""" MIT License Copyright (c) 2019 - 2025 Advanced Micro Devices, Inc. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. """ import os import subprocess # nosec import argparse import datetime import shutil import sys sys.dont_write_bytecode = True sys.path.append(os.path.join(os.path.dirname(__file__), "..")) from common import * # Set the timestamp timestamp = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S") scriptPath = os.path.dirname(os.path.realpath(__file__)) outFolderPath = os.getcwd() buildFolderPath = os.getcwd() caseMin = min(miscAugmentationMap.keys()) caseMax = max(miscAugmentationMap.keys()) errorLog = [{"notExecutedFunctionality": 0}] # Get a list of log files based on a flag for preserving output def get_log_file_list(): return [ outFolderPath + "/OUTPUT_PERFORMANCE_MISC_LOGS_HIP_" + timestamp + "/Tensor_misc_hip_raw_performance_log.txt", ] def case_file_check(CASE_FILE_PATH, new_file): try: case_file = open(CASE_FILE_PATH, "r") for line in case_file: print(line) if not (line.startswith('"Name"')): new_file.write(line) case_file.close() return True except IOError: print("Unable to open case results") return False # Generate performance reports based on counters and a list of types def generate_performance_reports(RESULTS_DIR): import pandas as pd pd.options.display.max_rows = None # Generate performance report df = pd.read_csv(RESULTS_DIR + "/consolidated_results.stats.csv") df["AverageMs"] = df["AverageNs"] / 1000000 dfPrint = df.drop(["Percentage"], axis=1) dfPrint["HIP Kernel Name"] = dfPrint.iloc[:, 0].str.lstrip("Hip_") dfPrint_noIndices = dfPrint.astype(str) dfPrint_noIndices.replace(["0", "0.0"], "", inplace=True) dfPrint_noIndices = dfPrint_noIndices.to_string(index=False) print(dfPrint_noIndices) def run_unit_test_cmd( numDims, case, numRuns, testType, toggle, batchSize, outFilePath, additionalArg ): bitDepths = list(BitDepthTestMode) if testType == TestType.UNIT_TEST.value: bitDepths = [BitDepthTestMode.U8_TO_U8, BitDepthTestMode.F32_TO_F32] if miscAugmentationMap[int(case)][0] == "log": bitDepths = [BitDepthTestMode.U8_TO_F32, BitDepthTestMode.F32_TO_F32] if miscAugmentationMap[int(case)][0] == "log1p": bitDepths = [BitDepthTestMode.I16_TO_F32] if miscAugmentationMap[int(case)][0] == "tensor_divide_tensor": bitDepths = [BitDepthTestMode.F32_TO_F32, BitDepthTestMode.U8_TO_F32] # Tensor AND/OR/XOR (cases 5/6/7) are only executed for integer bit depths if int(case) in (5, 6, 7): bitDepths = [BitDepthTestMode.U8_TO_U8] for bitDepth in bitDepths: print( "\n./Tensor_misc_hip " + str(case) + " " + str(testType) + " " + str(toggle) + " " + str(numDims) + " " + str(batchSize) + " " + str(numRuns) + " " + str(bitDepth) + " " + str(additionalArg) ) result = subprocess.Popen( [ buildFolderPath + "/build/Tensor_misc_hip", str(case), str(testType), str(toggle), str(numDims), str(batchSize), str(numRuns), str(bitDepth.value), str(additionalArg), outFilePath, scriptPath, ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, ) # nosec log_detected( result, errorLog, miscAugmentationMap[int(case)][0], get_bit_depth(int(bitDepth.value)), get_misc_func_name(int(case), numDims, additionalArg), ) print( "------------------------------------------------------------------------------------------" ) def run_performance_test_cmd( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, bitDepth, outFilePath, additionalArg, ): with open( loggingFolder + "/Tensor_misc_hip_raw_performance_log.txt", "a" ) as logFile: logFile.write( "./Tensor_misc_hip " + str(case) + " " + str(testType) + " " + str(toggle) + " " + str(numDims) + " " + str(batchSize) + " " + str(numRuns) + " " + str(bitDepth) + " " + str(additionalArg) + "\n" ) process = subprocess.Popen( [ buildFolderPath + "/build/Tensor_misc_hip", str(case), str(testType), str(toggle), str(numDims), str(batchSize), str(numRuns), str(bitDepth), str(additionalArg), outFilePath, scriptPath, ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, ) # nosec read_from_subprocess_and_write_to_log(process, logFile) log_detected( process, errorLog, miscAugmentationMap[int(case)][0], get_bit_depth(int(bitDepth)), get_misc_func_name(int(case), numDims, additionalArg), ) def run_performance_test_with_profiler_cmd( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, bitDepth, outFilePath, additionalArg, ): if not os.path.exists(outFilePath + "/case_" + str(case)): os.mkdir(outFilePath + "/case_" + str(case)) with open( loggingFolder + "/Tensor_misc_hip_raw_performance_log.txt", "a" ) as logFile: logFile.write( "\nrocprof --basenames on --timestamp on --stats -o " + outFilePath + "/case_" + str(case) + "/output_case" + str(case) + ".csv ./Tensor_misc_hip " + str(case) + " " + str(testType) + " " + str(toggle) + " " + str(numDims) + " " + str(batchSize) + " " + str(numRuns) + " " + str(additionalArg) + "\n" ) process = subprocess.Popen( [ "rocprof", "--basenames", "on", "--timestamp", "on", "--stats", "-o", outFilePath + "/case_" + str(case) + "/output_case" + str(case) + ".csv", "./Tensor_misc_hip", str(case), str(testType), str(toggle), str(numDims), str(batchSize), str(numRuns), str(bitDepth), str(additionalArg), outFilePath, scriptPath, ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, ) # nosec read_from_subprocess_and_write_to_log(process, logFile) log_detected( process, errorLog, miscAugmentationMap[int(case)][0], get_bit_depth(int(bitDepth)), get_misc_func_name(int(case), numDims, additionalArg), ) print( "------------------------------------------------------------------------------------------" ) def run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, additionalArg, profilingOption="NO", ): bitDepths = [0, 2] if int(case) == 2: bitDepths = [2, 4] elif int(case) == 4: bitDepths = [11] elif int(case) in (5, 6, 7) and testType == TestType.PERFORMANCE_TEST.value: bitDepths = [0, 5, 7, 8, 9, 10] elif testType == TestType.PERFORMANCE_TEST.value: bitDepths = [0, 1, 2, 5] if testType == TestType.UNIT_TEST.value: run_unit_test_cmd( numDims, case, numRuns, testType, toggle, batchSize, outFilePath, additionalArg, ) elif testType == TestType.PERFORMANCE_TEST.value: # U8, F16, F32 and I8 bit depths available for augmentations. Log and Log1p cases customized to run with available bit depths bitDepths = [ BitDepthTestMode.U8_TO_U8, BitDepthTestMode.F32_TO_F32, BitDepthTestMode.F16_TO_F16, BitDepthTestMode.I8_TO_I8, BitDepthTestMode.I16_TO_I16, BitDepthTestMode.U16_TO_U16, BitDepthTestMode.I32_TO_I32, BitDepthTestMode.U32_TO_U32, ] if miscAugmentationMap[int(case)][0] == "log": bitDepths = [ BitDepthTestMode.U8_TO_F32, BitDepthTestMode.F16_TO_F16, BitDepthTestMode.F32_TO_F32, BitDepthTestMode.I8_TO_I8, ] if miscAugmentationMap[int(case)][0] == "log1p": bitDepths = [BitDepthTestMode.I16_TO_F32] if ( miscAugmentationMap[int(case)][0] == "tensor_add_tensor" or miscAugmentationMap[int(case)][0] == "tensor_subtract_tensor" or miscAugmentationMap[int(case)][0] == "tensor_multiply_tensor" ): bitDepths = [ BitDepthTestMode.U8_TO_U8, BitDepthTestMode.F16_TO_F16, BitDepthTestMode.F32_TO_F32, BitDepthTestMode.I8_TO_I8, BitDepthTestMode.I16_TO_I16, BitDepthTestMode.U16_TO_U16, BitDepthTestMode.I32_TO_I32, BitDepthTestMode.U32_TO_U32, ] if miscAugmentationMap[int(case)][0] == "tensor_divide_tensor": bitDepths = [ BitDepthTestMode.U8_TO_F32, BitDepthTestMode.F32_TO_F32, BitDepthTestMode.F16_TO_F16, BitDepthTestMode.U16_TO_F32, BitDepthTestMode.U32_TO_F32, BitDepthTestMode.I8_TO_F32, BitDepthTestMode.I16_TO_F32, BitDepthTestMode.I32_TO_F32, ] if profilingOption == "NO": for bitDepth in bitDepths: run_performance_test_cmd( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, bitDepth.value, outFilePath, additionalArg, ) elif profilingOption == "YES": for bitDepth in bitDepths: run_performance_test_with_profiler_cmd( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, bitDepth.value, outFilePath, additionalArg, ) # Parse and validate command-line arguments for the RPP test suite def rpp_test_suite_parser_and_validator(): parser = argparse.ArgumentParser() parser.add_argument( "--case_start", type=int, default=caseMin, help="Testing start case # - Range must be in [" + str(caseMin) + ":" + str(caseMax) + "]", ) parser.add_argument( "--case_end", type=int, default=caseMax, help="Testing start case # - Range must be in [" + str(caseMin) + ":" + str(caseMax) + "]", ) parser.add_argument( "--test_type", type=int, default=0, help="Type of Test - (0 = QA tests / 1 = Performance tests)", ) parser.add_argument("--toggle", type=int, default=0, help="Toggle outputs") parser.add_argument( "--case_list", nargs="+", help="A list of specific case numbers to run separated by spaces", required=False, ) parser.add_argument( "--num_dims_list", type=int, nargs="+", default=[2, 3, 4], help="List of input dimensions (e.g., 2 3 4)", ) parser.add_argument( "--num_runs", type=int, default=1, help="Specifies the number of runs for running the performance tests", ) parser.add_argument( "--profiling", type=str, default="NO", help="Run with profiler? - (YES/NO)", required=False, ) parser.add_argument( "--qa_mode", type=int, default=0, help="Run with qa_mode? Outputs from tests will be compared with golden outputs - (0 / 1)", required=False, ) parser.add_argument( "--batch_size", type=int, default=1, help="Specifies the batch size to use for running tests. Default is 1.", ) parser.add_argument( "--broadcast", type=int, nargs="+", default=[0, 1, 2], help="Specifies the broadcast mode to be used (0 = Non-broadcast / 1 = second tensor broadcasted / 2 = First tensor broadcasted). Default is [0, 1, 2].", ) parser.add_argument( "--preserve_output", type=int, default=1, help="preserves the output of the program - (0 = override output / 1 = preserve output )", ) print_case_list(miscAugmentationMap, "HIP", parser) args = parser.parse_args() # validate the parameters passed by user if (args.case_start < caseMin or args.case_start > caseMax) or ( args.case_end < caseMin or args.case_end > caseMax ): print( "Starting case# and Ending case# must be in the [" + str(caseMin) + ":" + str(caseMax) + "] range. Aborting!" ) exit(0) elif args.case_end < args.case_start: print("Ending case# must be greater than starting case#. Aborting!") exit(0) elif args.test_type < 0 or args.test_type > 1: print("Test Type# must be in the 0 / 1. Aborting!") exit(0) elif args.qa_mode < 0 or args.qa_mode > 1: print("QA mode must be in the 0 / 1. Aborting!") exit(0) elif ( args.case_list is not None and args.case_start > caseMin and args.case_end < caseMax ): print( "Invalid input! Please provide only 1 option between case_list, case_start and case_end" ) exit(0) elif args.num_runs <= 0: print("Number of Runs must be greater than 0. Aborting!") exit(0) elif args.batch_size <= 0: print("Batch size must be greater than 0. Aborting!") exit(0) elif args.profiling != "YES" and args.profiling != "NO": print("Profiling option value must be either 'YES' or 'NO'.") exit(0) case_list = [] if args.case_list: for case in args.case_list: try: case_number = get_case_number(miscAugmentationMap, case) case_list.append(case_number) except ValueError as e: print(e) args.case_list = case_list if args.case_list is None or len(args.case_list) == 0: args.case_list = range(args.case_start, args.case_end + 1) args.case_list = [str(x) for x in args.case_list] else: for case in args.case_list: if int(case) < caseMin or int(case) > caseMax: print( "The case# must be in [" + str(caseMin) + ":" + str(caseMax) + "]" ) exit(0) return args args = rpp_test_suite_parser_and_validator() caseStart = args.case_start caseEnd = args.case_end testType = args.test_type toggle = args.toggle caseList = args.case_list numDimsList = args.num_dims_list numRuns = args.num_runs profilingOption = args.profiling batchSize = args.batch_size qaMode = args.qa_mode if qaMode: testType = 0 broadcast = args.broadcast preserveOutput = args.preserve_output outFilePath = " " if testType == TestType.UNIT_TEST.value and batchSize != 3: print("QA mode can only run with a batch size of 3.") exit(0) if preserveOutput == 0: validate_and_remove_folders(outFolderPath, "QA_RESULTS_MISC_HIP") validate_and_remove_folders(outFolderPath, "OUTPUT_PERFORMANCE_MISC_LOGS_HIP") if testType == TestType.UNIT_TEST.value: outFilePath = outFolderPath + "/QA_RESULTS_MISC_HIP_" + timestamp numRuns = 1 elif testType == TestType.PERFORMANCE_TEST.value: if "--num_runs" not in sys.argv: numRuns = 100 # default numRuns for running performance tests outFilePath = outFolderPath + "/OUTPUT_PERFORMANCE_MISC_LOGS_HIP_" + timestamp else: print( "Invalid TEST_TYPE specified. TEST_TYPE should be 0/1 (0 = QA tests / 1 = Performance tests)" ) exit(0) os.mkdir(outFilePath) loggingFolder = outFilePath dstPath = outFilePath # Enable extglob if os.path.exists(buildFolderPath + "/build"): shutil.rmtree(buildFolderPath + "/build") os.makedirs(buildFolderPath + "/build") os.chdir(buildFolderPath + "/build") # Run cmake and make commands run_cmake_build(scriptPath) supportedCaseList = [ key for key, values in miscAugmentationMap.items() if "HIP" in values ] noCaseSupported = all(int(case) not in supportedCaseList for case in caseList) if noCaseSupported: print("\ncase numbers %s are not supported" % caseList) exit(0) for case in caseList: if int(case) not in miscAugmentationMap: continue for numDims in numDimsList: # Runs transpose functionality for all transposeOrder values ranging from 1 to numDims - 1 if miscAugmentationMap[int(case)][0] == "transpose": for transposeOrder in range(1, numDims): run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, transposeOrder, profilingOption, ) # Runs normalize functionality for all axisMask values - 1 to 2^numDims - 1 elif miscAugmentationMap[int(case)][0] == "normalize": for axisMask in range(1, pow(2, numDims)): run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, axisMask, profilingOption, ) # Runs concat functionality for all axisMask values - 0 to numDims - 1 elif miscAugmentationMap[int(case)][0] == "concat": for axisMask in range(0, numDims): run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, axisMask, profilingOption, ) # Runs broadcast functionality for all broadcast values elif miscAugmentationMap[int(case)][0] in [ "tensor_and_tensor", "tensor_or_tensor", "tensor_xor_tensor", "tensor_add_tensor", "tensor_subtract_tensor", "tensor_multiply_tensor", "tensor_divide_tensor", ]: for broadcastFlag in broadcast: run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, broadcastFlag, profilingOption, ) # Runs all other functionalities else: run_test( loggingFolder, numDims, case, numRuns, testType, toggle, batchSize, outFilePath, "", profilingOption, ) if testType == TestType.PERFORMANCE_TEST.value and profilingOption == "YES": RESULTS_DIR = outFolderPath + "/OUTPUT_PERFORMANCE_MISC_LOGS_HIP_" + timestamp print("RESULTS_DIR = " + RESULTS_DIR) CONSOLIDATED_FILE = RESULTS_DIR + "/consolidated_results.stats.csv" CASE_NUM_LIST = caseList BIT_DEPTH_LIST = [2] OFT_LIST = [0] # Open csv file new_file = open(CONSOLIDATED_FILE, "w") new_file.write( '"HIP Kernel Name","Calls","TotalDurationNs","AverageNs","Percentage"\n' ) # Loop through cases for CASE_NUM in CASE_NUM_LIST: # Set results directory CASE_RESULTS_DIR = RESULTS_DIR + "/case_" + str(CASE_NUM) print("CASE_RESULTS_DIR = " + CASE_RESULTS_DIR) # Loop through bit depths for BIT_DEPTH in BIT_DEPTH_LIST: # Loop through output format toggle cases for OFT in OFT_LIST: # Write into csv file CASE_FILE_PATH = ( CASE_RESULTS_DIR + "/output_case" + str(CASE_NUM) + ".stats.csv" ) print("CASE_FILE_PATH = " + CASE_FILE_PATH) fileCheck = case_file_check(CASE_FILE_PATH, new_file) if fileCheck == False: continue new_file.close() subprocess.call( ["chown", str(os.getuid()) + ":" + str(os.getgid()), CONSOLIDATED_FILE] ) # nosec try: generate_performance_reports(RESULTS_DIR) except ImportError: print( "\nPandas not available! Results of GPU profiling experiment are available in the following files:\n" + CONSOLIDATED_FILE + "\n" ) except IOError: print("Unable to open results in " + CONSOLIDATED_FILE) # print the results of qa tests nonQACaseList = [] supportedCases = 0 for num in caseList: if int(num) in miscAugmentationMap: supportedCases += 1 caseInfo = ( "Tests are run for " + str(supportedCases) + " supported cases out of the " + str(len(caseList)) + " cases requested" ) if testType == TestType.UNIT_TEST.value: qaFilePath = os.path.join(outFilePath, "QA_results.txt") checkFile = os.path.isfile(qaFilePath) if checkFile: print( "---------------------------------- Results of QA Test - Tensor_misc_hip ----------------------------------\n" ) print_qa_tests_summary( qaFilePath, supportedCaseList, nonQACaseList, "Tensor_misc_hip" ) # Performance tests if testType == TestType.PERFORMANCE_TEST.value and profilingOption == "NO": logFileList = get_log_file_list() functionalityGroupList = ["statiscal_operations"] for logFile in logFileList: print_performance_tests_summary(logFile, functionalityGroupList, numRuns) if len(errorLog) > 1 or errorLog[0]["notExecutedFunctionality"] != 0: print( "\n---------------------------------- Log of function variants requested but not run - Tensor_misc_hip ----------------------------------\n" ) for i in range(1, len(errorLog)): print(errorLog[i]) if errorLog[0]["notExecutedFunctionality"] != 0: print( str(errorLog[0]["notExecutedFunctionality"]) + " functionality variants requested by test_suite_misc_hip were not executed since these sub-variants are not currently supported in RPP.\n" ) print( "-----------------------------------------------------------------------------------------------" )