Re: Floating Point Exception

From: Robert Redelmeier (redelm_at_ev1.net.invalid)
Date: 10/12/04


Date: Mon, 11 Oct 2004 23:57:25 GMT

Steve <steve@steve.com> wrote:
> You're right, I do want it to respond at PCI bus speed rather
> than CPU speed. The PCI card is non-compliant as it's only
> a prototype. It's basically an FPGA and some wiring.

Much as I assumed. Are you sure you have the card
memory marked as non-cacheable?

> No errors occur. The application just runs for ever and
> the PCI card keeps sending data out, which I've checked
> with a scope.

Is that data programmatically also correct?

> So if noise (or non compliance) was a problem I would expect
> it to fail on reads more often than writes, which is not
> the case. But I can't guarantee non-compliance is not the
> cause of the problem.

I'm not sure why you expect more trouble with reads than writes.

Try different data. Does it throw FPx if you just hammer
0x0 or 0xF..F ? What about walking bit with and without
saturation?

For some of this testing, you could just disable the FPx.

-- Robert