vp8: [loongson] optimize idct with mmi

1. vp8_dequant_idct_add_y_block_mmi
2. vp8_dequant_idct_add_uv_block_mmi

Change-Id: I9987147be2685ac79d4b045d1d56f6709ee1223c

diff --git a/vp8/common/mips/mmi/idct_blk_mmi.c b/vp8/common/mips/mmi/idct_blk_mmi.c
new file mode 100644 (file)
index 0000000..f6020ab
--- /dev/null
+++ b/vp8/common/mips/mmi/idct_blk_mmi.c
@@ -0,0 +1,71 @@
+/*
+ *  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 "./vp8_rtcd.h"
+#include "vpx_mem/vpx_mem.h"
+
+void vp8_dequant_idct_add_y_block_mmi(int16_t *q, int16_t *dq, uint8_t *dst,
+                                      int stride, int8_t *eobs) {
+  int i, j;
+
+  for (i = 0; i < 4; i++) {
+    for (j = 0; j < 4; j++) {
+      if (*eobs++ > 1) {
+        vp8_dequant_idct_add_mmi(q, dq, dst, stride);
+      } else {
+        vp8_dc_only_idct_add_mmi(q[0] * dq[0], dst, stride, dst, stride);
+        memset(q, 0, 2 * sizeof(q[0]));
+      }
+
+      q += 16;
+      dst += 4;
+    }
+
+    dst += 4 * stride - 16;
+  }
+}
+
+void vp8_dequant_idct_add_uv_block_mmi(int16_t *q, int16_t *dq, uint8_t *dstu,
+                                       uint8_t *dstv, int stride,
+                                       int8_t *eobs) {
+  int i, j;
+
+  for (i = 0; i < 2; i++) {
+    for (j = 0; j < 2; j++) {
+      if (*eobs++ > 1) {
+        vp8_dequant_idct_add_mmi(q, dq, dstu, stride);
+      } else {
+        vp8_dc_only_idct_add_mmi(q[0] * dq[0], dstu, stride, dstu, stride);
+        memset(q, 0, 2 * sizeof(q[0]));
+      }
+
+      q += 16;
+      dstu += 4;
+    }
+
+    dstu += 4 * stride - 8;
+  }
+
+  for (i = 0; i < 2; i++) {
+    for (j = 0; j < 2; j++) {
+      if (*eobs++ > 1) {
+        vp8_dequant_idct_add_mmi(q, dq, dstv, stride);
+      } else {
+        vp8_dc_only_idct_add_mmi(q[0] * dq[0], dstv, stride, dstv, stride);
+        memset(q, 0, 2 * sizeof(q[0]));
+      }
+
+      q += 16;
+      dstv += 4;
+    }
+
+    dstv += 4 * stride - 8;
+  }
+}

diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index be2ac00546d8825dd98ec39379d37e12c44eae83..ece2785eb7410fd2f1c67b5a74578eaccfe2f792 100644 (file)
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -28,10 +28,10 @@ add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char
 specialize qw/vp8_dequant_idct_add mmx neon dspr2 msa mmi/;
 
 add_proto qw/void vp8_dequant_idct_add_y_block/, "short *q, short *dq, unsigned char *dst, int stride, char *eobs";
-specialize qw/vp8_dequant_idct_add_y_block sse2 neon dspr2 msa/;
+specialize qw/vp8_dequant_idct_add_y_block sse2 neon dspr2 msa mmi/;
 
 add_proto qw/void vp8_dequant_idct_add_uv_block/, "short *q, short *dq, unsigned char *dst_u, unsigned char *dst_v, int stride, char *eobs";
-specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa/;
+specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa mmi/;
 
 #
 # Loopfilter

diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 5813c81c42e50c246302721e1c3f93c7768d3900..246fe6a6772c27a5523f5fd40c308fdd02c80afe 100644 (file)
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -122,6 +122,7 @@ VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/loopfilter_filters_mmi.c
 VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/idctllm_mmi.c
 VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/dequantize_mmi.c
 VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/copymem_mmi.c
+VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/idct_blk_mmi.c
 
 ifeq ($(CONFIG_POSTPROC),yes)
 VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/mfqe_msa.c