| Happy Birthday Arne! | page 1 2 3 4 5
|
|---|
|
|---|
Wojtek P
Poland
| |
Posts 1597
09 Jun 2011 00:19
| Gunnar von Boehn wrote:
|
Yes this is true, with fetching a line ahead you get quite good speed.
|
Do you use/will use (in L2 cache controller):
- "prefetch always" algorithm, when line X is accessed, check if line X+1 is in cache, if not - prefetch
- "prefetch on miss" algorithm - like above but only if line X required a miss
- something else.with such a huge cache prefetch always seems like best idea.
| |
Matt Hey
USA
| |
Posts 728
09 Jun 2011 11:25
| Wojtek P wrote:
|
Gunnar von Boehn wrote:
|
Yes this is true, with fetching a line ahead you get quite good speed.
|
Do you use/will use (in L2 cache controller):
- "prefetch always" algorithm, when line X is accessed, check if line X+1 is in cache, if not - prefetch
- "prefetch on miss" algorithm - like above but only if line X required a miss
- something else.
with such a huge cache prefetch always seems like best idea.
|
Obeying the post-index and pre-increment addressing mode "hints" as to what cache line needs to be pre-loaded next should do a pretty good job. Adding a base register update with post-index addressing mode would also tell the next cache line to be loaded if necessary as well as encouraging post-indexed addressing mode to be used on following instructions (also faster and reduces code size).
| |
Team Chaos Leader
USA
| | (Moderator)
Posts 2094
12 Jun 2011 01:55
| Out of curiosity, what is the memcopy speed of an XBOX-360 CPU? A PS3 CPU?
| |
Samuel D Crow
USA
| | (Natami Team)
Posts 1295
12 Jun 2011 04:09
| @TCL The PS3 would be faster than the XBox 360 since it has expensive interleaved memory so that sequential memory accesses run much much faster than non-sequential accesses. According to EXTERNAL LINK the maximum memory bandwidth is 25.6 GiB/s. The XBox 360 has 22.4 GiB/s memory interface bus bandwidth according to EXTERNAL LINK .
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
12 Jun 2011 09:18
| Samuel D Crow wrote:
|
@TCL
The PS3 would be faster than the XBox 360 since it has expensive interleaved memory so that sequential memory accesses run much much faster than non-sequential accesses. According to EXTERNAL LINK the maximum memory bandwidth is 25.6 GiB/s.
The XBox 360 has 22.4 GiB/s memory interface bus bandwidth according to EXTERNAL LINK .
|
Well, in real live 10% difference is not worth to talk about. And you need to mind that the numbers that you quoted are theoretical memory bandwidth. Its absolutely impossible to reach these numbers with normal program code. A normal PowerPC written C program can not even reach 5% of these values. (not even Five percent!)
| |
Claudio Wieland
Germany
| | (Natami Team)
Posts 703
12 Jun 2011 10:07
| It is much more important what one does with the available bandwidth. The less data needs to be transported through the bottleneck, the more *effective* bandwidth one gets.
| |
Thomas Hirsch
Germany
| | (MX-Board Owner)
Posts 647
12 Jun 2011 15:33
| Yes, I also should send you my congratulations. I only got my CPU card design running stable for more than three hours yesterday on one FE CPU card (not on the RC CPU card). And is just running at 110MHz.
It is a kind of impressive, working on the project for three years, four years and then watching someone else crossing the finish line in just two weeks.
| |
Wawa Tk
Germany
| |
Posts 581
14 Jun 2011 11:38
| could something be done with all those sticky threads that are obviously finished by now and only take most part of the front page, which cant be resized to make place for more topics?
| |
Peter K.
Germany
| | (MX-Board Owner)
Posts 439
15 Jun 2011 10:52
| Wawa, you are right - there were too many sticky threads.
Cleaned that up... .I could throw the links into our seldomly used news-blog to
make them more easy to find, if there is a demand.
| |
Deep Sub Micron
Germany
| | (MX-Board Owner)
Posts 567
15 Jun 2011 11:08
| Peter K. wrote:
|
I could throw the links into our seldomly used news-blog to
make them more easy to find, if there is a demand.
|
Good idea!
| |
Wawa Tk
Germany
| |
Posts 581
15 Jun 2011 11:53
| thanks, peter, that s much better.
| |
Vojin Vidanovic
| |
Posts 5
09 Nov 2011 19:04
| Its great to see Natami performs well as "super 68k".
No one expects its to be super computer as to compare to any today
x86 / PPC board. Just to be super Amiga Classic is enough.Even some unoptimized stuff like x5 performance of A4000 040 25Mhz is way nice - having 100Mhz CPU was Amiga dream. Not to mention newer test can only turn out better results. Hope its exciting times for all the Classic lovers.
| |
|
The team will post updates and news here