Latest master changes to baremetal #213
Merged
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 way, multi-codelet capture becomes possible - In order to register memory state from before capture, we save the protected mem zones and permissions. This is only dune during the initial protection phase, should we intercept mallocs & frees too ?
- dump_init is still the default and is uesd for single codelet capture - multi_dump_init can be used as an alternative for multi codelet capture => when used, the tracee will continue executing after a codelet and the trace rwill free its memory before terminating itself
- dump_init() / multi_dump_init() should still be called at the very beginning of the tracee. - Codelets should be placed between dump() and after_dump() calls. This will now automatically create a new tracer / tracee pair to prepare for the next codelet - A call to dump_close() should be placed at the end of the tracee.
The tracee now behaves "normally" in single capture. In multi capture, it will trigger mem lock for the new tracer everytime, and avoid exiting after non captured inocations.
Propagate master fixes
TODO : - Modify the dump function to accept multiple invocations to dump. Currently, we insert N calls that eah start capture for a single invocation, which is "slightly" incorrect. - Do the same for the OpenMP dump pass
Variable was only used from tracer application. Now it is also used in the Python scripts so these can be run from anywhere.
Variable was only used from tracer application.
Now it is also used in the Python scripts so these can be
run from anywhere.
…ontents" This reverts commit dd76789.
…sters' contents"" This reverts commit 38404e1.
This reverts commit 494c430.
Variable was only used from tracer application. Now it is also used in the Python scripts so these can be run from anywhere.
Graph generation etc... still seem to have issue, but report generation does run
These changes should make it easier to propagate the CERE_MODE command in complex build systems. Is should now be accessible by using $CERE_MODE, which was not available previously as the variable's definition was inlined in the command
These are optional variables that will default to CERE_WORKING_PATH when undefined. - CERE_BUILD_PATH will be used as the current dir to launch build commands - CERE_RUN_PATH will be used as the current dir to launch run commands
New CERE_BUILD/RUN_PATHS variables
Env variable changes to multi codelet branch
"dump" function now accepts a more simple string of invocations. It is parsed at runtime. While this needs to be doen for every dump call (even those not corresponding to an invocation to dump), this makes the LLVM pass simpler for now.
…into david-c-wong-multi-codelet-capture
New feature that allows to specify and capture multiple regions/invocations at once. - Modified the tracer / tracee interactions -> at the end of a region, we can now spawn a new tracee, on tracee becomes tracer, with memory unlocking system - Modified LLVM passes to insert "dump" and "after_dump" calls around multiple regions - Updated the cere capture script to parse / check a list of invocs passed as CLI argument
We now have cerec for C, cerec++ for C++ and ceref for Fortran (optional). Source language detection caused issues at link time in the past, especially when linking C++ : for instance, the C++ standard lib would not always be found because clang would be called called instead of clang++. Now, choosing the right compiler is the task of the target application's build system, instead of CERE's automatic language detection.
This could cause issues, for instance when CMake was trying to write .ll files to write protected directories when building test programs
Compiler frontend rework
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.
For now, baremetal will probably be kept as a separate/optional feature branch