home *** CD-ROM | disk | FTP | other *** search
- Path: informatik.tu-muenchen.de!fischerj
- From: fischerj@informatik.tu-muenchen.de (Juergen "Rally" Fischer)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: doubling pixels horizontally
- Date: 29 Feb 1996 16:06:57 GMT
- Organization: Technische Universitaet Muenchen, Germany
- Distribution: world
- Message-ID: <4h4iv1$n3s@sunsystem5.informatik.tu-muenchen.de>
- References: <4f4ibc$gl9@news.cs.tu-berlin.de> <591.6610T1165T2102@login.eunet.no><1045.6611T753T2256@vip.cybercity.dk><4faoe1$47@sunsystem5.informatik.tu-muenchen.de><2991.6612T1034T625@vip.cybercity.dk><576.6613T1070T1730@login.eunet.no><1257.6614T57T922 <4gkknd$b1v@brachio.zrz.TU-Berlin.DE>
- NNTP-Posting-Host: hphalle2i.informatik.tu-muenchen.de
- Originator: fischerj@hphalle2i.informatik.tu-muenchen.de
-
-
- In article <4gkknd$b1v@brachio.zrz.TU-Berlin.DE>, rawneiha@w350zrz.zrz.TU-Berlin.DE (Philipp Boerker) writes:
- |> Organization: Technical University Berlin, Germany
- |> Lines: 32
- |> Message-ID: <4gkknd$b1v@brachio.zrz.TU-Berlin.DE>
- |> References: <4f4ibc$gl9@news.cs.tu-berlin.de> <591.6610T1165T2102@login.eunet.no><1045.6611T753T2256@vip.cybercity.dk><4faoe1$47@sunsystem5.informatik.tu-muenchen.de><2991.6612T1034T625@vip.cybercity.dk><576.6613T1070T1730@login.eunet.no><1257.6614T57T922|>
- |> NNTP-Posting-Host: w350zrz.zrz.tu-berlin.de
- |>
- |> ludvigp@ifi.uio.no (Ludvig Pedersen) writes:
- |>
- |> >>Have you timed the routine and have you compared it to a plain copy?
- |> >>My routine takes 48c for conversion too, and I do proper pipelining
- |> >>and still it is not chipmem speed.
- |> >That means your routine should be just as fast as ours, I guess.
- |>
- |> I think, I made an logical error! If the routine takes 8.x c/pixel and it
- |> converts 3.5 M pixel/second that means that on the paper only 30M c should have
- |> been spent. Where are the other 20M c? Wasted while waiting for mem? That would
- |> mean that the routine is pretty free on 25 MHz too! :)
- |>
- |> >Your are not using the CPU and the BLITTER at the same time in chipram are
- |> >you??? Because that is very foolish! :) (read: SLOW!)
- |>
- |> No, I don't. I even toggle blitternasty-bit at the importent places...
-
- not good. blitter parallel (default) is as fast as nasty as long as
- cpu doesn't acess chipmem.
-
- blitternasty was mostly good in A500 times (cpu always acesses chipmem! ever
- tried a halt instruction ? should make bus free waiting for blitter...)
- and blitternasty is mostly bad in AGA times (even on vanilla A1200, due to cache).
-
- |>
- |> >Anyway, to make sure I timed it.
- |>
- |> >Using a PAL lowres screen with 256 colors, all interrupts off.
- |>
- |> >c2p: Used 74 rasterlines on a 160x128 screen
- |> >Copy: Used 73 rasterlines.
- |>
- |> >I consider that as pretty equal.
- yep.
- |>
- |> Inside the error-delta for timings, I'd say.
- |>
- |> Greets,
- |> Phil.
- |> grond/matrix
- |>
- ------------------------------------------------------------------------
- fischerj@Informatik.TU-Muenchen.DE (Juergen "Rally" Fischer) =:)
-
-