re: Altivec 'bug' or limitation?
re: Altivec 'bug' or limitation?
- Subject: re: Altivec 'bug' or limitation?
- From: George Warner <email@hidden>
- Date: Mon, 31 Jan 2005 09:14:02 -0800
On Sat, 29 Jan 2005 14:16:24 -0800, Scott Horn <email@hidden> wrote:
> This is a curiosity question. I've no experience in distributed or
high-precision software development. I just noticed a funny thing when playing
around with xgrid at home.
>
> In Xgrid Preview 2, with two CPUs, a 800mhz G3 and a 1.25ghz G4. Running
mandelbrot, manual zoom, the following coords:
>
> X: -0.7499982136839042
> Y: -0.008411978256855454
> Width: 4.903769581863252e-07
>
> With 'Use Altivec if available' is checked I get 7 squares with horizontal
colored lines, and two squares with the (apparently) correct rendering. I
surmise that the two correct squares were calculated by the G3 and the seven
incorrect ones by the G4.
>
> To test this I turn off Altivec and run 'Show Image' again. This time all
nine squares return apparently correct results.
>
> Question: Is this possibly a bug in the altivec libraries or hardware, or in
the mandelbrot app, or is it simply a mathematical precision limitation of
altivec?
This sounds like ether the app is trying to run AltiVec code on the G3
(unlikely) or the G3 side is using doubles and the G4 side floats (via the
AltiVec unit) and it's a resolution issue (most likely).
--
Enjoy,
George Warner,
Schizophrenic Optimization Scientist
Apple Developer Technical Support (DTS)
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Xcode-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden