[LLVMdev] Eliminating gotos
benedict.gaster at amd.com
Tue Aug 12 08:50:17 PDT 2008
That is interesting.
Do you have any pointers to the test cases in question?
The problem is that we don't have conditional branches and so we are going
to have to do some form of goto elimination and as such do you have any
alternatives in mind?
On 12/08/2008 16:40, "Daniel Berlin" <dberlin at dberlin.org> wrote:
>> [bg] Actually this does not need to be the case. The paper that I sighted
>> does not use code replication to resolve irreducible control flow but
>> instead introduces a loop construct.
> We implemented this in GCC back when we first started GIMPLE (since
> GIMPLE is based on the IL the authors of that paper used in their
> compiler), and the code size increases on a bunch of testcases were
> massive due to extra loop constructs.
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
More information about the llvm-dev