,
+                                                  y_offset, dst, dst_stride,
+                                                  64, &sse1);
   unsigned int sse2;
-  int se2 = vp9_sub_pixel_variance32xh_avx2(src + 32, src_stride,
-                                            x_offset, y_offset,
-                                            dst + 32, dst_stride,
-                                            64, &sse2);
-  se += se2;
-  sse += sse2;
-  *sse_ptr = sse;
-  return sse - (((int64_t)se * se) >> 12);
+  const int se2 = vp9_sub_pixel_variance32xh_avx2(src + 32, src_stride,
+                                                  x_offset, y_offset,
+                                                  dst + 32, dst_stride,
+                                                  64, &sse2);
+  const int se = se1 + se2;
+  *sse = sse1 + sse2;
+  return *sse - (((int64_t)se * se) >> 12);
 }
 
 unsigned int vp9_sub_pixel_variance32x32_avx2(const uint8_t *src,
@@ -211,14 +143,11 @@
                                               int y_offset,
                                               const uint8_t *dst,
                                               int dst_stride,
-                                              unsigned int *sse_ptr) {
-  // processing 32 element in parallel
-  unsigned int sse;
-  int se = vp9_sub_pixel_variance32xh_avx2(src, src_stride, x_offset,
-                                           y_offset, dst, dst_stride,
-                                           32, &sse);
-  *sse_ptr = sse;
-  return sse - (((int64_t)se * se) >> 10);
+                                              unsigned int *sse) {
+  const int se = vp9_sub_pixel_variance32xh_avx2(src, src_stride, x_offset,
+                                                 y_offset, dst, dst_stride,
+                                                 32, sse);
+  return *sse - (((int64_t)se * se) >> 10);
 }
 
 unsigned int vp9_sub_pixel_avg_variance64x64_avx2(const uint8_t *src,
@@ -227,24 +156,22 @@
                                                   int y_offset,
                                                   const uint8_t *dst,
                                                   int dst_stride,
-                                                  unsigned int *sseptr,
+                                                  unsigned int *sse,
                                                   const uint8_t *sec) {
-  // processing 32 elements in parallel
-  unsigned int sse;
-
-  int se = vp9_sub_pixel_avg_variance32xh_avx2(src, src_stride, x_offset,
-                                               y_offset, dst, dst_stride,
-                                               sec, 64, 64, &sse);
+  unsigned int sse1;
+  const int se1 = vp9_sub_pixel_avg_variance32xh_avx2(src, src_stride, x_offset,
+                                                      y_offset, dst, dst_stride,
+                                                      sec, 64, 64, &sse1);
   unsigned int sse2;
-  // processing the next 32 elements in parallel
-  int se2 = vp9_sub_pixel_avg_variance32xh_avx2(src + 32, src_stride, x_offset,
-                                                y_offset, dst + 32, dst_stride,
-                                                sec + 32, 64, 64, &sse2);
-  se += se2;
-  sse += sse2;
-  *sseptr = sse;
+  const int se2 =
+      vp9_sub_pixel_avg_variance32xh_avx2(src + 32, src_stride, x_offset,
+                                          y_offset, dst + 32, dst_stride,
+                                          sec + 32, 64, 64, &sse2);
+  const int se = se1 + se2;
 
-  return sse - (((int64_t)se * se) >> 12);
+  *sse = sse1 + sse2;
+
+  return *sse - (((int64_t)se * se) >> 12);
 }
 
 unsigned int vp9_sub_pixel_avg_variance32x32_avx2(const uint8_t *src,
@@ -253,15 +180,11 @@
                                                   int y_offset,
                                                   const uint8_t *dst,
                                                   int dst_stride,
-                                                  unsigned int *sseptr,
+                                                  unsigned int *sse,
                                                   const uint8_t *sec) {
   // processing 32 element in parallel
-  unsigned int sse;
-  int se = vp9_sub_pixel_avg_variance32xh_avx2(src, src_stride, x_offset,
-                                                 y_offset, dst, dst_stride,
-                                                 sec, 32, 32, &sse);
-  *sseptr = sse;
-  return sse - (((int64_t)se * se) >> 10);
+  const int se = vp9_sub_pixel_avg_variance32xh_avx2(src, src_stride, x_offset,
+                                                     y_offset, dst, dst_stride,
+                                                     sec, 32, 32, sse);
+  return *sse - (((int64_t)se * se) >> 10);
 }
-
-