Just came across this:
https://dl.acm.org/doi/10.1145/3837730.3837734
5 Likes
The paper’s title is " Are We There Yet? Evaluating AMDFlang Fortran Compiler Readiness on Production Workloads", so was curious if the paper conforms to the Betteridge’s Law of Headlines and the answer seems to be that it does.
It sounds like they did significantly better than when I last tested it (March '25). I had a great time at a workshop/hackathon reporting bugs in openMP offload for code such as taking the transpose of a 3x3 matrix or doing a +1 reduction. It was still nice to see the progress in the compiler though!
I’m glad to see it’s taken further steps!
1 Like