llvm dev - Jun 2011 - [LLVMdev] LLVMdev Digest, Vol 84, Issue 44

Chris,
            good to see the AbstractType stuff vaporize, but I still  
have a question, the Type system seems to have
a "shadow type" system, every user Type created also has a  
corresponding XYZValType,  is the "shadow type"
system still retained ?   It's purpose seems(?) to be for factoring  
the "unique-ing" logic out of the rest of the type
system logic, but I'm wondering how much sense that makes (or perhaps  
I'm not looking at it correctly yet).

sincerely,
Peter Lawrence.



On Jun 25, 2011, at 9:42 AM, llvmdev-request at cs.uiuc.edu wrote:
>
> Once that lands, the constant uniquing is greatly simplified  
> because AbstractType stuff all vaporizes.  When the branch lands,  
> the constant uniquing maps should all be replaced with folding sets.
>
> -Chris
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20110627/d6047e8c/attachment.html>

On Jun 27, 2011, at 9:41 AM, Peter Lawrence wrote:
> Chris,
>            good to see the AbstractType stuff vaporize, but I still have a
question, the Type system seems to have
> a "shadow type" system, every user Type created also has a
corresponding XYZValType,  is the "shadow type"
> system still retained ?   It's purpose seems(?) to be for factoring the
"unique-ing" logic out of the rest of the type
> system logic, but I'm wondering how much sense that makes (or perhaps
I'm not looking at it correctly yet).
Hi Peter,

Most of that completely disappears and is replaced with simple DenseMaps.  It is
much faster and easier to understand.

-Chris
> 
> sincerely,
> Peter Lawrence.
> 
> 
> 
> On Jun 25, 2011, at 9:42 AM, llvmdev-request at cs.uiuc.edu wrote:
> 
>> 
>> Once that lands, the constant uniquing is greatly simplified because
AbstractType stuff all vaporizes.  When the branch lands, the constant uniquing
maps should all be replaced with folding sets.
>> 
>> -Chris
>> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20110702/a266b761/attachment.html>