OESF Portables Forum
General Forums => General Discussion => Topic started by: danboid on January 23, 2007, 09:02:43 am
-
Thanks to Tyrannozaurus.com for pointing out that uCLinux is now running on the PSP, now with frambuffer and SDL support!
http://df38.dot5hosting.com/~remember/chris/ (http://df38.dot5hosting.com/~remember/chris/)
As you can see from dmesg, the PSP only packs a measley 2.42 Bogomips which makes the PS3's 158.72 and a (non-overclocked!) SL-CXX00's 400+ Bogomips look really damn impressive! But this I don't understand, I thought that the PS3s Cell CPU was supposed to be on a par with at least a 1Ghz G4 CPU and surely they manage more than 158 bogomips? I understand this score doesn't take account of the Cells many co-pro's or whatever Sony are calling them but how can my Zaurus be 3 or 4 times faster than a PS3 (if I overclocked it), even if this is just at integer calcs?
-
impressive
if ONLY the psp had a touchscreen
... and higher res
... and bluetooth, and more memory, and a keyboard, THEN it'd be a Z killer
-
impressive
if ONLY the psp had a touchscreen
... and higher res
... and bluetooth, and more memory, and a keyboard, THEN it'd be a Z killer
[div align=\"right\"][a href=\"index.php?act=findpost&pid=152154\"][{POST_SNAPBACK}][/a][/div]
Don't forget USB 2.0 host and SD/CF slots (they're cheaper than Sony's formats)!
-
awwww MAN! Sweet...
DOWN WITH SONY!
-
actually that is a very wierd resault, 2.34 or whatever is just crazy, me thinks somthing hasent been set up properly (perhaps cpu scaling, perhaps theintro stuff they have clocks down when not doing anything)
note the lack of driver stuff seems he only got the led working. also uclinux is nice but i really like a mmu for several reasons, but i could be misinformed about that (mainly security reasons but perhaps there can be some sort of workaround to prevent apps snooping on each other, not that its reall required but i do like security )
i wonder if the ps3 downclocks as well, i havent double checked you figures however. perhaps it is running with debugging options and not optimised, that would really knock the performance down on some things. also compiling for an older architecture can sometimes inccur a performance hit.
i cant remeber the exact instruction however i belive that ARM processors had an intsruction that hada penelty that was not present in the earlier versions, but grain of salt without varificatoin and double checking (i mean i am remebering this from memorie wihch is "short term" at the best of times)
another thing is what does bogomips mean? well its a mesure of how many times a second you can do "nothing", its used as a calabration loop by the kernel for timing and such. perhaps in this kernel it dosent use a nop but a sleep instead and gets woken by the timer, that really seems like it would fit with the "get it to work no matter what" that people try before relasing
well that was my $10 for the discounted rate of $0.02
-
looks like my optimisation bits were correct to a certin extent
CC = mipsel-uclibc-gcc
CFLAGS += -O0 -static -elf2flt -static-libgcc -lpthread -msoft-float
LDFLAGS += -O0 -static -elf2flt -static-libgcc -lpthread -msoft-float
-
Da_Blitz / Meanie:
Have you guys ever met up for a super Z geek out or have you ever met at a Sydney LUG meeting? Whom else have you amongst the prominent members of the Sydney LUG, if either of you is a member?
-
<--- post 486, almost Top 50 material
-
actually that is a very wierd resault, 2.34 or whatever is just crazy, me thinks somthing hasent been set up properly (perhaps cpu scaling, perhaps theintro stuff they have clocks down when not doing anything)
[div align=\"right\"][a href=\"index.php?act=findpost&pid=152224\"][{POST_SNAPBACK}][/a][/div]
he did say that he hadn't enabled the cache; I would guess in arm cpu that'd make a huge difference. I imagine too that the PSP is going to be very dependent on coprocessor chips to accelerate the graphics and sound..
I was very impressed how they managed to write a useable text editor on the PSP using only the gaming buttons!
-
i still wonder about that low performance, probelly somthing bizare or mundane
still bogomips rely on nop's which are normally one cycle so thats why they follow clockspped quite closly (ecept for x86 which is 2x but not 2x performance scince its a mesure of "no operation")
i have been thinking of going to a Sydney lug meeting, anyone else frome here go at all?
-
remeber that the processor's clock speed on the psp is software configurable, so most likely it starts out at a slow base speed and has to be changed to a higher speed by the software(firmare). the guy who did this might not have taken this into account.
this is the only thing i can think of that would cause such a poor performance
-
stampsm:
Yes, you're probably right about that, but what of the PS3's piffling 158 bogomips?
There are some nbench results running on the PS3 here
http://type-x.ddo.jp/pukiwiki/index.php?%A...%C6%A5%B9%A5%C8 (http://type-x.ddo.jp/pukiwiki/index.php?%A5%D9%A5%F3%A5%C1%A5%DE%A1%BC%A5%AF%A5%C6%A5%B9%A5%C8)
and here
http://matsuu.blogspot.com/search/label/ps3 (http://matsuu.blogspot.com/search/label/ps3)
Although I'm not entirely sure whats going on there as why are the scores much higher for the 'old index' Pentium 90 compared to the new index of a AMD K6/233?
-
stampsm:
Yes, you're probably right about that, but what of the PS3's piffling 158 bogomips?
There are some nbench results running on the PS3 here
http://type-x.ddo.jp/pukiwiki/index.php?%A...%C6%A5%B9%A5%C8 (http://type-x.ddo.jp/pukiwiki/index.php?%A5%D9%A5%F3%A5%C1%A5%DE%A1%BC%A5%AF%A5%C6%A5%B9%A5%C8)
and here
http://matsuu.blogspot.com/search/label/ps3 (http://matsuu.blogspot.com/search/label/ps3)
Although I'm not entirely sure whats going on there as why are the scores much higher for the 'old index' Pentium 90 compared to the new index of a AMD K6/233?
[div align=\"right\"][a href=\"index.php?act=findpost&pid=153002\"][{POST_SNAPBACK}][/a][/div]
probably something to do with the old processors not having a FPU so they might not be benchmarked on it while newer processors have one. the code might not be using the FPU and doing the calculations the hard way.
-
Da_Blitz / Meanie:
Have you guys ever met up for a super Z geek out or have you ever met at a Sydney LUG meeting? Whom else have you amongst the prominent members of the Sydney LUG, if either of you is a member?
[div align=\"right\"][a href=\"index.php?act=findpost&pid=152264\"][{POST_SNAPBACK}][/a][/div]
Sydney LUG? why would I want to hang out with a bunch of geeks when I could be enjoying a movie with my girlfriend instead?
-
perhaps the idea of the main processor was only to keep the child processor fed? therefore not much performance is needed (as the calculation part iis supposed to be offloaded)
i wonder how it would do in a data shuffling experiment, i bet that processor kicks the intel chip when it comes to moving data around.
wouldnt supprise me as tehy are designed as computaion power houses and not data shufflurs
-
According to Geekbench 2006, PS3 Linux currently performs about the same as a PowerMac G5 1.6Ghz, which is of course disappointing considering the PS3s Cell CPU has twice that clock rate but its much better than the 158 bogomips rating would suggest. Hopefully this will improve as gcc's cell support matures?
http://www.geekpatrol.ca/2006/11/playstation-3-performance/ (http://www.geekpatrol.ca/2006/11/playstation-3-performance/)