diff options
| author | Shivraj Patil <shivraj.patil@imgtec.com> | 2015-06-29 20:57:13 +0530 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2015-07-06 18:25:01 +0200 |
| commit | 2f3f98af2b3215b7f3ab302275a0b3b4acaf84a5 (patch) | |
| tree | 820512a13f2a87942a1f1b4f4b6a76e7c880e8ec /libavutil/mips | |
| parent | 115c14c3b6644181e0331bfbda7f86c61e939a7d (diff) | |
| download | ffmpeg-2f3f98af2b3215b7f3ab302275a0b3b4acaf84a5.tar.gz | |
avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for mpegvideoencdsp functions
This patch adds MSA (MIPS-SIMD-Arch) optimizations for mpegvideoencdsp functions in new file mpegvideoencdsp_msa.c
Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavutil/mips')
| -rw-r--r-- | libavutil/mips/generic_macros_msa.h | 34 |
1 files changed, 34 insertions, 0 deletions
diff --git a/libavutil/mips/generic_macros_msa.h b/libavutil/mips/generic_macros_msa.h index 3f44ec955c..b1e62b667d 100644 --- a/libavutil/mips/generic_macros_msa.h +++ b/libavutil/mips/generic_macros_msa.h @@ -1262,6 +1262,15 @@ } #define HADD_UB3_UH(...) HADD_UB3(v8u16, __VA_ARGS__) +#define HADD_UB4(RTYPE, in0, in1, in2, in3, out0, out1, out2, out3) \ +{ \ + HADD_UB2(RTYPE, in0, in1, out0, out1); \ + HADD_UB2(RTYPE, in2, in3, out2, out3); \ +} +#define HADD_UB4_UB(...) HADD_UB4(v16u8, __VA_ARGS__) +#define HADD_UB4_UH(...) HADD_UB4(v8u16, __VA_ARGS__) +#define HADD_UB4_SH(...) HADD_UB4(v8i16, __VA_ARGS__) + /* Description : Horizontal subtraction of unsigned byte vector elements Arguments : Inputs - in0, in1 Outputs - out0, out1 @@ -1771,6 +1780,15 @@ #define SPLATI_H2_SB(...) SPLATI_H2(v16i8, __VA_ARGS__) #define SPLATI_H2_SH(...) SPLATI_H2(v8i16, __VA_ARGS__) +#define SPLATI_H3(RTYPE, in, idx0, idx1, idx2, \ + out0, out1, out2) \ +{ \ + SPLATI_H2(RTYPE, in, idx0, idx1, out0, out1); \ + out2 = (RTYPE) __msa_splati_h((v8i16) in, idx2); \ +} +#define SPLATI_H3_SB(...) SPLATI_H3(v16i8, __VA_ARGS__) +#define SPLATI_H3_SH(...) SPLATI_H3(v8i16, __VA_ARGS__) + #define SPLATI_H4(RTYPE, in, idx0, idx1, idx2, idx3, \ out0, out1, out2, out3) \ { \ @@ -2823,4 +2841,20 @@ tmp_m = __msa_pckev_b((v16i8) in1, (v16i8) in0); \ ST_SB(tmp_m, (pdst)); \ } + +/* Description : Horizontal 2 tap filter kernel code + Arguments : Inputs - in0, in1, mask, coeff, shift +*/ +#define HORIZ_2TAP_FILT_UH(in0, in1, mask, coeff, shift) \ +( { \ + v16i8 tmp0_m; \ + v8u16 tmp1_m; \ + \ + tmp0_m = __msa_vshf_b((v16i8) mask, (v16i8) in1, (v16i8) in0); \ + tmp1_m = __msa_dotp_u_h((v16u8) tmp0_m, (v16u8) coeff); \ + tmp1_m = (v8u16) __msa_srari_h((v8i16) tmp1_m, shift); \ + tmp1_m = __msa_sat_u_h(tmp1_m, shift); \ + \ + tmp1_m; \ +} ) #endif /* AVUTIL_MIPS_GENERIC_MACROS_MSA_H */ |