Re: MPE 64-bit bench
- From: brian.fox@xxxxxxxxxx
- Date: Mon, 25 Jun 2007 17:11:36 -0700
On Jun 25, 5:30 pm, m...@xxxxxx (Marcel Hendrix) wrote:
Here are the results for a slightly adapted MPE benchmark
(made 64 bit clean). The source code passed to the three
compilers is the same, the hardware is also equal.
Compilers:
----------
Gforth 0.6.9-20070604 (fast, 64bit, Linux)
eForth64 7.2 (64bit, Linux)
iForth 2.0 (32bit, Linux)
System:
-------
AMD X2 3 GHz (not tuned), 2 GB DDR2, Linux.
(BIOS does not recognize CPU).
Benchmark output:
-----------------
This system's primitives using no extensions
gForth eForth iForth
Test time excluding overhead ms ms ms
DO LOOP 3 2 2
+ 4 2 2
M+ 5 1 2
* 3 2 2
/ 29 2 16
M* 14 2 2
M/ 31 1 16
/MOD 29 2 16
*/ 37 1 17
ARRAY fill 11 2 3
Total: 166 17 78
This system's O/S interface using no extensions
Test time excluding overhead ms ms ms
Win32 API: SendMessage 0 0 0
Win32 API: GetTickCount 119 117 124
Total: 119 117 124
This system's application performance using no extensions
Test time excluding overhead ms ms ms
Eratosthenes sieve 1899 Primes 344 71 61
Fibonacci recursion ( 35 -> 9227465 ) 476 161 176
Hoare's quick sort (reverse order) 231 75 44
Generate random numbers (1024 kb array) 219 49 49
LZ77 Comp. (400 kb Random Data Mem>Mem) 463 96 81
Dhrystone (integer) 375 95 65
Total: 2108 547 476
About 15% difference between eForth and iForth. Maybe that is
just because handling 32 bits is faster than handling 64.
-marcel
Does this indicate that we should consider creating an optional
replacement kernel for gForth for intel machines?
Brian Fox
.
- References:
- MPE 64-bit bench
- From: Marcel Hendrix
- MPE 64-bit bench
- Prev by Date: ISO/IEC 15145:1997 Forth Programming Language
- Next by Date: Re: New Holon for Windows
- Previous by thread: MPE 64-bit bench
- Next by thread: Re: MPE 64-bit bench
- Index(es):
Relevant Pages
|
