Reshape Layer implementation for the GPU Architecture #8
Conversation
This file contains hidden or 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
steremma
changed the title
|
I renamed this PR to add the |
steremma
changed the title
|
I believe this is ready to be merged :) |
- use auto
in case gDebug is greater than 0.
both in the sequential and MT case. The sequential case has also been fixed since a non-triggered lazy snapshot caused a crash.
Bloom filter in header section of so files is well described by: https://blogs.oracle.com/solaris/gnu-hash-elf-sections-v2 and lld/ELF/SyntheticSections.cpp The point is that the static linker puts bloom filter value to .gnu.hash section in so files. We just have to read this value and compare to the mangled_name hash that we're looking for. Bloom filter is a false posive probability data structure, so it might say "yes" to library which "doesn't" contain mangled_name, but it won't say "no" to library which "do" contain mangled_name. Modules W/O this patch ``` Processing tutorials/hsimple.C... hsimple : Real Time = 0.04 seconds Cpu Time = 0.03 seconds (TFile *) 0x562b37a14fe0 Processing /home/yuka/CERN/ROOT/memory.C... cpu time = 0.362307 seconds sys time = 0.039741 seconds res memory = 278.215 Mbytes vir memory = 448.973 Mbytes ``` Modules With this patch ``` Processing tutorials/hsimple.C... hsimple : Real Time = 0.05 seconds Cpu Time = 0.05 seconds (TFile *) 0x564410677780 Processing /home/yuka/CERN/ROOT/memory.C... cpu time = 0.356471 seconds sys time = 0.079519 seconds res memory = 266.73 Mbytes vir memory = 423.59 Mbytes ``` This difference become bigger when we need to lookup more libraries in experiments.
rather than re-creating the full set every time it's needed. This is a performance optimisation. The routine for finding branches names is not lightweight, especially when dealing with trees with tens of branches. The names of the branches is now cached in the RInterface and its ownership shared among the RInterface<T> instances.
- use auto
- formating
- use auto
- spell check
- spell check - formating
- formating
If current pointer changes due to reallocation offset can be used as is
Introduce dummy for it
…minate code duplication
…tectures. The tests include overlapping local views, as well as the case where depth > 0.
…xpect the same arguments
Since some time back the gradient boosting option of TMVA is called "BoostType=Grad", not "GradBoost". This updated textual output to use the new name.
…st on different architectures
…d in-method comments
…architectures to prove correctness
…the `ReshapeLayer` class
…or arguments" This change needs more thought This reverts commit 48fc442.
…s. Added tests for the CPU architecture
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
22 participants
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 PR implements and tests all the functions of the
Reshape Layerin CUDA. Those are:Flattenfunction.Deflattenfunction.Reshapefunction.I additionally refactored the respective testing suite to remove code duplication between the same tests implemented in different architectures.