Add test for ilog2 #93
Conversation
| // assembly-output: bpf-linker | ||
| // compile-flags: --crate-type cdylib -C link-arg=--unroll-loops -C link-arg=-O3 | ||
| #![no_std] | ||
| // bpf target did not historically support the ilog2 operation because LLVM did |
There was a problem hiding this comment.
what are you testing here? There's no CHECK; can you add some?
There was a problem hiding this comment.
TBH I do not remember the details of how to use this testing. To my memory, I wanted this to build successfully, but it currently doesn't because ilog2 IL is refused by LLVM.
There was a problem hiding this comment.
No, because this is actually affecting some of my code. Right now, I have to manually compute ilog2. If I ever depended on a crate which uses this functionality, I would be screwed 😅
There was a problem hiding this comment.
How do I correctly write this testcase to check that the compilation is successful?
There was a problem hiding this comment.
What is it you're looking for, here?
There was a problem hiding this comment.
How do I correctly write this testcase to check that the compilation is successful?
Read the llvm-filecheck docs? Have you tried running this locally?
There was a problem hiding this comment.
What is it you're looking for, here?
Historically, ilog2 would cause a linkage error that was very difficult to diagnose.
Have you tried running this locally?
Not recently, no. I'll try to take a look at it this week and update it according to the llvm-filecheck docs.
| @@ -0,0 +1,13 @@ | |||
| // assembly-output: bpf-linker | |||
| // compile-flags: --crate-type cdylib -C link-arg=--unroll-loops -C link-arg=-O3 | |||
There was a problem hiding this comment.
do we need unroll-loops and 03? seems superfluous, see https://en.wikipedia.org/wiki/Chekhov%27s_gun
There was a problem hiding this comment.
Almost certainly copy-pasted this 😀
It seems that LLVM does not have an instruction pattern for ilog2/leading bits/trailing bits. This adds a test case for this behaviour.
This change is