[LLVMdev] Temporary register subscript

[Jun Mendoza]

Hi Eli,

Ok, I understand the SSA meaning is intact regardless of the representation.
Thanks for the interesting clarification on the demo behavior regarding
asserts!

-Jun

On Mon, Aug 8, 2011 at 12:15 PM, Eli Friedman <eli.friedman at gmail.com>wrote:

> On Sun, Aug 7, 2011 at 9:01 PM, Jun Mendoza <junmendoza78 at gmail.com>
> wrote:
> > Hi Folks,
> > I noticed that recently in the online code gen demo that generating the
> > unique temporaries have been modified.
> > I don't believe i changed any options. (C++, demangle names, no
> > optimization)
> >
> > Where:
> > if(x >= y){
> > }
> >
> > Previously Compiles to:
> > %0 = load i32* %x, align 4
> > %1 = load i32* %y, align 4
> > %2 = icmp sge i32 %0, %1
> > br i1 %2, label %bb, label %bb1
> > bb:
> > ; true goes here
> >
> > Currently Compiles to:
> >
> > %tmp = load i32* %x
> > %tmp1 = load i32* %y
> > %cmp = icmp sge i32 %tmp, %tmp1
> > br i1 %cmp, label %if.then, label %if.end
> >
> > if.then:
> > ; true goes here
> >
> > If this is the case then is there a mailing list or Status update page
> that
> > reflects this behavior?
> >
> > My question may seem trivial however, im still in llvm exploration mode
> :)
> >
> > Im working on a compiler that generates LLVM IR and for the moment i dont
> > use the code gen API. I rely on the Assembly manual and demo for
> reference
> > (Although at some point i will need to use the API in order to take full
> > advantage).
>
> Yes, we recently replaced the compiler on the demo page.  Both syntax
> forms have the same meaning; changing the name of anything that is not
> a global does not affect code generation.  Giving the values names
> tends to be a bit more convenient for compiler developers to read.
> Most values are named in builds with asserts enabled, but are not
> named in builds without assertions for performance reasons.  I guess
> the build the demo page is using is a build with assertions enabled at
> the moment.
>
> -Eli
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110808/a441421b/attachment.html>