wide-int

[Kenneth Zadeck]

I am sorry that in the haste of battle that mike did not have an 
opportunity to write a proper introduction to the is patch.  The patch 
was submitted last night so that it could be formally submitted by the 
end of stage 1.

This patch is the same as the top of the wide-int branch that has been 
publicly available for public view and comment for the last several 
months.   During that time it has significantly evolved and improved.   
The wide-int interface has changed in significant ways as well as it has 
now been tested on at least one instance of every active architecture 
and has been tested on one or two architectures for every single language.

The wide-int branch deals with a fundamental data structure problem in 
the current gcc trunk.    Integer math larger than 64 bits is generally 
buggy and beyond 128 bits is completely wrong and generally causes the 
compiler to ICE.   This patch fixes all of the 128 bits and below 
problems for every port, and if certain changes are made to the port 
(see the part of the diff relating to the RS6000 for an example) can 
support integers of any size.

Currently there are a few public ports that have some code that uses 
OImode.   The usage of OImode in those ports is currently limited. I 
personally do not know if it is because the architectures make very 
limited use of OImode, or if the port maintainers could not justify 
doing the work that we have done here. However,  when this patch is 
accepted, GCC will be able support any length integer for any kind of 
operation.

Kenny

On 11/21/2013 05:08 PM, Mike Stump wrote:
> This patch adds support for ints wider than double_int.
>
> Ok?
>