diff options
author | gxw <guxiwei-hf@loongson.cn> | 2019-08-07 12:52:00 +0300 |
---|---|---|
committer | Michael Niedermayer <michael@niedermayer.cc> | 2019-08-13 17:48:38 +0300 |
commit | a3e572d96fd1dd6291f6b28e173db858c08ff8d8 (patch) | |
tree | 85807f6ec1442cc362cf8946e67f564c92267e07 /libavutil/mips | |
parent | 8f92eb05e063e6c4d6e36521020620d4e6e1c21d (diff) |
avutil/mips: refine msa macros CLIP_*.
Changing details as following:
1. Remove the local variable 'out_m' in 'CLIP_SH' and store the result in
source vector.
2. Refine the implementation of macro 'CLIP_SH_0_255' and 'CLIP_SW_0_255'.
Performance of VP8 decoding has speed up about 1.1%(from 7.03x to 7.11x).
Performance of H264 decoding has speed up about 0.5%(from 4.35x to 4.37x).
Performance of Theora decoding has speed up about 0.7%(from 5.79x to 5.83x).
3. Remove redundant macro 'CLIP_SH/Wn_0_255_MAX_SATU' and use 'CLIP_SH/Wn_0_255'
instead, because there are no difference in the effect of this two macros.
Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Diffstat (limited to 'libavutil/mips')
-rw-r--r-- | libavutil/mips/generic_macros_msa.h | 119 |
1 files changed, 49 insertions, 70 deletions
diff --git a/libavutil/mips/generic_macros_msa.h b/libavutil/mips/generic_macros_msa.h index 9ac0583765..681d87c458 100644 --- a/libavutil/mips/generic_macros_msa.h +++ b/libavutil/mips/generic_macros_msa.h @@ -933,99 +933,78 @@ /* Description : Clips all halfword elements of input vector between min & max out = ((in) < (min)) ? (min) : (((in) > (max)) ? (max) : (in)) - Arguments : Inputs - in (input vector) - - min (min threshold) - - max (max threshold) - Outputs - out_m (output vector with clipped elements) + Arguments : Inputs - in (input vector) + - min (min threshold) + - max (max threshold) + Outputs - in (output vector with clipped elements) Return Type - signed halfword */ -#define CLIP_SH(in, min, max) \ -( { \ - v8i16 out_m; \ - \ - out_m = __msa_max_s_h((v8i16) min, (v8i16) in); \ - out_m = __msa_min_s_h((v8i16) max, (v8i16) out_m); \ - out_m; \ -} ) +#define CLIP_SH(in, min, max) \ +{ \ + in = __msa_max_s_h((v8i16) min, (v8i16) in); \ + in = __msa_min_s_h((v8i16) max, (v8i16) in); \ +} /* Description : Clips all signed halfword elements of input vector between 0 & 255 - Arguments : Inputs - in (input vector) - Outputs - out_m (output vector with clipped elements) - Return Type - signed halfword + Arguments : Inputs - in (input vector) + Outputs - in (output vector with clipped elements) + Return Type - signed halfwords */ -#define CLIP_SH_0_255(in) \ -( { \ - v8i16 max_m = __msa_ldi_h(255); \ - v8i16 out_m; \ - \ - out_m = __msa_maxi_s_h((v8i16) in, 0); \ - out_m = __msa_min_s_h((v8i16) max_m, (v8i16) out_m); \ - out_m; \ -} ) +#define CLIP_SH_0_255(in) \ +{ \ + in = __msa_maxi_s_h((v8i16) in, 0); \ + in = (v8i16) __msa_sat_u_h((v8u16) in, 7); \ +} + #define CLIP_SH2_0_255(in0, in1) \ { \ - in0 = CLIP_SH_0_255(in0); \ - in1 = CLIP_SH_0_255(in1); \ + CLIP_SH_0_255(in0); \ + CLIP_SH_0_255(in1); \ } + #define CLIP_SH4_0_255(in0, in1, in2, in3) \ { \ CLIP_SH2_0_255(in0, in1); \ CLIP_SH2_0_255(in2, in3); \ } -#define CLIP_SH_0_255_MAX_SATU(in) \ -( { \ - v8i16 out_m; \ - \ - out_m = __msa_maxi_s_h((v8i16) in, 0); \ - out_m = (v8i16) __msa_sat_u_h((v8u16) out_m, 7); \ - out_m; \ -} ) -#define CLIP_SH2_0_255_MAX_SATU(in0, in1) \ -{ \ - in0 = CLIP_SH_0_255_MAX_SATU(in0); \ - in1 = CLIP_SH_0_255_MAX_SATU(in1); \ -} -#define CLIP_SH4_0_255_MAX_SATU(in0, in1, in2, in3) \ -{ \ - CLIP_SH2_0_255_MAX_SATU(in0, in1); \ - CLIP_SH2_0_255_MAX_SATU(in2, in3); \ +#define CLIP_SH8_0_255(in0, in1, in2, in3, \ + in4, in5, in6, in7) \ +{ \ + CLIP_SH4_0_255(in0, in1, in2, in3); \ + CLIP_SH4_0_255(in4, in5, in6, in7); \ } /* Description : Clips all signed word elements of input vector between 0 & 255 - Arguments : Inputs - in (input vector) - Outputs - out_m (output vector with clipped elements) + Arguments : Inputs - in (input vector) + Outputs - in (output vector with clipped elements) Return Type - signed word */ -#define CLIP_SW_0_255(in) \ -( { \ - v4i32 max_m = __msa_ldi_w(255); \ - v4i32 out_m; \ - \ - out_m = __msa_maxi_s_w((v4i32) in, 0); \ - out_m = __msa_min_s_w((v4i32) max_m, (v4i32) out_m); \ - out_m; \ -} ) +#define CLIP_SW_0_255(in) \ +{ \ + in = __msa_maxi_s_w((v4i32) in, 0); \ + in = (v4i32) __msa_sat_u_w((v4u32) in, 7); \ +} -#define CLIP_SW_0_255_MAX_SATU(in) \ -( { \ - v4i32 out_m; \ - \ - out_m = __msa_maxi_s_w((v4i32) in, 0); \ - out_m = (v4i32) __msa_sat_u_w((v4u32) out_m, 7); \ - out_m; \ -} ) -#define CLIP_SW2_0_255_MAX_SATU(in0, in1) \ -{ \ - in0 = CLIP_SW_0_255_MAX_SATU(in0); \ - in1 = CLIP_SW_0_255_MAX_SATU(in1); \ +#define CLIP_SW2_0_255(in0, in1) \ +{ \ + CLIP_SW_0_255(in0); \ + CLIP_SW_0_255(in1); \ } -#define CLIP_SW4_0_255_MAX_SATU(in0, in1, in2, in3) \ -{ \ - CLIP_SW2_0_255_MAX_SATU(in0, in1); \ - CLIP_SW2_0_255_MAX_SATU(in2, in3); \ + +#define CLIP_SW4_0_255(in0, in1, in2, in3) \ +{ \ + CLIP_SW2_0_255(in0, in1); \ + CLIP_SW2_0_255(in2, in3); \ +} + +#define CLIP_SW8_0_255(in0, in1, in2, in3, \ + in4, in5, in6, in7) \ +{ \ + CLIP_SW4_0_255(in0, in1, in2, in3); \ + CLIP_SW4_0_255(in4, in5, in6, in7); \ } /* Description : Addition of 4 signed word elements |