[PATCH] D76107: AMDGPU: Don't handle kernarg.segment.ptr in functions

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Mar 12 17:43:08 PDT 2020


rampitec added a comment.

In D76107#1920749 <https://reviews.llvm.org/D76107#1920749>, @arsenm wrote:

> In D76107#1920704 <https://reviews.llvm.org/D76107#1920704>, @rampitec wrote:
>
> > A trap would be better, we can easily spot it.
>
>
>
>
> In D76107#1920704 <https://reviews.llvm.org/D76107#1920704>, @rampitec wrote:
>
> > A trap would be better, we can easily spot it.
>
>
> A trap isn’t particularly different from the null dereference if you try to actually use it. I was leaning more towards a compile time error


It is actually very different. A null pointer dereference means debugging, tracking the pointer, analyzing IR step by step until you see why was null emitted. A trap on the contrary is immediately visible in the ISA dump and immediately triggers the alarm, that compiler have spotted the code which should never been executed.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D76107/new/

https://reviews.llvm.org/D76107





More information about the llvm-commits mailing list