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

@@ -43,7 +43,8 @@ MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o            \
                                              mips/hevc_idct_msa.o          \
                                              mips/hevc_lpf_sao_msa.o       \
                                              mips/hevcpred_msa.o
-MSA-OBJS-$(CONFIG_VP9_DECODER)            += mips/vp9_mc_msa.o
+MSA-OBJS-$(CONFIG_VP9_DECODER)            += mips/vp9_mc_msa.o             \
+                                             mips/vp9_lpf_msa.o
 MSA-OBJS-$(CONFIG_H264DSP)                += mips/h264dsp_msa.o            \
                                              mips/h264idct_msa.o
 MSA-OBJS-$(CONFIG_H264QPEL)               += mips/h264qpel_msa.o

libavcodec/mips/vp9dsp_init_mips.c

@@ -80,9 +80,34 @@ static av_cold void vp9dsp_mc_init_msa(VP9DSPContext *dsp, int bpp)
     }
 }
 
+static av_cold void vp9dsp_loopfilter_init_msa(VP9DSPContext *dsp, int bpp)
+{
+    if (bpp == 8) {
+        dsp->loop_filter_8[0][0] = ff_loop_filter_h_4_8_msa;
+        dsp->loop_filter_8[0][1] = ff_loop_filter_v_4_8_msa;
+        dsp->loop_filter_8[1][0] = ff_loop_filter_h_8_8_msa;
+        dsp->loop_filter_8[1][1] = ff_loop_filter_v_8_8_msa;
+        dsp->loop_filter_8[2][0] = ff_loop_filter_h_16_8_msa;
+        dsp->loop_filter_8[2][1] = ff_loop_filter_v_16_8_msa;
+
+        dsp->loop_filter_16[0] = ff_loop_filter_h_16_16_msa;
+        dsp->loop_filter_16[1] = ff_loop_filter_v_16_16_msa;
+
+        dsp->loop_filter_mix2[0][0][0] = ff_loop_filter_h_44_16_msa;
+        dsp->loop_filter_mix2[0][0][1] = ff_loop_filter_v_44_16_msa;
+        dsp->loop_filter_mix2[0][1][0] = ff_loop_filter_h_48_16_msa;
+        dsp->loop_filter_mix2[0][1][1] = ff_loop_filter_v_48_16_msa;
+        dsp->loop_filter_mix2[1][0][0] = ff_loop_filter_h_84_16_msa;
+        dsp->loop_filter_mix2[1][0][1] = ff_loop_filter_v_84_16_msa;
+        dsp->loop_filter_mix2[1][1][0] = ff_loop_filter_h_88_16_msa;
+        dsp->loop_filter_mix2[1][1][1] = ff_loop_filter_v_88_16_msa;
+    }
+}
+
 static av_cold void vp9dsp_init_msa(VP9DSPContext *dsp, int bpp)
 {
     vp9dsp_mc_init_msa(dsp, bpp);
+    vp9dsp_loopfilter_init_msa(dsp, bpp);
 }
 #endif  // #if HAVE_MSA
 

libavcodec/mips/vp9dsp_mips.h

@@ -88,4 +88,37 @@ VP9_COPY_AVG_MIPS_MSA_FUNC(4);
 #undef VP9_8TAP_MIPS_MSA_FUNC
 #undef VP9_COPY_AVG_MIPS_MSA_FUNC
 
+void ff_loop_filter_h_4_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                              int32_t i, int32_t h);
+void ff_loop_filter_h_8_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                              int32_t i, int32_t h);
+void ff_loop_filter_h_16_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                               int32_t i, int32_t h);
+void ff_loop_filter_v_4_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                              int32_t i, int32_t h);
+void ff_loop_filter_v_8_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                              int32_t i, int32_t h);
+void ff_loop_filter_v_16_8_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                               int32_t i, int32_t h);
+void ff_loop_filter_h_44_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_h_88_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_h_16_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_v_44_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_v_88_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_v_16_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_h_48_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_h_84_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_v_48_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+void ff_loop_filter_v_84_16_msa(uint8_t *dst, ptrdiff_t stride, int32_t e,
+                                int32_t i, int32_t h);
+
 #endif  // #ifndef AVCODEC_MIPS_VP9DSP_MIPS_H

libavutil/mips/generic_macros_msa.h

@@ -24,6 +24,9 @@
 #include <stdint.h>
 #include <msa.h>
 
+#define ALIGNMENT           16
+#define ALLOC_ALIGNED(align) __attribute__ ((aligned((align) << 1)))
+
 #define LD_B(RTYPE, psrc) *((RTYPE *)(psrc))
 #define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
 #define LD_SB(...) LD_B(v16i8, __VA_ARGS__)