Click Here
home features news forums classifieds faqs links search
6056 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
67 crawler(s) on-line.
 15 guest(s) on-line.
 1 member(s) on-line.


 cdimauro

You are an anonymous user.
Register Now!
 cdimauro:  2 mins ago
 Karlos:  6 mins ago
 Hammer:  11 mins ago
 Paul:  53 mins ago
 thomas:  58 mins ago
 terminills:  1 hr 3 mins ago
 Yssing:  1 hr 8 mins ago
 Hypex:  1 hr 16 mins ago
 sananaman:  1 hr 40 mins ago
 zipper:  1 hr 41 mins ago

/  Forum Index
   /  Classic Amiga Hardware
      /  Speedtest for Hardwares that are able to run Amiga OS 68k Software
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 )
PosterThread
pavlor 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 2-Sep-2010 19:22:54
#81 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9370
From: Unknown

@bernd_afa

Quote:
so again here, this test run only in cache and its no real world situation, because in real world 2. level cache give speedup


No benchmark simulates real world situation...

Some applications really benefit from big L2 cache (eg. emulators), some not.

Quote:
SAM 440 ep or a 604e without second level cache or a G3 750FX with 512 kb 2. Level cache give near same performance.


SAM is significantly slower, but not that slow as pointed in your selected benchmark(s). Sure, there are applications demanding more speed (and L2 cache). However, 440EP 666 MHz is fast enough to emulate A500 (with 440EP optimized UAE) and even to run DosBox with 3000 CPU cycles.

I can give also some Efika numbers for comparison:
OGR-NG (2.9101-507c): 3.2 milions nodes/sec
1500 CPU cycles in DosBox (the sound is jerky with more cycles)

There are some other Efika benchmarks eg. here

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 3-Sep-2010 17:35:26
#82 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

@pavlor
>No benchmark simulates real world situation...
>Some applications really benefit from big L2 cache (eg. emulators), some not.

but the mp3 encoding performance is a practice test and show simular results to this bench.do you think 40ep FPU is same fast as a 604 G3 or G4 FPU ?

Problem is also, i know no practice tests that need only integer performance.
all thats slow on todays systems need FPU.

 Status: Offline
Profile     Report this post  
pavlor 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 3-Sep-2010 17:46:09
#83 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9370
From: Unknown

@bernd_afa

You can also compare benchmark (eg. automotive or consumer) results on the www.eembc.org page (405, 440, 603e, G3, G4, e500 and e600 cores). Benchmarks contain audio/video encoding, image conversion etc.

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 4-Sep-2010 11:56:02
#84 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

@pavlor
>You can also compare benchmark (eg. automotive or consumer) results on the >www.eembc.org page (405, 440, 603e, G3, G4, e500 and e600 cores). Benchmarks >contain audio/video encoding, image conversion etc.

The eembc benches for PPC are not compile with GCC, so in Fact the benchmarks are not usable to see what CPU is the best.

A good compiler can rearange the code or preload data etc so a bad designed CPU is not lots slower.

remember Powerr 6 was a in order design, its was not too much slower, due IBM enhance the compiler for that, but IBM change in Power 7 now to out of order design, because they see the compilers are not able to do the job as good as a out of order design.

maybe you can look what performance a power 6 get with a GCC.

For amiga PPC systems only GCC compiles count.

the green Hills Compilers are not available for any amiga OS PPC system, but the green Hill Compiler are faster and maybe support the special 440ep Instructions.read here.

"""
http://www.ghs.com/news/20050711_eembc.html

“When you combine the performance of AMCC’s 440EP processor with the efficiency of Green Hills compilers, the results are impressive,” said Markus Levy, EEMBC president. “These scores will provide designers with additional objective data to help in selecting the optimal processor for their given application.”
""""

Last edited by bernd_afa on 04-Sep-2010 at 12:00 PM.

 Status: Offline
Profile     Report this post  
pavlor 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 4-Sep-2010 12:41:44
#85 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9370
From: Unknown

@bernd_afa

Quote:
and maybe support the special 440ep Instructions.read here.


Special 440EP instructions are supported by the last OS4 SDK.

And there are AMCC 440GP 500 MHz results with GNU GCC 3.0.4, difference is minimal...

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 4-Sep-2010 16:49:12
#86 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

@pavlor
>And there are AMCC 440GP 500 MHz results with GNU GCC 3.0.4, difference is >minimal...

have you link ?

>Special 440EP instructions are supported by the last OS4 SDK.

GCC do not use this commands, only when you write a program in asm you can use

GCC only can use ISSE2 instead of FPU, but it cant use altivec in portable C code or AMCC instructions.

 Status: Offline
Profile     Report this post  
minator 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 5-Sep-2010 1:41:58
#87 ]
Cult Member
Joined: 23-Mar-2004
Posts: 989
From: Cambridge

@bernd_afa

Quote:
@minator
>2) E-UAE on the Mac sucks really, really badly (I don't think E-UAE includes the 68K JIT).

Yes your results show that no JIT is used.

only when a JIT is used, then it measure really CPu speed.

The test use asm instructions that every CPU have.

the JIT of the target system need only translate the instruction to a simular instruction of the target CPU.

thats a simple process and every JIT do in same way.


er, no.
We are talking about multiple different JITs written by different people, they could (and probably are) done in wildly different ways.

Your test is a test of JIT/Emulator + CPU.

Quote:
So this make sure, that no diffrent results due to diffrent compilers can happen and so can see how good the CPU out of order execution work.


Out of order execution generally isn't much use for FFTs. There are plenty of processors which are pretty much designed to run FFTs and none of these are out-of-order.

---

BTW some new results for you:

MacBook Pro Core i7 2.66GHz (+ turbo boost)

FFT float 1859ms
FFT integer 2211ms
FFT assembly 863ms

These results are a *lot* better and put it in the middle somewhere.
The floating point is beats the Mac mini / MOS machine but the int results are closer to the 1GHz Pegasos.

One difference I did notice it it sent a CPU core to 100%.
I did tweak the settings a bit so it looks like UAE setting can have *major* impacts on the results.

Last edited by minator on 05-Sep-2010 at 02:13 AM.

_________________
Whyzzat?

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 5-Sep-2010 11:03:09
#88 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

@minator
>er, no.
>We are talking about multiple different JITs written by different people, they could >(and probably are) done in wildly different ways.

that doesnt matter, a JIT is same as a cross assembler but it work on realtime.

http://encyclopedia2.thefreedictionary.com/cross-assembler

All RISC CPu can only do few things.(modern X86 is too a RISC CPU, but it translate the X86 instructions so compatible X86 CISC Code can use to make assembler code shorter and have better overview of asm code for compiler developers or OS developers.Here asm programming is still need.In FACT a modern X86 CPU is a RISC CPU with a unknown RISC Instruction set and there is a JIT in the X86 core that translate the X86 instructions to the RISC Core.translating can done paralell, same as you can translate a book in diffrent languages by diffrent translaters to be faster.

only the book reader need keep the order of the text to understand the book


1.put a long, word, byte, float from memory to a register.
2. put a long,word, byte,float from register to memory
3. add, subtract multiply, dividide teh values of the registers.
4. compare values in a register with other registers to detect if a value is > < =
5. jump to a memory address if the compare is < > =
the next is not use in this benchmark.
6. jump to a subroutine frequently

so you see all CPU do only few things, only whats diffrent between CPU is the command Hex syntax.

programs are store in memory and every of the above instructions, have a number to show the CPU what they should do.

And the number of the instruction is diffrent for each of the command type on a diffrent CPU.

But a crossassembler or a JIT do the job and translate the number from a diffrent CPU to another CPU.

amiblitz is written in that way, that all available JIT for PPC can do that in same instruction counts as the 68k code.

in short, when the program written for 68k need 100 asm instructions, when the MOS or OS4 JIT translate it, it use also 100 asm instructions.only on UAE JIT on X86 it use more instructions, because on instruction type 5 (jump to a memory address if the compare is < > =)there is check if a chipset event happen and if so, it jump to chipset.

Last edited by bernd_afa on 05-Sep-2010 at 11:06 AM.

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 5-Sep-2010 11:44:31
#89 ]
Elite Member
Joined: 9-Jun-2004
Posts: 12197
From: Norway

@bernd_afa

Quote:
So you see all CPU do only few things, only what’s different between CPU is the command Hex syntax.


Only some cpu’s have the push and pop instruction, the PowerPC does this by using indirect addressing, where the stack pointer is normal cpu register instead.

The power/powerpc also has some instructions to read and write reverse byte order.

But in general you are right.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 5-Sep-2010 12:24:35
#90 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

@NutsAboutAmiga
>Only some cpu’s have the push and pop instruction, the PowerPC does this by >using indirect addressing, where the stack pointer is normal cpu register instead.

The push and pop instructions can too replace with 1 PPC instruction.
but the benchmark use no function call in the speed test loop, so there is no push and pop need.




 Status: Offline
Profile     Report this post  
stachu100 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 15-Sep-2010 23:30:08
#91 ]
New Member
Joined: 26-Jan-2010
Posts: 1
From: Unknown

@bernd_afa

Quote:
the PPC 604e seem the best PPC ever build in FPU performance /MHZ, sad it do not run on higher clockrates as 375 MHZ


Now it is:
Cyberstorm PPC 400MHz, 66.7MHz bus speed, 70ns ram, OS4:

FFTDemo
Speed test for FFT + iFFT: (float)
time needed 7478ms
Speed test for FFT + iFFT: (integer)
time needed 6419ms
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 3045ms

http://www.youtube.com/watch?v=gxsdHQN6K1Q

 Status: Offline
Profile     Report this post  
delshay 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 5-Apr-2013 23:04:32
#92 ]
Regular Member
Joined: 20-Sep-2008
Posts: 447
From: Unknown

Blizzard PPC 375Mhz

Speed test for FFT + iFFT: (float)
time needed 14472ms for 413696 samples, => .324103832244873x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 12381ms for 413696 samples, => .378841042518615x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 10383ms for 413696 samples, => .451741397380828x speed @44100Hz/stereo

Last edited by delshay on 06-Apr-2013 at 05:04 AM.

_________________
The Machine: Bride Of The Pin•Bot by Williams Electronics

 Status: Offline
Profile     Report this post  
bernd_afa 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 23-Jul-2017 20:06:00
#93 ]
Cult Member
Joined: 14-Apr-2006
Posts: 829
From: Unknown

now there is X5000 here. maybe somebody with a X5000 can do the test, so it can add to the list in 1. page. in 1. page is testprogram and how it can test

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: Speedtest for Hardwares that are able to run Amiga OS 68k Software
Posted on 23-Jul-2017 23:52:26
#94 ]
Elite Member
Joined: 9-Jun-2004
Posts: 12197
From: Norway

@bernd_afa

Maybe you can have the C sources of that, so we can compile some native versions of it?

Last edited by NutsAboutAmiga on 24-Jul-2017 at 01:11 AM.
Last edited by NutsAboutAmiga on 24-Jul-2017 at 12:18 AM.
Last edited by NutsAboutAmiga on 24-Jul-2017 at 12:17 AM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle