comp_avg_pred neon: used by sub pixel avg variance

BUG=webm:1423

Change-Id: I33de537f238f58f89b7a6c1c2d6e8110de4b8804

test/comp_avg_pred_test.cc

@@ -156,6 +156,12 @@ INSTANTIATE_TEST_CASE_P(C, AvgPredTest,
 INSTANTIATE_TEST_CASE_P(SSE2, AvgPredTest,
                         ::testing::Values(&vpx_comp_avg_pred_sse2));
 #endif  // HAVE_SSE2
+
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(NEON, AvgPredTest,
+                        ::testing::Values(&vpx_comp_avg_pred_neon));
+#endif  // HAVE_NEON
+
 #if HAVE_VSX
 INSTANTIATE_TEST_CASE_P(VSX, AvgPredTest,
                         ::testing::Values(&vpx_comp_avg_pred_vsx));

vpx_dsp/arm/avg_pred_neon.c

@@ -0,0 +1,55 @@
+/*
+ *  Copyright (c) 2017 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <arm_neon.h>
+#include <assert.h>
+
+#include "./vpx_dsp_rtcd.h"
+#include "vpx_dsp/arm/mem_neon.h"
+
+void vpx_comp_avg_pred_neon(uint8_t *comp, const uint8_t *pred, int width,
+                            int height, const uint8_t *ref, int ref_stride) {
+  if (width > 8) {
+    int x, y;
+    for (y = 0; y < height; ++y) {
+      for (x = 0; x < width; x += 16) {
+        const uint8x16_t p = vld1q_u8(pred + x);
+        const uint8x16_t r = vld1q_u8(ref + x);
+        const uint8x16_t avg = vrhaddq_u8(p, r);
+        vst1q_u8(comp + x, avg);
+      }
+      comp += width;
+      pred += width;
+      ref += ref_stride;
+    }
+  } else {
+    int i;
+    for (i = 0; i < width * height; i += 16) {
+      const uint8x16_t p = vld1q_u8(pred);
+      uint8x16_t r;
+
+      if (width == 4) {
+        r = load_unaligned_u8q(ref, ref_stride);
+        ref += 4 * ref_stride;
+      } else {
+        const uint8x8_t r_0 = vld1_u8(ref);
+        const uint8x8_t r_1 = vld1_u8(ref + ref_stride);
+        assert(width == 8);
+        r = vcombine_u8(r_0, r_1);
+        ref += 2 * ref_stride;
+      }
+      r = vrhaddq_u8(r, p);
+      vst1q_u8(comp, r);
+
+      pred += 16;
+      comp += 16;
+    }
+  }
+}

vpx_dsp/vpx_dsp.mk

@@ -325,6 +325,7 @@ ifneq ($(filter yes,$(CONFIG_ENCODERS) $(CONFIG_POSTPROC) $(CONFIG_VP9_POSTPROC)
 DSP_SRCS-yes            += variance.c
 DSP_SRCS-yes            += variance.h
 
+DSP_SRCS-$(HAVE_NEON)   += arm/avg_pred_neon.c
 DSP_SRCS-$(HAVE_NEON)   += arm/subpel_variance_neon.c
 DSP_SRCS-$(HAVE_NEON)   += arm/variance_neon.c
 

vpx_dsp/vpx_dsp_rtcd_defs.pl

@@ -1175,7 +1175,7 @@ add_proto qw/unsigned int vpx_get4x4sse_cs/, "const unsigned char *src_ptr, int
   specialize qw/vpx_get4x4sse_cs neon msa vsx/;
 
 add_proto qw/void vpx_comp_avg_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride";
-  specialize qw/vpx_comp_avg_pred sse2 vsx/;
+  specialize qw/vpx_comp_avg_pred neon sse2 vsx/;
 
 #
 # Subpixel Variance