This forum uses cookies
This forum makes use of cookies to store your login information if you are registered, and your last visit if you are not. Cookies are small text documents stored on your computer; the cookies set by this forum can only be used on this website and pose no security risk. Cookies on this forum also track the specific topics you have read and when you last read them. Please confirm whether you accept or reject these cookies being set.

A cookie will be stored in your browser regardless of choice to prevent you being asked this question again. You will be able to change your cookie settings at any time using the link in the footer.

Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
JPCSP Official Requirements?
#11
Strange, I get a completely different performance when run on my Intel core @2.66GHz with Java 7 32-bit:
Code:
Empty loop               :  199 ms ( 457%) @ 222 MHz
Simple loop              :  287 ms ( 397%) @ 222 MHz
read32                   :  359 ms ( 338%) @ 222 MHz
read16                   :  352 ms ( 291%) @ 222 MHz
read8                    :  394 ms ( 251%) @ 222 MHz
write32                  :  347 ms ( 354%) @ 222 MHz
write16                  :  402 ms ( 239%) @ 222 MHz
write8                   :  475 ms ( 212%) @ 222 MHz
Function call no params  :  561 ms ( 190%) @ 222 MHz
Function call with params:  380 ms ( 359%) @ 222 MHz
FPU add.s                :  195 ms ( 350%) @ 222 MHz
FPU mul.s                :  193 ms ( 353%) @ 222 MHz
VFPU vadd.s              :  334 ms ( 245%) @ 222 MHz
VFPU vadd.p              :  441 ms ( 186%) @ 222 MHz
VFPU vadd.t              :  573 ms ( 143%) @ 222 MHz
VFPU vadd.q              :  700 ms ( 117%) @ 222 MHz
VFPU vadd.q sequence     : 1307 ms (  52%) @ 222 MHz
LWC1                     :  422 ms ( 288%) @ 222 MHz
SWC1                     :  394 ms ( 312%) @ 222 MHz
memcpy (native)          :   70 ms (1237%) @ 222 MHz
memset (native)          :   90 ms (1191%) @ 222 MHz
strcpy (native)          :  175 ms ( 778%) @ 222 MHz
memcpy (non-native)      :  401 ms ( 198%) @ 222 MHz
memset (non-native)      :  327 ms ( 235%) @ 222 MHz
strcpy (non-native)      :  404 ms ( 209%) @ 222 MHz
Overall performance index: 359%

Do you get different values depending on
- java 6 or 7
- 32-bit or 64-bit
Always include a complete log file at INFO level in your reports. Thanks! How to post a log
Reply
#12
If you've read this post before I scrapped old results...., I had the profiler.txt enabled and that was what was causing the strange values.

New results (with profiler.txt off) :p
I posted screenshots since compilerPerf.log doesn't generate the 'overall performance index' value for some reason.

Java 6 (1.6.0_35-b10) @ 64bit // 1392% (Overall performance Index)
   

Java 7 (1.7.0_07-b11) @ 64bit // 1407%
   

Java 8 (1.8.0-ea-b54) @ 64bit // 1460%
   
---------------------------------------------
Java 6 (1.6.0_31-b05) @ 32bit // 1072%
   

Java 7 (1.7.0_07-b11) @ 32bit // 891%
   

Java 8 (not installed) -

[Image: 2541114.png]
Reply
#13
Yes, i got different values according the Java version and the JPCSP architecture.

First, gotta say that i perform test with CPU normal stage and CPU OverCloked; all pictures on the right are OC, the ones on the left Normal CPU Stage. I use JPCSP r2767 to perform the tests.

K10_States:
   

Java 7/6 x64:
   

Java 7/6 x86:
   

It looks like my CPU is not compatible with JPCSP Sad. Best result was Java_7x64_OC_CPU getting only 60% overall.
[Image: montcer.png]
Reply
#14
(10-12-2012, 08:23 PM)montcer9012 Wrote: Yes, i got different values according the Java version and the JPCSP architecture.

First, gotta say that i perform test with CPU normal stage and CPU OverCloked; all pictures on the right are OC, the ones on the left Normal CPU Stage. I use JPCSP r2767 to perform the tests.

K10_States:


Java 7/6 x64:


Java 7/6 x86:


It looks like my CPU is not compatible with JPCSP Sad. Best result was Java_7x64_OC_CPU getting only 60% overall.
This is not normal that you get such poor results...
Could you post a log file at INFO level? This will show your options...
Are you using the standard start script? or special java command line options?
Is it a Windows platform?
Always include a complete log file at INFO level in your reports. Thanks! How to post a log
Reply
#15
(10-12-2012, 08:42 PM)gid15 Wrote: This is not normal that you get such poor results...
Could you post a log file at INFO level? This will show your options...
Are you using the standard start script? or special java command line options?
Is it a Windows platform?
Yes, i am using the normal start script; none special command line option.
Yes, is Windows 7x64 Ultimate SP1.

NEWS: The options i was using on past result was r2767 default; since this test is about the compiler i change the MaximumCompilerMethod from 3000 to 50/100 or even 500 because it gave me better results on games and look the differences: (Compiler Method set to 100)
Without OC:
   
With OC attached on last post:
   

Still being poor if i compare with yours, but at least is something.

EDIT: Well, i don't know what happen here, but since i start change the MMSize, if i return to 3000 i get similar results (250~300 on overall). The only thing i change on JPCSP options since last post was the Maximum Compiler method Confused

Confirm: When i extract the JPCSP files to any folder, the first and next test i perform without change the CompilerMethod lower than 3000 will gave me the poor results attached before; however, once i change the CompilerMethodSize does not matter if i return to 3000, i always get an overall between 200%~300%. Maybe Windows Registry have something to do here, i don't know.

EDIT 2: If i perform the test with the CPU OC and on earlier rev 2758 i got an overall between 500%~560%. I guess the test added on r2765 are exhaustive.
   


Attached Files
.html   r2767x64 - CompilerTestMMSize_3000.html (Size: 25.75 KB / Downloads: 162)
[Image: montcer.png]
Reply
#16
here are my tests updated with latest compilerPerf
(tested using jpcsp r2768)
java 32 6-7 (32bit jpcsp build)
java 64 6-8

I used maximum method size 50

I might update the post again with another test set at 3000


Attached Files
.rar   compilerPerf.log (all tests).rar (Size: 4.11 KB / Downloads: 537)
.rar   compilerPerf.log (all tests @ 3000 Method).rar (Size: 4.16 KB / Downloads: 134)
Reply
#17
(10-12-2012, 08:50 PM)montcer9012 Wrote: Confirm: When i extract the JPCSP files to any folder, the first and next test i perform without change the CompilerMethod lower than 3000 will gave me the poor results attached before; however, once i change the CompilerMethodSize does not matter if i return to 3000, i always get an overall between 200%~300%. Maybe Windows Registry have something to do here, i don't know.
I was also able to reproduce the problem: when no Settings.properties was defined, the profiler was enabled by default! Going once into the configuration (even without changing anything) and saving it was disabling the profiler. Could it be the reason why some users were reporting poor performance? This is the case if they never opened the configuration.
Anyway, good finding! Now fixed in r2769 Smile
Always include a complete log file at INFO level in your reports. Thanks! How to post a log
Reply
#18
Good finding for you! You were the one who found the bug hahahaha. Anyway, do you think that 3 GHz is the average to use JPCSP, or in some future it will be more compatible?
[Image: montcer.png]
Reply
#19
(10-13-2012, 07:34 AM)montcer9012 Wrote: Good finding for you! You were the one who found the bug hahahaha. Anyway, do you think that 3 GHz is the average to use JPCSP, or in some future it will be more compatible?
It depends on the game, some are more CPU intensive than others...
Posting profiler information might help me to improve the CPU performance.
Always include a complete log file at INFO level in your reports. Thanks! How to post a log
Reply
#20
(10-13-2012, 07:41 AM)gid15 Wrote: It depends on the game, some are more CPU intensive than others...
Posting profiler information might help me to improve the CPU performance.
I try to do it most times i post some game test! But i know i am not the only one who attached information so you are the one who get overloaded hahahaha Sad

Anyway, was solved in r2769; thanks!
[Image: montcer.png]
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)