|
Chineasthetik by Scarab [web]
[nfo]
|
||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
|||||||||||||
|
popularity : 59% |
|||||||||||||
alltime top: #11567 |
|
|||||||||||||
added on the 2002-08-19 11:05:55 by scicco  |
||||||||||||||
popularity helper
comments
ah, finally a 68k demo again :) i liked the music and parts of the synching, but overall it felt too static.
added on the 2002-08-19 14:18:15 by robotriot 
this is a test on concentration, "how long can you wait for the demo until you press rmb?" , not to long iam afraid ;-)
hmm... the concept was good, but it gets boring when there's nothing else than just flares :/
Obviously not really fast-paced but it's nice to watch anyway... and it's one of the few recent demos that are watchable on a 030.
in a strange way i liked this demo even if really nothing much was in.. really weird. definately not among my favorites, but a nice watch one per year.
drawback is it is way too slow on 040/25..
Not many effects inside, more a design demo but I like it anyway.
...
liked the gfx and design and flow, a nice melancholy feeling :)
I like this neat piece of design.
Got boring after a while, but its nice enough for a thumb!
I think I saw a flare there
China :)
Slow code, the rest is quite OK.
Nice demo, I like it.
design and music
Quote:
Slow code
Thank you for your constructive criticism. Could you please give me some hints how to make the code faster? I'd be really grateful receiving lots of optimisation hints from you. Thanks in advance!
Pro tip: All you need is a really fast RAM. That's the panacea! :]
@StingRay
Not only me noticed a bit of slowness, too bad because the prod is interesting.
A way to improve it could be to reshape the code to take advantage of the cpu cache memories.
Or just raise the requirements (060).
Not only me noticed a bit of slowness, too bad because the prod is interesting.
A way to improve it could be to reshape the code to take advantage of the cpu cache memories.
Or just raise the requirements (060).
And the award for the most clueless comment goes to: Alientech
@StingRay
It would be actually cool to see a version optimized for 040 where the cpu caches are not that limited (like in 020, 030) (...).
It would be actually cool to see a version optimized for 040 where the cpu caches are not that limited (like in 020, 030) (...).
Quote:
A way to improve it could be to reshape the code to take advantage of the cpu cache memories.
Thank your for your optimisation tips. Here's a part of the code:
Code:
FX_C2P1x1INIT
lea .TAB(pc),a0
lea FX_C2P1x1_Routine(pc),a1
move.l FX_CPUTYPE(pc),d7
and.w #$ff,d7
ror.b #4,d7
move.l a0,a2
add.w 2(a2,d7.w*4),a2 ; offset to c2p routine
move.l a2,(a1) ; store ptr to c2p routine
add.w (a0,d7.w*4),a0 ; offset to c2p init routine
jmp (a0) ; execute c2p_init routine
.TAB dc.w c2p1x1_000_init-.TAB,c2p1x1_000-.TAB ; 68000
dc.w c2p1x1_010_init-.TAB,c2p1x1_010-.TAB ; 68010
dc.w c2p1x1_030_init-.TAB,c2p1x1_030-.TAB ; 68020
dc.w c2p1x1_030_init-.TAB,c2p1x1_030-.TAB ; 68030
dc.w c2p1x1_040_init-.TAB,c2p1x1_040-.TAB ; 68040
dc.w 0,0 ; Dummy
dc.w c2p1x1_040_init-.TAB,c2p1x1_040-.TAB ; 68060
FX_C2P1x1_Routine dc.l 0
CNOP 0,4
FX_DO_C2P1x1
move.l FX_C2P1x1_Routine(pc),a2
jmp (a2)
I will now inform myself about cpu cache memories so I can write faster code in the future. Thank you for your tips, they are appreciated!
Ham: Thanks for the pro tip. :D
@StingRay
Thanks for the explanation.
I guess the c2p is only the final stage of the rendering pipeline.
Do you think other optimizations are possible at early stages?
I was also meaning a code entirely written for the 040.
Thanks to the pipeline, after a write access to the memory, the 040 can actually execute other instructions which don' t access memory (the so called "free instructions"), without waiting for the previous write to complete, if the code to run is in the cache.
This basically means that when copying data from fast to chip ram, it is also possible to perform a processing of data, without additional time.
Do you think the speed would benefit from it?
Thanks for the explanation.
I guess the c2p is only the final stage of the rendering pipeline.
Do you think other optimizations are possible at early stages?
I was also meaning a code entirely written for the 040.
Thanks to the pipeline, after a write access to the memory, the 040 can actually execute other instructions which don' t access memory (the so called "free instructions"), without waiting for the previous write to complete, if the code to run is in the cache.
This basically means that when copying data from fast to chip ram, it is also possible to perform a processing of data, without additional time.
Do you think the speed would benefit from it?
submit changes
if this prod is a fake, some info is false or the download link is broken,
do not post about it in the comments, it will get lost.
instead, click here !