Reduce excessive loop unrolling in lbgpu velocity interpolation #2982
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This caused excessive register usage
Codecov Report
@@ Coverage Diff @@
## python #2982 +/- ##
======================================
Coverage 82% 82%
======================================
Files 525 525
Lines 26807 26807
======================================
Hits 22015 22015
Misses 4792 4792Continue to review full report at Codecov.
|
Merged
KaiSzuttor
reviewed
Jul 10, 2019
KaiSzuttor
approved these changes
Jul 10, 2019
|
bors r+ |
bors bot
added a commit
that referenced
this pull request
Jul 10, 2019
2982: Reduce excessive loop unrolling in lbgpu velocity interpolation r=KaiSzuttor a=mkuron This caused excessive register usage, especially when combined with thrust. Issue discovered by @fweik in #2878. It turns out that this is a problem for CUDA too, it just exhibits a different behavior. Instead of crashing like on HIP, CUDA just produces a large binary and slower code. In a perfect world, the compiler should display a warning, but I guess neither AMD nor Nvidia operate in a perfect world. Co-authored-by: Michael Kuron <mkuron@users.noreply.github.com>
Build succeeded |
bors bot
added a commit
that referenced
this pull request
Jul 24, 2019
2984: Don't unroll loops with nontrivial loop bodies r=KaiSzuttor a=mkuron Simple assignments and arithmetic are fine to unroll, but if there's a function call or a nested loop, unrolling is probably not useful Follow-up to #2982. Co-authored-by: Michael Kuron <mkuron@icp.uni-stuttgart.de>
bors bot
added a commit
that referenced
this pull request
Jul 25, 2019
2984: Don't unroll loops with nontrivial loop bodies r=KaiSzuttor a=mkuron Simple assignments and arithmetic are fine to unroll, but if there's a function call or a nested loop, unrolling is probably not useful Follow-up to #2982. 2993: Update Doxygen landing page r=RudolfWeeber a=jngrad Fixes #2600, closes #294 Update Doxygen landing page and add a link to the Related Pages where detailed instructions are provided for developing new features in the C++ core. 2998: Benchmarks reliability r=RudolfWeeber a=jngrad Since 4.0 a few things changed in the CMake workflow (either in espresso CMake files, or in newer CMake releases) that broke the benchmark functionality: - changing myconfig.hpp doesn't always cause a full project recompilation (rare) - ctest sometimes runs benchmarks in parallel The benchmark bash scripts were also brittle. They were both refactored based on `build_cmake.sh`: - trap mechanism with clear error message - another trap with cleanup action - simplified CMake logic - ctest runs benchmarks in serial - erase /build/src before each compilation Other changes: - don't print debug energy values in the CSV output - some column names were changed - the [wiki documentation](https://github.com/espressomd/espresso/wiki/Development#benchmarking) was updated accordingly - and it now discusses condor on ICP machines - reduce risk of LJ simulation crashing with a longer energy minimization - add a benchmark for bonded interactions - print more info in stdout (config file being processed, energy during LJ minimization and simulation) Co-authored-by: Michael Kuron <mkuron@icp.uni-stuttgart.de> Co-authored-by: Kai Szuttor <2150555+kaiszuttor@users.noreply.github.com> Co-authored-by: Jean-Noël Grad <jgrad@icp.uni-stuttgart.de>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This caused excessive register usage, especially when combined with thrust. Issue discovered by @fweik in #2878.
It turns out that this is a problem for CUDA too, it just exhibits a different behavior. Instead of crashing like on HIP, CUDA just produces a large binary and slower code. In a perfect world, the compiler should display a warning, but I guess neither AMD nor Nvidia operate in a perfect world.