[Ffmpeg-devel] yet another silly int vs. float benchmark

[Michael Niedermayer]

Hi

On Saturday 21 May 2005 18:06, Michael Niedermayer wrote:
> Hi
>
> heres another benchmark proggy, advantages over the others
> 1. pure c
> 2. ~40 lines of code, can be easily done in less i know ...
> 3. tries to test both the case where each instruction depends upon the
> previous one and where the instructions are a little more independant

500mhz pentium 3:
100 ; needed    32 cycles ->    32 cycles per operation
100 iv[0]+=iv[1];iv[1]+=iv[0]; needed   233 cycles ->   116 cycles per operation
100 iv[0]*=iv[1];iv[1]*=iv[0]; needed   832 cycles ->   416 cycles per operation
100 fv[0]+=fv[1];fv[1]+=fv[0]; needed   884 cycles ->   442 cycles per operation
100 fv[0]*=fv[1];fv[1]*=fv[0]; needed  1032 cycles ->   516 cycles per operation
100 iv[0]+=iv[1];iv[1]+=iv[2];iv[2]+=iv[3];iv[3]+=iv[4];iv[4]+=iv[5]; needed   301 cycles ->    60 cycles per operation
100 iv[0]*=iv[1];iv[1]*=iv[2];iv[2]*=iv[3];iv[3]*=iv[4];iv[4]*=iv[5]; needed   555 cycles ->   111 cycles per operation
100 fv[0]+=fv[1];fv[1]+=fv[2];fv[2]+=fv[3];fv[3]+=fv[4];fv[4]+=fv[5]; needed   542 cycles ->   108 cycles per operation
100 fv[0]*=fv[1];fv[1]*=fv[2];fv[2]*=fv[3];fv[3]*=fv[4];fv[4]*=fv[5]; needed  1039 cycles ->   207 cycles per operation

800mhz duron
100 ; needed    11 cycles ->    11 cycles per operation
100 iv[0]+=iv[1];iv[1]+=iv[0]; needed   204 cycles ->   102 cycles per operation
100 iv[0]*=iv[1];iv[1]*=iv[0]; needed   796 cycles ->   398 cycles per operation
100 fv[0]+=fv[1];fv[1]+=fv[0]; needed   720 cycles ->   360 cycles per operation
100 fv[0]*=fv[1];fv[1]*=fv[0]; needed   720 cycles ->   360 cycles per operation
100 iv[0]+=iv[1];iv[1]+=iv[2];iv[2]+=iv[3];iv[3]+=iv[4];iv[4]+=iv[5]; needed   211 cycles ->    42 cycles per operation
100 iv[0]*=iv[1];iv[1]*=iv[2];iv[2]*=iv[3];iv[3]*=iv[4];iv[4]*=iv[5]; needed  1014 cycles ->   202 cycles per operation
100 fv[0]+=fv[1];fv[1]+=fv[2];fv[2]+=fv[3];fv[3]+=fv[4];fv[4]+=fv[5]; needed   680 cycles ->   136 cycles per operation
100 fv[0]*=fv[1];fv[1]*=fv[2];fv[2]*=fv[3];fv[3]*=fv[4];fv[4]*=fv[5]; needed   509 cycles ->   101 cycles per operation

both compiled with gcc -O3 -fomit-frame-pointer sillybench.c -o sillybench

-- 
Michael