Nvbit support #2
Open
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
Fix Link order on Linux
…equire compressed trace file to end in .gz so we can use symlinks etc). Got rid of compiler warnings.
-- Checkpoint prior to c+++ --
…tate, remove some statics and unused variables. Tested.
…istics and 2nd pass.
… library separation.
… pin trace file pattern generation. Simplfied signature of some library functions which now use MemPatterns interfaces. Simplified gs_patterns.cpp main to use the new MemPatternsForPin and interfaces.
…to gs_patterns.cpp which handles trace file based input. Refatored second_pass handling so it can be done with any code able to get a collection of trace_t's.
Improved filename creation logic as well.
…ethods. First cut of MemPatternsForNV (gsnv_patterns.cpp) -- no functional but compiles.
Build libgs_patterns_core as shared lib - required for nvbit linkage. gsnv_patterns converted to header file to simplify nvbit build. Some missing default implemetations for gsnv_patterns.
Instroduced a file runner for gsnv_patterns which can write and read an nvbit trace binary output file checkpoint for nvbit <non-functional>
Call nvgs_patterns on each mem_trace callback
…t <non-functional>
…put patterns looking reasonable.
… still need to setup temp dump file for second pass use.
… and MemPatternsForNV into separate namespaces to avoid namespace conflicts. Renamed a few files to be more descriptive. Deleted old gsnv_test as its no longer needed.
…_trace_out_file() in nvbit trace mode.
…MT with -d 3,3,3 -b 1
…en Pin and NVBit. InstrWindow is now dynamically allocated to increase max supported size(required to support 2k). Reset and intialization now within InstrWindow class (rather than in gs_patters_core).
…by nvgs_trace. Use iaddr field rather than base_addr for CTAs. Use base_addr comparison to addr to determine mem instructions which are part of warp for CTAs. Removed remaining malloc calls in gs_patterns.h (a couple are still left in unused functions in utils.h). Added partial string match to kernel name lookup to make it easier to specify kernel to trace in config. Fixed name of config variable. Renamed min_size to max_access_size in InstrAddrAdapter to be consistent.
…iaddr for lookup.
Add support for optional one_warp mode (handles traces from warp 0 only) which can be set via cmd line or config setting. Pattern size length (PSIZE) can now dynamically grow between INITIAL_PSIZE(2^15) and MAX_PSIZE(2^30) on each pattern array.
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.
Major refactoring of gs_patterns into 3 components gs_patterns binary, libgs_patterns_core.so and libgsnv_trace.so.
Migrated to C++.
Added support for instrumenting CUDA kernels via nvbit/libgsnv_trace.so and also extracting an nvbit trace file.
The extracted nvbit trace file can be used to run gs_patterns similarly to how pin traces are used.
Updated various readme files with details on how to use the new functionality.