support for addressing units which are not 8 bits

Project Infrastructure LLVM Dev List Archives
1

Hi!

I want to write a LLVM back-end for a bit addressing target architecture.
http://llvm.org/docs/GetElementPtr.html#i-m-writing-a-backend-for-a-target-which-needs-custom-lowering-for-gep-how-do-i-do-this states quite clearly that some additional work must be done for this sort of architecture. Is this still true? Is there a project/branch to make it more flexible? Support for any memory addressing?

I had a short look at SelectionDAGBuilder::visitGetElementPtr(const User &I) and either this method handles it properly already or customization with some target attributes should not be difficult.

And what means "a lot of code in the backend" in the link above? Which other parts are involved?

Thanks,
Boris

2

Hi!

I want to write a LLVM back-end for a bit addressing target architecture.

The Often Misunderstood GEP Instruction — LLVM 18.0.0git documentation quite clearly that some additional work must be done for this sort
of architecture. Is this still true? Is there a project/branch to make it
more flexible? Support for any memory addressing?

I had a short look at SelectionDAGBuilder::visitGetElementPtr(const User
&I) and either this method handles it properly already or customization
with some target attributes should not be difficult.

Yes, you're right.

And what means "a lot of code in the backend" in the link above? Which
other parts are involved?

The tricky part here isn't addressing, it's that is that each address
points at 32 bits, so you have to track down every single place LLVM
hardcodes "i8" and fix it.

CC'ing Philipp Brüschweiler, who had a patch series a while back (see
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20120702/146050.html).

-Eli

3

Hi!

I want to write a LLVM back-end for a bit addressing target architecture.

...

And what means "a lot of code in the backend" in the link above? Which other parts are involved?

The tricky part here isn't addressing, it's that is that each address points at 32 bits, so you have to track down every single place LLVM hardcodes "i8" and fix it.

CC'ing Philipp Brüschweiler, who had a patch series a while back (see http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20120702/146050.html).

Finally I had some time to give it a try. I can't apply the patches to llvm 3.2 because the files include/llvm/Target/TargetData.h and lib/Target/TargetData.cpp don't exist anymore. Probably they have been split into multiple new files. But applying the patches to the new files by hand should not be to hard.

Anyway, are there plans to add this kind of feature to the trunk? Isn't there some demand?

Boris