[llvm-dev] CFLAndersAliasAnalysis print implementation
Kenneth Adam Miller via llvm-dev
llvm-dev at lists.llvm.org
Wed May 23 15:35:14 PDT 2018
Ok, the command necessary to get past this is:
opt ./example.ll -disable-basicaa -cfl-anders-aa -aa-eval -disable-output
-print-alias-sets
It was non-intuitive, because I thought turning on cfl-anders-aa would also
enable printing the alias sets automatically. It also helps to add this
flag: -mem2reg
On Tue, May 22, 2018 at 12:50 PM, Kenneth Adam Miller <
kennethadammiller at gmail.com> wrote:
> Hello all,
>
> I'm having trouble getting this analysis to print out it's graph of
> aliases. I am processing an example C file into llvm ir, like this:
>
> int main(void) {
> int x=1, y=2, z=3;
> int *p;
> int **p1, **p2;
> if (x==z) { //0x100000f40
> p=&x;
> p1=&p;
> x*=2;
> *p1+=z;
> }
> }
>
> clang -S -emit-llvm example.c -o example.ll
>
> Then, I am trying to run the Andersen alias analysis on it:
>
> opt -analyze --cfl-anders-aa -aa example.ll
>
> I wrote a new print function for the anders alias analysis, and the print
> function parameter Module *M and the Result calculated each indicate that
> there are no aliases. I'm certain I'm missing something, but I don't know
> what. To be absolutely clear, a Result is calculated of type:
>
> std::unique_ptr<CFLAndersAAResult> Result;
>
> And CFLAndersAAResult contains member fields:
>
> DenseMap<const Function *, Optional<FunctionInfo>> Cache;
>
> std::forward_list<cflaa::FunctionHandle<CFLAndersAAResult>> Handles;
>
> The DenseMap Cache is empty, and so is Handles.
>
> I can do -print-alias-sets, and that prints something, but I don't think
> it's the result of the Andersen analysis. Because when I do commands with --cfl-anders-aa
> taken out, I still get the same alias sets.
>
> Can anybody help me get the Andersen alias results?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20180523/2d3e0ddb/attachment.html>
More information about the llvm-dev
mailing list