kokkos / kokkos/mdspan

Vectorization issue Intel compiler

Open
#308 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
507
Forks
87
Avg merge
15h 27m
Merged PRs (30d)
2

Description

Hello, I'm having problem vectorizing this very simple loop where dolic_c and tke_Lmix are mdspan objects:

for (int jc = start_index; jc <= end_index; jc++)
    tke_Lmix(blockNo, dolic_c(blockNo, jc), jc) = 0.0;

I'm using intel compiler and from the report I get the following output:

   remark #15389: vectorization support: reference dolic_c.__members.__members[blockNo*dolic_c.__members.__members.__members.__members.__members.__members.__members[1]+jc] has unaligned access   [ /home/k/k202136/YAOP/src/backends/CPU/cpu_kernels.hpp(68,38) ]
   remark #15381: vectorization support: unaligned access used inside loop body
   remark #15335: loop was not vectorized: vectorization possible but seems inefficient. Use vector always directive or -vec-threshold0 to override
   remark #15329: vectorization support: irregularly indexed store was emulated for the variable <tke_Lmix.__members.__members[(blockNo*tke_Lmix.__members.__membe>, part of index is read from memory   [ /home/k/k202136/YAOP/src/backends/CPU/cpu_kernels.hpp(68,21) ]
   remark #15305: vectorization support: vector length 4
   remark #15309: vectorization support: normalized vectorization overhead 0.049
   remark #15450: unmasked unaligned unit stride loads: 1
   remark #15463: unmasked indexed (or scatter) stores: 1
   remark #15475: --- begin vector cost summary ---
   remark #15476: scalar cost: 11
   remark #15477: vector cost: 25.750
   remark #15478: estimated potential speedup: 0.420
   remark #15488: --- end vector cost summary ---
   remark #25439: unrolled with remainder by 2

I think that this is simple enough that should be handled by the compiler and I also don't understand the "irregularly indexed store" issue mentioned in the report

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the loop reported at cpu_kernels.hpp line 68 and reproduce the Intel compiler vectorization report for the mdspan objects dolic_c and tke_Lmix. Trace their indexing and layout to determine why the store is treated as irregular and whether the behavior comes from mdspan or the compiler; document the finding and any actionable workaround.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
compilers, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
28/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.