avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/mips/Makefile

@@ -77,4 +77,5 @@ MMI-OBJS-$(CONFIG_MPEG4_DECODER)          += mips/xvid_idct_mmi.o
 MMI-OBJS-$(CONFIG_BLOCKDSP)               += mips/blockdsp_mmi.o
 MMI-OBJS-$(CONFIG_PIXBLOCKDSP)            += mips/pixblockdsp_mmi.o
 MMI-OBJS-$(CONFIG_H264QPEL)               += mips/h264qpel_mmi.o
+MMI-OBJS-$(CONFIG_VP8_DECODER)            += mips/vp8dsp_mmi.o
 MMI-OBJS-$(CONFIG_HPELDSP)                += mips/hpeldsp_mmi.o

libavcodec/mips/constants.c

@@ -28,6 +28,7 @@ DECLARE_ALIGNED(8, const uint64_t, ff_pw_2) =       {0x0002000200020002ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_3) =       {0x0003000300030003ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_4) =       {0x0004000400040004ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_5) =       {0x0005000500050005ULL};
+DECLARE_ALIGNED(8, const uint64_t, ff_pw_6) =       {0x0006000600060006ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_8) =       {0x0008000800080008ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_9) =       {0x0009000900090009ULL};
 DECLARE_ALIGNED(8, const uint64_t, ff_pw_10) =      {0x000A000A000A000AULL};

libavcodec/mips/constants.h

@@ -29,6 +29,7 @@ extern const uint64_t ff_pw_2;
 extern const uint64_t ff_pw_3;
 extern const uint64_t ff_pw_4;
 extern const uint64_t ff_pw_5;
+extern const uint64_t ff_pw_6;
 extern const uint64_t ff_pw_8;
 extern const uint64_t ff_pw_9;
 extern const uint64_t ff_pw_10;

libavcodec/mips/vp8dsp_init_mips.c

@@ -1,5 +1,6 @@
 /*
  * Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale@imgtec.com)
+ * Copyright (c) 2016 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
  *
  * This file is part of FFmpeg.
  *
@@ -105,9 +106,97 @@ static av_cold void vp8dsp_init_msa(VP8DSPContext *dsp)
 }
 #endif  // #if HAVE_MSA
 
+#if HAVE_MMI
+static av_cold void vp8dsp_init_mmi(VP8DSPContext *dsp)
+{
+    dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_mmi;
+    dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_mmi;
+    dsp->vp8_idct_add       = ff_vp8_idct_add_mmi;
+    dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_mmi;
+    dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_mmi;
+    dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmi;
+
+    dsp->put_vp8_epel_pixels_tab[0][0][1] = ff_put_vp8_epel16_h4_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][1][0] = ff_put_vp8_epel16_v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][1][1] = ff_put_vp8_epel16_h4v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][1][2] = ff_put_vp8_epel16_h6v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][2][1] = ff_put_vp8_epel16_h4v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_mmi;
+
+    dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_mmi;
+
+    dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_mmi;
+    dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_mmi;
+
+    dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilinear16_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilinear16_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilinear16_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilinear16_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilinear16_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilinear16_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilinear16_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilinear16_hv_mmi;
+
+    dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilinear8_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilinear8_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilinear8_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilinear8_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilinear8_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilinear8_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilinear8_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilinear8_hv_mmi;
+
+    dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilinear4_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilinear4_h_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilinear4_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilinear4_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilinear4_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilinear4_v_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilinear4_hv_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilinear4_hv_mmi;
+
+    dsp->put_vp8_epel_pixels_tab[0][0][0]     = ff_put_vp8_pixels16_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmi;
+
+    dsp->put_vp8_epel_pixels_tab[1][0][0]     = ff_put_vp8_pixels8_mmi;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmi;
+
+    dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_mmi;
+    dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_mmi;
+    dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mmi;
+    dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mmi;
+
+    dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_mmi;
+    dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_mmi;
+    dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmi;
+    dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmi;
+
+    dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmi;
+    dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmi;
+}
+#endif /* HAVE_MMI */
+
 av_cold void ff_vp8dsp_init_mips(VP8DSPContext *dsp)
 {
 #if HAVE_MSA
     vp8dsp_init_msa(dsp);
 #endif  // #if HAVE_MSA
+#if HAVE_MMI
+    vp8dsp_init_mmi(dsp);
+#endif /* HAVE_MMI */
 }

libavcodec/mips/vp8dsp_mips.h

@@ -1,5 +1,6 @@
 /*
  * Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale@imgtec.com)
+ * Copyright (c) 2016 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
  *
  * This file is part of FFmpeg.
  *
@@ -21,6 +22,11 @@
 #ifndef AVCODEC_MIPS_VP8DSP_MIPS_H
 #define AVCODEC_MIPS_VP8DSP_MIPS_H
 
+#include "libavutil/mem.h"
+#include "libavcodec/vp8dsp.h"
+#include "libavcodec/mathops.h"
+#include "constants.h"
+
 void ff_put_vp8_pixels4_msa(uint8_t *dst, ptrdiff_t dststride,
                             uint8_t *src, ptrdiff_t srcstride,
                             int h, int x, int y);
@@ -169,4 +175,115 @@ void ff_vp8_idct_dc_add4uv_msa(uint8_t *dst, int16_t block[4][16],
 void ff_vp8_idct_dc_add4y_msa(uint8_t *dst, int16_t block[4][16],
                               ptrdiff_t stride);
 
+void ff_vp8_luma_dc_wht_mmi(int16_t block[4][4][16], int16_t dc[16]);
+void ff_vp8_luma_dc_wht_dc_mmi(int16_t block[4][4][16], int16_t dc[16]);
+void ff_vp8_idct_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4y_mmi(uint8_t *dst, int16_t block[4][16],
+        ptrdiff_t stride);
+void ff_vp8_idct_dc_add4uv_mmi(uint8_t *dst, int16_t block[4][16],
+        ptrdiff_t stride);
+
+void ff_put_vp8_pixels4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int x, int y);
+void ff_put_vp8_pixels8_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int x, int y);
+void ff_put_vp8_pixels16_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int x, int y);
+
+void ff_put_vp8_epel16_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
+        ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel16_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+void ff_put_vp8_epel8_h4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_h6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel8_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+void ff_put_vp8_epel4_h4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_h6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_epel4_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+void ff_put_vp8_bilinear16_h_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear16_v_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear16_hv_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+void ff_put_vp8_bilinear8_h_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear8_v_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear8_hv_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+void ff_put_vp8_bilinear4_h_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear4_v_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+void ff_put_vp8_bilinear4_hv_mmi(uint8_t *dst, ptrdiff_t dststride,
+        uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my);
+
+// loop filter applied to edges between macroblocks
+void ff_vp8_v_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E,
+        int flim_I, int hev_thresh);
+void ff_vp8_h_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E,
+        int flim_I, int hev_thresh);
+void ff_vp8_v_loop_filter8uv_mmi(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
+        int flim_E, int flim_I, int hev_thresh);
+void ff_vp8_h_loop_filter8uv_mmi(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
+        int flim_E, int flim_I, int hev_thresh);
+
+// loop filter applied to inner macroblock edges
+void ff_vp8_v_loop_filter16_inner_mmi(uint8_t *dst, ptrdiff_t stride,
+        int flim_E, int flim_I, int hev_thresh);
+void ff_vp8_h_loop_filter16_inner_mmi(uint8_t *dst, ptrdiff_t stride,
+        int flim_E, int flim_I, int hev_thresh);
+void ff_vp8_v_loop_filter8uv_inner_mmi(uint8_t *dstU, uint8_t *dstV,
+        ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh);
+void ff_vp8_h_loop_filter8uv_inner_mmi(uint8_t *dstU, uint8_t *dstV,
+        ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh);
+
+void ff_vp8_v_loop_filter_simple_mmi(uint8_t *dst, ptrdiff_t stride, int flim);
+void ff_vp8_h_loop_filter_simple_mmi(uint8_t *dst, ptrdiff_t stride, int flim);
+
 #endif  // #ifndef AVCODEC_MIPS_VP8DSP_MIPS_H