Is there any code about generating the results of MiBench and PolyBench as shown in figs. 10 and 11 of the paper?

Question

Is there any code about generating the results of MiBench and PolyBench as shown in figs. 10 and 11 of the paper?

Closed this issue 3 years ago · 3 comments

As shown in Figs.10 and 11 of the paper "NeuroVectorizer: End-to-End Vectorization with Deep
Reinforcement Learning", could you provide source codes about generating the results of MiBench and PolyBench?

Answer 1 · 2021-08-23T12:12:36.000Z

Hi @gloaming2dawn , thanks for asking.
Can you try using the artifact description in the paper?
https://dl.acm.org/doi/pdf/10.1145/3368826.3377928

Answer 2 · 2021-08-25T21:35:19.000Z

Hi, @AmeerHajAli , Fig.9 can be reproduced as the artifact description. However, there is no artifact description about the result of PolyBench (i.e., fig. 10). When I try to reproduce the result of PolyBench, I cannot get the similar result of fig. 10. Pragmas are invalid for many loops in PolyBench (shown "loop not vectorized" when compiling).

Answer 3 · 2021-11-27T01:04:57.000Z

Can you please provide more details on how are you running it and which version of LLVM are your running?