| [Files] New Amithlon 1.29 update available | ANN.lu |
| Posted on 11-Mar-2004 20:42 GMT by top (Edited on 2004-03-12 18:05:13 GMT by Christophe Decanini) | 73 comments
View flat
View list |
the New update for Amithlon to run AmigaOS3.9 under PC hardware via Linux Drivers :
° Amithlon Kernel compatible Linux PCI drivers : PC network chipsets, PC sound chipsets
° Added new version of the XCat Utility (Thanks to Bernd Meyer)
° FIX: Somehow the name of the installed amithlon1_com.device was wrong!
Available for download in Aminet
link file:
Amithlon 1.29 update
Author: geit@gmx.de (Guido Mersmann)
Author: amithlon@amithlon.net (Bernd Meyer)
and
Author: support@vmc.de (Harald Frank, VMC)
Author: bvernoux@wanadoo.fr (Benjamin Vernoux, Titan)
|
|
| List of all comments to this article |
| New Amithlon 1.29 update available : Comment 43 of 73 | ANN.lu |
| Posted by Bernie Meyer on 12-Mar-2004 13:41 GMT | In reply to Comment 22 (Kjetil):
The whole "the x86 needs to swap byte order when emulating 68k" thing is pretty much a non-issue.When Intel introduced the 486, they gave it the "bswap" instruction, which will do just what's needed (for longwords; For shorts, you use a rotate, and bytes don't need swapping). On any modern x86, that instruction is blindingly fast, and thanks to the whole out-of-order execution, usually happens while the processor is waiting for something else, anyway.Anyone who ever looked closely at the JIT compiler will have seen that there are a bunch of ways in which operations on registers are delayed, just in case they can be optimized away. For example, adding a constant offset to a register will not necessarily produce any "add" in x86 code; The offset is just remembered, and only once it really needs doing is the "add" done.For a while, I toyed with the idea to have another such state which stored for any x86 register whether it contained its real value, or the byte-swapped version of it --- to avoid doing double-swaps if they weren't necessary. Then I went and hacked things so that instead of one bswap, 3 were output each time --- and the change in speed was negligible. So getting rid of even *all* bswap would also result in negligible speedup.This, of course, is only true for the integer part. In the FPU code, the whole byte-swapping really sucks. A large part of that is due to the fact that the x87 FPU sucks. If I were to do the whole thing over again today, I'd use SSE2 for the FPU, ignore the x87 crap, and make a modern XP or P4 the minimum requirement... |
|
| List of all comments to this article (continued) |
|
- User Menu
-
- About ANN archives
- The ANN archives is powered by #AmigaZeux. It was updated daily (news last: 22-Oct-2004; comments last: 18-May-2005).
ANN.lu was created, previously owned and maintained by Christian Kemp, www.ckemp.com.
- Contribute
- Not possible at this time!
- Search ANN archives
- Advanced search
- Hosting
- ANN.lu was hosted by Dreamhost. Sign up through this link, mention "ckemp" as referrer and he will get a 10% commission on any account you purchase.
Please show your appreciation for any past, present and future work on ANN.lu by making a contribution via PayPal.
|
