Long workload name layout fix #269

* initial gui commit with workloads folder * Insert Workload directory, and Drag/drop works, options/values update correctly * min points is now a slider and updates on each workload. still doesnt filter * min points filter, now filters * points filter works with point selection * globs workloads in current dir and one level down * regex filters * regex filtering works * changed index tuples in dataframes * added correct indexing to data * Remove pycache files * ignore .pyc files * format python files * deleted unnecessary workload files * remove unused python scripts * add requirements.txt * remove unused imports + unnecessary code * Relocated source files * support pip installation * changed layout to handle more data, updated to new json format * return of smooth lines * scrollable histogram * fixed for throughput * formatting * miscellaneous fixes * precalc sorting values for less work in callback - added compute_sorts to precalculate dataframe sort columns -renamed columns in compute_speedups to same format as output -removed sorting algorithms from callback -fixed func_regex and exp_regex to use func_list and exp_list * Min/Max sort works -min/max sort fixed, -For some reason dataframe is calculated twice... * Merge with new compute scripts -create settings if nonexistent -write current path, and clil flag to settings after argparse -add new flags from validation script -use impact avg from validation scripts -use new classes for calculating throughput/latency/impact -uses regex from cli for progress points/experiments * Remove outdated code -removed unused/commented code -fixed import errors * reformatting * Pasting path in gui fix -fixed parser default values -pasting path to workload folder in box fixed to new format * drag and drop file fix * Updated python headers + v0.1.0 + fixed --version - set version to 0.1.0 - moved source/python/gui/VERSION to source/python/gui/source/VERSION - populated __init__.py with __version__ - populated all python files with author, copyright, license, maintainer, and status - fixed usage from installation * formatting --------- Co-authored-by: JoseSantosAMD <Josantos@amd.com>

* Update workflow paths-ignore * Fix call-stack histogram + reorder display - read in call-stack samples and generate bar graph - reorder display so that call-stack histogram is at bottom

* refresh button - globally saves workload path given from main - add refresh button - re runs parsefile with workload path * refresh button works with insert path - refreshes data in workload path - accepts both directory and file - if directory only stores first file found for now - some renaming * formatting --------- Co-authored-by: JoseSantosAMD <Jose.Santos@amd.com>

* Added error bars - from impact stddev * formatting - NOTE: a future PR will resolve this PR using the incorrect data for the error bars --------- Co-authored-by: JoseSantosAMD <Jose.Santos@amd.com>

* Gui can be started with no workload. - some renaming of functions - if empty data, empty div is returned * workloads can be directory also * filtering errors fix - fix errors from applying sort/filters when data is empty --------- Co-authored-by: JoseSantosAMD <Jose.Santos@amd.com>

* Multiple files now read from directory -Now accepts multiple jsons in directory -samples error fix -light mode now an option -select which file to present data from dropdown (Does not work with histogram yet) -workload added to dataframe -More color options * Reformat * Requested Changes -Uses os.path.basename -removed data for runs_dict -default is now sorted by impact -naming convention consistency fix -No more filt_kernel_names -gui console output determined by verbosity * terminal output if verbosity > 3 or only CLI * Use Imapct sum instead of avg -verbosity help menu update

* Multiple files now read from directory -Now accepts multiple jsons in directory -samples error fix -light mode now an option -select which file to present data from dropdown (Does not work with histogram yet) -workload added to dataframe -More color options * Reformat * terminal output if verbosity > 3 or only CLI * Terminal Verbosity can now change from gui -Drop down selects verbose level -1: Shows input from gui on terminal -2: shows points in data -3: Used for debugging -Also fixed path vs directory bug in input path from gui * multiple verbose modes in terminal set by gui -multi selection in sort by is removed, kept in workloads -verbose mode 2 shows data from print_speedup_info -sort by changed to if/elif -removed unused fig -fixed errorbars -line_stddev now in dataframe * Reformatting gui labels and filters - resizing filters - renamed labels in graphs for consistency - renamed labels in dataframe for consistency - moved styling to styling sheets * Formatting

* Fixes for PR ROCm#260 - Remove duplicate -l / --light argument - Add missing "multi" dict entries for input filters * Update RedHat workflow - ignore changes to source/python/gui/** * Fixed sizing for drop downs --------- Co-authored-by: JoseSantosAMD <Jose.Santos@amd.com>

* Fully remove verbosity tab * parser, gui, main updates - parser.parse_files returns filenames - parser.parse_files handles verbosity - parser.parse_files handles settings - gui fixes removing verbosity - improved main - call parse_files - support recursive glob - improved settings support - support ~/.omnitrace-causal-plot.json

- added span to show full name - shortened name to fit dropwdown - changed layout for added consistency

- refresh button is to the right - header is more consistent across different width screens

- center div makes turns into multiple lines if not all items fit

- moved find_causal_files to parser so that main and gui can access - resized refresh width to allow for same shape across different screens

- graphs now have same width - chart headers start well below the header with filters

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Long workload name layout fix #269

Long workload name layout fix #269

Commits on Apr 2, 2023