avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for VP9 intra functions

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/mips/Makefile

@@ -45,7 +45,8 @@ MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o            \
                                              mips/hevcpred_msa.o
 MSA-OBJS-$(CONFIG_VP9_DECODER)            += mips/vp9_mc_msa.o             \
                                              mips/vp9_lpf_msa.o            \
-                                             mips/vp9_idct_msa.o
+                                             mips/vp9_idct_msa.o           \
+                                             mips/vp9_intra_msa.o
 MSA-OBJS-$(CONFIG_H264DSP)                += mips/h264dsp_msa.o            \
                                              mips/h264idct_msa.o
 MSA-OBJS-$(CONFIG_H264QPEL)               += mips/h264qpel_msa.o

libavcodec/mips/vp9dsp_init_mips.c

@@ -24,6 +24,36 @@
 #include "vp9dsp_mips.h"
 
 #if HAVE_MSA
+static av_cold void vp9dsp_intrapred_init_msa(VP9DSPContext *dsp, int bpp)
+{
+    if (bpp == 8) {
+#define init_intra_pred_msa(tx, sz)                             \
+    dsp->intra_pred[tx][VERT_PRED]    = ff_vert_##sz##_msa;     \
+    dsp->intra_pred[tx][HOR_PRED]     = ff_hor_##sz##_msa;      \
+    dsp->intra_pred[tx][DC_PRED]      = ff_dc_##sz##_msa;       \
+    dsp->intra_pred[tx][LEFT_DC_PRED] = ff_dc_left_##sz##_msa;  \
+    dsp->intra_pred[tx][TOP_DC_PRED]  = ff_dc_top_##sz##_msa;   \
+    dsp->intra_pred[tx][DC_128_PRED]  = ff_dc_128_##sz##_msa;   \
+    dsp->intra_pred[tx][DC_127_PRED]  = ff_dc_127_##sz##_msa;   \
+    dsp->intra_pred[tx][DC_129_PRED]  = ff_dc_129_##sz##_msa;   \
+    dsp->intra_pred[tx][TM_VP8_PRED]  = ff_tm_##sz##_msa;       \
+
+    init_intra_pred_msa(TX_16X16, 16x16);
+    init_intra_pred_msa(TX_32X32, 32x32);
+#undef init_intra_pred_msa
+
+#define init_intra_pred_msa(tx, sz)                             \
+    dsp->intra_pred[tx][DC_PRED]      = ff_dc_##sz##_msa;       \
+    dsp->intra_pred[tx][LEFT_DC_PRED] = ff_dc_left_##sz##_msa;  \
+    dsp->intra_pred[tx][TOP_DC_PRED]  = ff_dc_top_##sz##_msa;   \
+    dsp->intra_pred[tx][TM_VP8_PRED]  = ff_tm_##sz##_msa;       \
+
+    init_intra_pred_msa(TX_4X4, 4x4);
+    init_intra_pred_msa(TX_8X8, 8x8);
+#undef init_intra_pred_msa
+    }
+}
+
 static av_cold void vp9dsp_itxfm_init_msa(VP9DSPContext *dsp, int bpp)
 {
     if (bpp == 8) {
@@ -129,6 +159,7 @@ static av_cold void vp9dsp_loopfilter_init_msa(VP9DSPContext *dsp, int bpp)
 
 static av_cold void vp9dsp_init_msa(VP9DSPContext *dsp, int bpp)
 {
+    vp9dsp_intrapred_init_msa(dsp, bpp);
     vp9dsp_itxfm_init_msa(dsp, bpp);
     vp9dsp_mc_init_msa(dsp, bpp);
     vp9dsp_loopfilter_init_msa(dsp, bpp);

libavcodec/mips/vp9dsp_mips.h

@@ -149,4 +149,57 @@ void ff_idct_iadst_16x16_add_msa(uint8_t *pu8Dest, ptrdiff_t stride,
 void ff_iwht_iwht_4x4_add_msa(uint8_t *dst, ptrdiff_t stride,
                               int16_t *block, int eob);
 
+void ff_vert_16x16_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                       const uint8_t *top);
+void ff_vert_32x32_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                       const uint8_t *top);
+void ff_hor_16x16_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                      const uint8_t *top);
+void ff_hor_32x32_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                      const uint8_t *top);
+void ff_dc_4x4_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                   const uint8_t *top);
+void ff_dc_8x8_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                   const uint8_t *top);
+void ff_dc_16x16_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                     const uint8_t *top);
+void ff_dc_32x32_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                     const uint8_t *top);
+void ff_dc_left_4x4_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                        const uint8_t *top);
+void ff_dc_left_8x8_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                        const uint8_t *top);
+void ff_dc_left_16x16_msa(uint8_t *dst, ptrdiff_t stride,
+                          const uint8_t *left, const uint8_t *top);
+void ff_dc_left_32x32_msa(uint8_t *dst, ptrdiff_t stride,
+                          const uint8_t *left, const uint8_t *top);
+void ff_dc_top_4x4_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                       const uint8_t *top);
+void ff_dc_top_8x8_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                       const uint8_t *top);
+void ff_dc_top_16x16_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_top_32x32_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_128_16x16_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_128_32x32_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_127_16x16_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_127_32x32_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_129_16x16_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_dc_129_32x32_msa(uint8_t *dst, ptrdiff_t stride,
+                         const uint8_t *left, const uint8_t *top);
+void ff_tm_4x4_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                   const uint8_t *top);
+void ff_tm_8x8_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                   const uint8_t *top);
+void ff_tm_16x16_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                     const uint8_t *top);
+void ff_tm_32x32_msa(uint8_t *dst, ptrdiff_t stride, const uint8_t *left,
+                     const uint8_t *top);
+
 #endif  // #ifndef AVCODEC_MIPS_VP9DSP_MIPS_H