diff mbox series

[FFmpeg-devel] avcodec/mathops: Optimize generic mid_pred function

Message ID 20221201064137.1406-1-zhujunxian@oss.cipunited.com
State New
Headers show
Series [FFmpeg-devel] avcodec/mathops: Optimize generic mid_pred function | expand

Checks

Context Check Description
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished

Commit Message

Junxian Zhu Dec. 1, 2022, 6:42 a.m. UTC 
From: Junxian Zhu <zhujunxian@oss.cipunited.com>

Rewrite mid_pred function in generic mathops.h, reduce branch jump to improve performance. And because nowadays new version compiler can compile enough short asmbbely code as handwritting in these function, so remove specified optimized mips inline asmbbely mathops.h.

Signed-off-by: Junxian Zhu <zhujunxian@oss.cipunited.com>
---
 libavcodec/mathops.h      | 20 ++++--------
 libavcodec/mips/mathops.h | 67 ---------------------------------------
 2 files changed, 6 insertions(+), 81 deletions(-)
 delete mode 100644 libavcodec/mips/mathops.h

Comments

Michael Niedermayer Dec. 1, 2022, 11:25 a.m. UTC | #1 
On Thu, Dec 01, 2022 at 02:42:20PM +0800, Junxian Zhu wrote:
> From: Junxian Zhu <zhujunxian@oss.cipunited.com>
> 
> Rewrite mid_pred function in generic mathops.h, reduce branch jump to improve performance. And because nowadays new version compiler can compile enough short asmbbely code as handwritting in these function, so remove specified optimized mips inline asmbbely mathops.h.
> 
> Signed-off-by: Junxian Zhu <zhujunxian@oss.cipunited.com>
> ---
>  libavcodec/mathops.h      | 20 ++++--------
>  libavcodec/mips/mathops.h | 67 ---------------------------------------
>  2 files changed, 6 insertions(+), 81 deletions(-)
>  delete mode 100644 libavcodec/mips/mathops.h
> 
> diff --git a/libavcodec/mathops.h b/libavcodec/mathops.h
> index c89054d6ed..526ffe0eec 100644
> --- a/libavcodec/mathops.h
> +++ b/libavcodec/mathops.h
> @@ -41,8 +41,6 @@ extern const uint8_t ff_zigzag_scan[16+1];
>  #   include "arm/mathops.h"
>  #elif ARCH_AVR32
>  #   include "avr32/mathops.h"
> -#elif ARCH_MIPS
> -#   include "mips/mathops.h"
>  #elif ARCH_PPC
>  #   include "ppc/mathops.h"
>  #elif ARCH_X86
> @@ -98,18 +96,12 @@ static av_always_inline unsigned UMULH(unsigned a, unsigned b){
>  #define mid_pred mid_pred
>  static inline av_const int mid_pred(int a, int b, int c)
>  {
> -    if(a>b){
> -        if(c>b){
> -            if(c>a) b=a;
> -            else    b=c;
> -        }
> -    }else{
> -        if(b>c){
> -            if(c>a) b=c;
> -            else    b=a;
> -        }
> -    }
> -    return b;
> +    int t0,t1,t2,t3;
> +    t0 = (a > b) ? b : a ;
> +    t1 = (a > b) ? a : b ;
> +    t2 = (t0 > c) ? t0 : c;

> +    t3 = (t1 > t2) ? t2 : t1;
> +    return t3; 

t3 intermediate is unneeded
return (t1 > t2) ? t2 : t1;

also benchmarks would be usefull

thx

[...]
Junxian Zhu Dec. 4, 2022, 1:06 p.m. UTC | #2 
Compile with generic mid_pred function written in the old way on the mips platform, and objdump its .o file, it will show like this, with more instructions and branch jump. ```code mid_pred(int, int,
diff mbox series

Patch

diff --git a/libavcodec/mathops.h b/libavcodec/mathops.h
index c89054d6ed..526ffe0eec 100644
--- a/libavcodec/mathops.h
+++ b/libavcodec/mathops.h
@@ -41,8 +41,6 @@  extern const uint8_t ff_zigzag_scan[16+1];
 #   include "arm/mathops.h"
 #elif ARCH_AVR32
 #   include "avr32/mathops.h"
-#elif ARCH_MIPS
-#   include "mips/mathops.h"
 #elif ARCH_PPC
 #   include "ppc/mathops.h"
 #elif ARCH_X86
@@ -98,18 +96,12 @@  static av_always_inline unsigned UMULH(unsigned a, unsigned b){
 #define mid_pred mid_pred
 static inline av_const int mid_pred(int a, int b, int c)
 {
-    if(a>b){
-        if(c>b){
-            if(c>a) b=a;
-            else    b=c;
-        }
-    }else{
-        if(b>c){
-            if(c>a) b=c;
-            else    b=a;
-        }
-    }
-    return b;
+    int t0,t1,t2,t3;
+    t0 = (a > b) ? b : a ;
+    t1 = (a > b) ? a : b ;
+    t2 = (t0 > c) ? t0 : c;
+    t3 = (t1 > t2) ? t2 : t1;
+    return t3; 
 }
 #endif
 
diff --git a/libavcodec/mips/mathops.h b/libavcodec/mips/mathops.h
deleted file mode 100644
index bb9dc8375a..0000000000
--- a/libavcodec/mips/mathops.h
+++ /dev/null
@@ -1,67 +0,0 @@ 
-/*
- * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
- * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_MIPS_MATHOPS_H
-#define AVCODEC_MIPS_MATHOPS_H
-
-#include <stdint.h>
-#include "config.h"
-#include "libavutil/common.h"
-
-#if HAVE_INLINE_ASM
-
-#if HAVE_LOONGSON3
-
-#define MULH MULH
-static inline av_const int MULH(int a, int b)
-{
-    int c;
-    __asm__ ("dmult %1, %2      \n\t"
-             "mflo %0           \n\t"
-             "dsrl %0, %0, 32   \n\t"
-             : "=r"(c)
-             : "r"(a),"r"(b)
-             : "hi", "lo");
-    return c;
-}
-
-#define mid_pred mid_pred
-static inline av_const int mid_pred(int a, int b, int c)
-{
-    int t = b;
-    __asm__ ("sgt $8, %1, %2    \n\t"
-             "movn %0, %1, $8   \n\t"
-             "movn %1, %2, $8   \n\t"
-             "sgt $8, %1, %3    \n\t"
-             "movz %1, %3, $8   \n\t"
-             "sgt $8, %0, %1    \n\t"
-             "movn %0, %1, $8   \n\t"
-             : "+&r"(t),"+&r"(a)
-             : "r"(b),"r"(c)
-             : "$8");
-    return t;
-}
-
-#endif /* HAVE_LOONGSON3 */
-
-#endif /* HAVE_INLINE_ASM */
-
-#endif /* AVCODEC_MIPS_MATHOPS_H */