Setting Up Core Frequencies
Posted 04 February 2008 - 01:50 PM
with the help of http://documentation...081_sh7709s.pdf
i am now able to set the core-frequencies in real-time. but i can only switch between 33MHz and 66MHz.
every other combination of bits lead the classpad to crash. but i have read that somebody did a 3 and 4 times faster clock speed. so does anybody know how? i could not find any bit combination that worked.
i updatet '1001', know it is possible to control the tracer with the cross-pad and to switch between low and high speed via 'backarrow' for low speed and 'clear' for high speed so the use of exec-speed is not necessary.
the frequency at startup is restored after leaving the benchmark succesfully.
also interesting if someone know if it's possible to overclock the lcd-display.
Posted 04 February 2008 - 11:42 PM
so i got the code i wanted and it turned out that i used the wrong hardware documentation. there was a talk about 6 times multiplier instead of four.
so i tried the right code for speeding up classpad with factor 4 and its running quiet well for an hour now.
more is not possible within software, or? doesn't seem so.
i also tried the very well gameboy emulator, and uploaded some roms on the classpad via the ebook-reader tool. don't know if it's the right way, but it works. the graphic is awesome but the speed is low. i think thats because of the grayscales. i guess the display looks only sharp within a 200ms refresh cycle. so 5fps. and that is nearly how gameboy on classpad is to be played. i think faster rendering would result in flicker or so and you would not be in sync with the display.
for my benchmark i get know with 4 time speed a total overdraw of 2.3 at 20 fps.
so you can render more than twice to screen to get a smooth framerate of 20fps.
if someone knows plasma pong. this would be really possible. even in black and white with a good performance. the fast speed and the low refreshrate would give it a natural blur.
aussuming the cp can do the background calculation in reasonable time. with the use of parallel instruction, perhaps.
for myself i want to calculate and visualize behaviour of gravitation. so thats why i'm interessted in speeding up classpad.
but before i do that, i think i will learn to write a small 3d-renderer with depth-buffer and two-color rendering.
on pc it's useless because there is directx and opengl but on a calculator it could be funny.
grayscales would be to slow i think. but a small bsp-engine with per-trinangle-sorting should do it.
so if someone has already a real-3dengine done, put it on the table:). would be really interesting to hear what is possible on this calculator.
Posted 09 February 2008 - 11:56 PM
are you really sure about the 33 MHz?
Probably you have already noticed that Kucalc had found out the clock speed of the FX-9860G @ 14.74 MHz.
I have measured the power consumption of the FX-9860G and the CP-300+. Considering the same value,
I guess the clock frequencies are also identical.
If you need a benchmark to test the ClassPad versus FX-9860G, have a look here.
I have used the unstructured C version with 10000 iterations to test the FX-9860G (24.8 sec).
Posted 11 February 2008 - 04:38 PM
so default clock-frequency seems to be 16 2/3 MHz and goes up to 66 2/3 MHz withz 4 times multiplier.
first i just wanted to know the graphical-performance. now as i am trying to code a sofware-renderer the classpad turned out to be rather low in performance in respect to floating-point calculations. could it be? so i think i have to switch to fix-point numbers.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users