vp8: [loongson] optimize copymen with mmi

1. vp8_copy_mem16x16_mmi
2. vp8_copy_mem8x8_mmi
3. vp8_copy_mem8x4_mmi

Change-Id: I3de29a11fa7402df0e48bbb944440b1e66498a65

3 files changed, 118 insertions, 3 deletions

diff --git a/vp8/common/mips/mmi/copymem_mmi.c b/vp8/common/mips/mmi/copymem_mmi.c
new file mode 100644
index 000000000..86a32aa9e
--- /dev/null
+++ b/vp8/common/mips/mmi/copymem_mmi.c
@@ -0,0 +1,114 @@
+/*
+ *  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_ports/asmdefs_mmi.h"
+
+#define COPY_MEM_16X2 \
+  "gsldlc1    %[ftmp0],   0x07(%[src])                    \n\t" \
+  "gsldrc1    %[ftmp0],   0x00(%[src])                    \n\t" \
+  "ldl        %[tmp0],    0x0f(%[src])                    \n\t" \
+  "ldr        %[tmp0],    0x08(%[src])                    \n\t" \
+  MMI_ADDU(%[src],     %[src],         %[src_stride])           \
+  "gssdlc1    %[ftmp0],   0x07(%[dst])                    \n\t" \
+  "gssdrc1    %[ftmp0],   0x00(%[dst])                    \n\t" \
+  "sdl        %[tmp0],    0x0f(%[dst])                    \n\t" \
+  "sdr        %[tmp0],    0x08(%[dst])                    \n\t" \
+  MMI_ADDU(%[dst],      %[dst],        %[dst_stride])           \
+  "gsldlc1    %[ftmp1],   0x07(%[src])                    \n\t" \
+  "gsldrc1    %[ftmp1],   0x00(%[src])                    \n\t" \
+  "ldl        %[tmp1],    0x0f(%[src])                    \n\t" \
+  "ldr        %[tmp1],    0x08(%[src])                    \n\t" \
+  MMI_ADDU(%[src],     %[src],         %[src_stride])           \
+  "gssdlc1    %[ftmp1],   0x07(%[dst])                    \n\t" \
+  "gssdrc1    %[ftmp1],   0x00(%[dst])                    \n\t" \
+  "sdl        %[tmp1],    0x0f(%[dst])                    \n\t" \
+  "sdr        %[tmp1],    0x08(%[dst])                    \n\t" \
+  MMI_ADDU(%[dst],     %[dst],         %[dst_stride])
+
+#define COPY_MEM_8X2 \
+  "gsldlc1    %[ftmp0],   0x07(%[src])                    \n\t" \
+  "gsldrc1    %[ftmp0],   0x00(%[src])                    \n\t" \
+  MMI_ADDU(%[src],     %[src],         %[src_stride])           \
+  "ldl        %[tmp0],    0x07(%[src])                    \n\t" \
+  "ldr        %[tmp0],    0x00(%[src])                    \n\t" \
+  MMI_ADDU(%[src],     %[src],         %[src_stride])           \
+                                                                \
+  "gssdlc1    %[ftmp0],   0x07(%[dst])                    \n\t" \
+  "gssdrc1    %[ftmp0],   0x00(%[dst])                    \n\t" \
+  MMI_ADDU(%[dst],      %[dst],        %[dst_stride])           \
+  "sdl        %[tmp0],    0x07(%[dst])                    \n\t" \
+  "sdr        %[tmp0],    0x00(%[dst])                    \n\t" \
+  MMI_ADDU(%[dst],     %[dst],         %[dst_stride])
+
+void vp8_copy_mem16x16_mmi(unsigned char *src, int src_stride,
+                           unsigned char *dst, int dst_stride) {
+  double ftmp[2];
+  uint64_t tmp[2];
+  uint8_t loop_count = 4;
+
+  /* clang-format off */
+  __asm__ volatile (
+    "1:                                                     \n\t"
+    COPY_MEM_16X2
+    COPY_MEM_16X2
+    MMI_ADDIU(%[loop_count], %[loop_count], -0x01)
+    "bnez       %[loop_count],    1b                        \n\t"
+    : [ftmp0]"=&f"(ftmp[0]),            [ftmp1]"=&f"(ftmp[1]),
+      [tmp0]"=&r"(tmp[0]),              [tmp1]"=&r"(tmp[1]),
+      [loop_count]"+&r"(loop_count),
+      [dst]"+&r"(dst),                  [src]"+&r"(src)
+    : [src_stride]"r"((mips_reg)src_stride),
+      [dst_stride]"r"((mips_reg)dst_stride)
+    : "memory"
+  );
+  /* clang-format on */
+}
+
+void vp8_copy_mem8x8_mmi(unsigned char *src, int src_stride, unsigned char *dst,
+                         int dst_stride) {
+  double ftmp[2];
+  uint64_t tmp[1];
+  uint8_t loop_count = 4;
+
+  /* clang-format off */
+  __asm__ volatile (
+    "1:                                                     \n\t"
+    COPY_MEM_8X2
+    MMI_ADDIU(%[loop_count], %[loop_count], -0x01)
+    "bnez       %[loop_count],    1b                        \n\t"
+    : [ftmp0]"=&f"(ftmp[0]),            [ftmp1]"=&f"(ftmp[1]),
+      [tmp0]"=&r"(tmp[0]),              [loop_count]"+&r"(loop_count),
+      [dst]"+&r"(dst),                  [src]"+&r"(src)
+    : [src_stride]"r"((mips_reg)src_stride),
+      [dst_stride]"r"((mips_reg)dst_stride)
+    : "memory"
+  );
+  /* clang-format on */
+}
+
+void vp8_copy_mem8x4_mmi(unsigned char *src, int src_stride, unsigned char *dst,
+                         int dst_stride) {
+  double ftmp[2];
+  uint64_t tmp[1];
+
+  /* clang-format off */
+  __asm__ volatile (
+    COPY_MEM_8X2
+    COPY_MEM_8X2
+    : [ftmp0]"=&f"(ftmp[0]),            [ftmp1]"=&f"(ftmp[1]),
+      [tmp0]"=&r"(tmp[0]),
+      [dst]"+&r"(dst),                  [src]"+&r"(src)
+    : [src_stride]"r"((mips_reg)src_stride),
+      [dst_stride]"r"((mips_reg)dst_stride)
+    : "memory"
+  );
+  /* clang-format on */
+}
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index ce5782363..7a04ef17f 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -104,13 +104,13 @@ specialize qw/vp8_dc_only_idct_add mmx neon dspr2 msa mmi/;
 # RECON
 #
 add_proto qw/void vp8_copy_mem16x16/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
-specialize qw/vp8_copy_mem16x16 sse2 neon dspr2 msa/;
+specialize qw/vp8_copy_mem16x16 sse2 neon dspr2 msa mmi/;
 
 add_proto qw/void vp8_copy_mem8x8/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
-specialize qw/vp8_copy_mem8x8 mmx neon dspr2 msa/;
+specialize qw/vp8_copy_mem8x8 mmx neon dspr2 msa mmi/;
 
 add_proto qw/void vp8_copy_mem8x4/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
-specialize qw/vp8_copy_mem8x4 mmx neon dspr2 msa/;
+specialize qw/vp8_copy_mem8x4 mmx neon dspr2 msa mmi/;
 
 #
 # Postproc
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 3aa3b7db6..5813c81c4 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -121,6 +121,7 @@ VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/sixtap_filter_mmi.c
 VP8_COMMON_SRCS-$(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
 
 ifeq ($(CONFIG_POSTPROC),yes)
 VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/mfqe_msa.c