avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for block functions

This patch adds MSA (MIPS-SIMD-Arch) optimizations for block functions in new file blockdsp_msa.c

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/blockdsp.c

@@ -77,4 +77,6 @@ av_cold void ff_blockdsp_init(BlockDSPContext *c, AVCodecContext *avctx)
 #else
         ff_blockdsp_init_x86(c, high_bit_depth);
 #endif /* FF_API_XVMC */
+    if (ARCH_MIPS)
+        ff_blockdsp_init_mips(c, high_bit_depth);
 }

libavcodec/blockdsp.h

@@ -49,5 +49,6 @@ void ff_blockdsp_init_x86(BlockDSPContext *c, unsigned high_bit_depth,
 #else
 void ff_blockdsp_init_x86(BlockDSPContext *c, unsigned high_bit_depth);
 #endif /* FF_API_XVMC */
+void ff_blockdsp_init_mips(BlockDSPContext *c, unsigned high_bit_depth);
 
 #endif /* AVCODEC_BLOCKDSP_H */

libavcodec/mips/Makefile

@@ -27,6 +27,7 @@ OBJS-$(CONFIG_H264PRED)                   += mips/h264pred_init_mips.o
 OBJS-$(CONFIG_H263DSP)                    += mips/h263dsp_init_mips.o
 OBJS-$(CONFIG_QPELDSP)                    += mips/qpeldsp_init_mips.o
 OBJS-$(CONFIG_HPELDSP)                    += mips/hpeldsp_init_mips.o
+OBJS-$(CONFIG_BLOCKDSP)                   += mips/blockdsp_init_mips.o
 MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o            \
                                              mips/hevc_mc_uni_msa.o        \
                                              mips/hevc_mc_uniw_msa.o       \
@@ -43,5 +44,6 @@ MSA-OBJS-$(CONFIG_H264PRED)               += mips/h264pred_msa.o
 MSA-OBJS-$(CONFIG_H263DSP)                += mips/h263dsp_msa.o
 MSA-OBJS-$(CONFIG_QPELDSP)                += mips/qpeldsp_msa.o
 MSA-OBJS-$(CONFIG_HPELDSP)                += mips/hpeldsp_msa.o
+MSA-OBJS-$(CONFIG_BLOCKDSP)               += mips/blockdsp_msa.o
 LOONGSON3-OBJS-$(CONFIG_H264DSP)          += mips/h264dsp_mmi.o
 LOONGSON3-OBJS-$(CONFIG_H264CHROMA)       += mips/h264chroma_mmi.o

libavcodec/mips/blockdsp_init_mips.c

@@ -0,0 +1,40 @@
+/*
+ * Copyright (c) 2015 Parag Salasakar (parag.salasakar@imgtec.com)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "blockdsp_mips.h"
+
+#if HAVE_MSA
+static av_cold void blockdsp_init_msa(BlockDSPContext *c,
+                                      unsigned high_bit_depth)
+{
+    c->clear_block = ff_clear_block_msa;
+    c->clear_blocks = ff_clear_blocks_msa;
+
+    c->fill_block_tab[0] = ff_fill_block16_msa;
+    c->fill_block_tab[1] = ff_fill_block8_msa;
+}
+#endif  // #if HAVE_MSA
+
+void ff_blockdsp_init_mips(BlockDSPContext *c, unsigned high_bit_depth)
+{
+#if HAVE_MSA
+    blockdsp_init_msa(c, high_bit_depth);
+#endif  // #if HAVE_MSA
+}

libavcodec/mips/blockdsp_mips.h

@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2015 Parag Salasakar (parag.salasakar@imgtec.com)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_MIPS_BLOCKDSP_MIPS_H
+#define AVCODEC_MIPS_BLOCKDSP_MIPS_H
+
+#include "../mpegvideo.h"
+
+void ff_fill_block16_msa(uint8_t *src, uint8_t val, int stride, int height);
+void ff_fill_block8_msa(uint8_t *src, uint8_t val, int stride, int height);
+void ff_clear_block_msa(int16_t *block);
+void ff_clear_blocks_msa(int16_t *block);
+
+#endif  // #ifndef AVCODEC_MIPS_BLOCKDSP_MIPS_H

libavcodec/mips/blockdsp_msa.c

@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) 2015 Parag Salasakar (parag.salasakar@imgtec.com)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/mips/generic_macros_msa.h"
+#include "blockdsp_mips.h"
+
+static void copy_8bit_value_width8_msa(uint8_t *src, uint8_t val,
+                                       int32_t src_stride, int32_t height)
+{
+    int32_t cnt;
+    uint64_t dst0;
+    v16u8 val0;
+
+    val0 = (v16u8) __msa_fill_b(val);
+    dst0 = __msa_copy_u_d((v2i64) val0, 0);
+
+    for (cnt = (height >> 2); cnt--;) {
+        SD4(dst0, dst0, dst0, dst0, src, src_stride);
+        src += (4 * src_stride);
+    }
+}
+
+static void copy_8bit_value_width16_msa(uint8_t *src, uint8_t val,
+                                        int32_t src_stride, int32_t height)
+{
+    int32_t cnt;
+    v16u8 val0;
+
+    val0 = (v16u8) __msa_fill_b(val);
+
+    for (cnt = (height >> 3); cnt--;) {
+        ST_UB8(val0, val0, val0, val0, val0, val0, val0, val0, src, src_stride);
+        src += (8 * src_stride);
+    }
+}
+
+static void memset_zero_16width_msa(uint8_t *src, int32_t stride,
+                                    int32_t height)
+{
+    int8_t cnt;
+    v16u8 zero = { 0 };
+
+    for (cnt = (height / 2); cnt--;) {
+        ST_UB(zero, src);
+        src += stride;
+        ST_UB(zero, src);
+        src += stride;
+    }
+}
+
+void ff_fill_block16_msa(uint8_t *src, uint8_t val, int stride, int height)
+{
+    copy_8bit_value_width16_msa(src, val, stride, height);
+}
+
+void ff_fill_block8_msa(uint8_t *src, uint8_t val, int stride, int height)
+{
+    copy_8bit_value_width8_msa(src, val, stride, height);
+}
+
+void ff_clear_block_msa(int16_t *block)
+{
+    memset_zero_16width_msa((uint8_t *) block, 16, 8);
+}
+
+void ff_clear_blocks_msa(int16_t *block)
+{
+    memset_zero_16width_msa((uint8_t *) block, 16, 8 * 6);
+}