Message 190068 - Python tracker

Message190068

Author	mark.dickinson
Recipients	mark.dickinson, mirabilos, skrah
Date	2013-05-26.07:40:48
SpamBayes Score	-1.0
Marked as misclassified	Yes
Message-id	<1369554048.65.0.534599691478.issue18062@psf.upfronthosting.co.za>
In-reply-to

Content
> The problem with changing the FPUCW on i387 is that it changes > from 64/15 bit mantissa/exponent to 53/15 bit which is still > not the 53/11 bit of IEEE double, so you still get double- > rounding issues (with denormal numbers only, I guess) because > the internal precision is still higher. That's not a problem for dtoa.c, at least: dtoa.c avoids any use of subnormals in intermediate calculations. It's not really too much of a problem for Python in general, either. Windows typically operates in this mode.

Content

> The problem with changing the FPUCW on i387 is that it changes
> from 64/15 bit mantissa/exponent to 53/15 bit which is still
> not the 53/11 bit of IEEE double, so you *still* get double-
> rounding issues (with denormal numbers only, I guess) because
> the internal precision is still higher.

That's not a problem for dtoa.c, at least: dtoa.c avoids any use of subnormals in intermediate calculations.  It's not really too much of a problem for Python in general, either.  Windows typically operates in this mode.

History
Date	User	Action	Args
2013-05-26 07:40:48	mark.dickinson	set	recipients: + mark.dickinson, skrah, mirabilos
2013-05-26 07:40:48	mark.dickinson	set	messageid: <1369554048.65.0.534599691478.issue18062@psf.upfronthosting.co.za>
2013-05-26 07:40:48	mark.dickinson	link	issue18062 messages
2013-05-26 07:40:48	mark.dickinson	create