visualization_utils/plotting-ying.py [1420:2096]: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads/nav_breakdown") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.tight_layout() plt.savefig(os.path.join(output_dir,'_'.join(column_name.split(" "))+".png"), bbox_inches='tight') # plt.show() plt.close('all') column_column_value_experiment_frequency_dict[column_name] = copy.deepcopy(column_value_experiment_frequency_dict) """ # multi-stack plot here index = experiment_names plotdata = pd.DataFrame(column_column_value_experiment_frequency_dict, index=index) df_g = plotdata.groupby(["transformation_metric", "move name"]) plotdata.plot(kind='bar', stacked=True, figsize=(12, 10)) plt.rc('font', **axis_font) plt.xlabel("experiments", **axis_font) plt.ylabel(column_name, **axis_font) plt.xticks(fontsize=fontSize, rotation=45) plt.yticks(fontsize=fontSize) plt.title("experiment vs " + column_name, **axis_font) plt.legend(bbox_to_anchor=(1, 1), loc='upper left', fontsize=fontSize) # dump in the top folder output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads/nav_breakdown") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.tight_layout() plt.savefig(os.path.join(output_dir,'column____'.join(column_name.split(" "))+".png"), bbox_inches='tight') # plt.show() plt.close('all') """ return column_column_value_experiment_frequency_dict # the function to plot distance to goal vs. iteration cnt def plotDistToGoalVSitr(input_dir_names, all_res_column_name_number): itrColNum = all_res_column_name_number["iteration cnt"] distColNum = all_res_column_name_number["dist_to_goal_non_cost"] trueNum = all_res_column_name_number["move validity"] experiment_itr_dist_to_goal_dict = {} # iterate through directories, get data and store in a dictionary for dir_name in input_dir_names: itr = [] distToGoal = [] file_full_addr = os.path.join(dir_name, "result_summary/FARSI_simple_run_0_1_all_reults.csv") with open(file_full_addr, newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') experiment_name = get_experiments_name(file_full_addr, all_res_column_name_number) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i > 1: itr.append(int(row[itrColNum])) distToGoal.append(float(row[distColNum])) experiment_itr_dist_to_goal_dict[experiment_name] = (itr[:], distToGoal[:]) plt.figure() # iterate and plot for experiment_name, value in experiment_itr_dist_to_goal_dict.items(): itr, distToGoal = value[0], value[1] if len(itr) == 0 or len(distToGoal) == 0: # no valid move continue plt.plot(itr, distToGoal, label=experiment_name) plt.xlabel("Iteration Cnt") plt.ylabel("Distance to Goal") plt.title("Distance to Goal vs. Iteration Cnt") # decide on the output dir if len(input_dir_names) == 1: output_dir = input_dir_names[0] else: # dump in the top folder output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.savefig(os.path.join(output_dir, "distToGoalVSitr.png")) # plt.show() plt.close('all') # the function to plot distance to goal vs. iteration cnt def plotRefDistToGoalVSitr(dirName, fileName, itrColNum, refDistColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') itr = [] refDistToGoal = [] for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i > 1: itr.append(int(row[itrColNum])) refDistToGoal.append(float(row[refDistColNum])) plt.figure() plt.plot(itr, refDistToGoal) plt.xlabel("Iteration Cnt") plt.ylabel("Reference Design Distance to Goal") plt.title("Reference Design Distance to Goal vs. Iteration Cnt") plt.savefig(dirName + fileName + "/refDistToGoalVSitr-" + fileName + ".png") # plt.show() plt.close('all') # the function to do the zonal partitioning def zonalPartition(comparedValue, zoneNum, maxValue): unit = maxValue / zoneNum if comparedValue > maxValue: return zoneNum - 1 if comparedValue < 0: return 0 for i in range(0, zoneNum): if comparedValue <= unit * (i + 1): return i raise Exception("zonalPartition is fed by a strange value! maxValue: " + str(maxValue) + "; comparedValue: " + str(comparedValue)) # the function to plot simulation time vs. move name in a zonal format def plotSimTimeVSmoveNameZoneDist(dirName, fileName, zoneNum, moveColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') splitSwapSim = np.zeros(zoneNum, dtype = float) splitSim = np.zeros(zoneNum, dtype = float) migrateSim = np.zeros(zoneNum, dtype = float) swapSim = np.zeros(zoneNum, dtype = float) tranSim = np.zeros(zoneNum, dtype = float) routeSim = np.zeros(zoneNum, dtype = float) identitySim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): # print('"' + row[trueNum] + '"\t"' + row[moveColNum] + '"\t"' + row[distColNum] + '"\t"' + row[simColNum] + '"') if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[moveColNum] == "split_swap": splitSwapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "split": splitSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "migrate": migrateSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "swap": swapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "transfer": tranSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "routing": routeSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "identity": identitySim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("move name is not split_swap or split or migrate or swap or transfer or routing or identity! The new type: " + row[moveColNum]) # plt.figure() plotdata = pd.DataFrame({ "split_swap":splitSwapSim, "split":splitSim, "migrate":migrateSim, "swap":swapSim, "transfer":tranSim, "routing":routeSim, "identity":identitySim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Move Name") plt.savefig(dirName + fileName + "/simTimeVSmoveNameZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot move generation time vs. move name in a zonal format def plotMovGenTimeVSmoveNameZoneDist(dirName, fileName, zoneNum, moveColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') splitSwapMov = np.zeros(zoneNum, dtype = float) splitMov = np.zeros(zoneNum, dtype = float) migrateMov = np.zeros(zoneNum, dtype = float) swapMov = np.zeros(zoneNum, dtype = float) tranMov = np.zeros(zoneNum, dtype = float) routeMov = np.zeros(zoneNum, dtype = float) identityMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): # print('"' + row[trueNum] + '"\t"' + row[moveColNum] + '"\t"' + row[distColNum] + '"\t"' + row[movGenColNum] + '"') if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[moveColNum] == "split_swap": splitSwapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "split": splitMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "migrate": migrateMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "swap": swapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "transfer": tranMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "routing": routeMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "identity": identityMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("move name is not split_swap or split or migrate or swap or transfer of routing or identity! The new type: " + row[moveColNum]) # plt.figure() plotdata = pd.DataFrame({ "split_swap":splitSwapMov, "split":splitMov, "migrate":migrateMov, "swap":swapMov, "transfer":tranMov, "routing":routeMov, "identity":identityMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Move Generation Time") plt.title("Move Generation Time in Each Zone based on Move Name") plt.savefig(dirName + fileName + "/movGenTimeVSmoveNameZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. comm_comp in a zonal format def plotSimTimeVScommCompZoneDist(dirName, fileName, zoneNum, commcompColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') commSim = np.zeros(zoneNum, dtype = float) compSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[commcompColNum] == "comm": commSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[commcompColNum] == "comp": compSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("comm_comp is not giving comm or comp! The new type: " + row[colNum]) # plt.figure() plotdata = pd.DataFrame({ "comm":commSim, "comp":compSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on comm_comp") plt.savefig(dirName + fileName + "/simTimeVScommCompZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. comm_comp in a zonal format def plotMovGenTimeVScommCompZoneDist(dirName, fileName, zoneNum, commcompColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') commMov = np.zeros(zoneNum, dtype = float) compMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[commcompColNum] == "comm": commMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[commcompColNum] == "comp": compMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("comm_comp is not giving comm or comp! The new type: " + row[colNum]) # plt.figure() plotdata = pd.DataFrame({ "comm":commMov, "comp":compMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Move Generation Time") plt.title("Move Generation Time in Each Zone based on comm_comp") plt.savefig(dirName + fileName + "/movGenTimeVScommCompZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. high level optimization name in a zonal format def plotSimTimeVShighLevelOptZoneDist(dirName, fileName, zoneNum, optColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') topoSim = np.zeros(zoneNum, dtype = float) tunSim = np.zeros(zoneNum, dtype = float) mapSim = np.zeros(zoneNum, dtype = float) idenOptSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[optColNum] == "topology": topoSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "customization": tunSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "mapping": mapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "identity": idenOptSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("high level optimization name is not giving topology or customization or mapping or identity! The new type: " + row[optColNum]) # plt.figure() plotdata = pd.DataFrame({ "topology":topoSim, "customization":tunSim, "mapping":mapSim, "identity":idenOptSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Optimation Name") plt.savefig(dirName + fileName + "/simTimeVShighLevelOptZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. high level optimization name in a zonal format def plotMovGenTimeVShighLevelOptZoneDist(dirName, fileName, zoneNum, optColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') topoMov = np.zeros(zoneNum, dtype = float) tunMov = np.zeros(zoneNum, dtype = float) mapMov = np.zeros(zoneNum, dtype = float) idenOptMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[optColNum] == "topology": topoMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "customization": tunMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "mapping": mapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "identity": idenOptMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("high level optimization name is not giving topology or customization or mapping or identity! The new type: " + row[optColNum]) # plt.figure() plotdata = pd.DataFrame({ "topology":topoMov, "customization":tunMov, "mapping":mapMov, "identity":idenOptMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Transformation Generation Time") plt.title("Transformation Generation Time in Each Zone based on Optimization Name") plt.savefig(dirName + fileName + "/movGenTimeVShighLevelOptZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. architectural principle in a zonal format def plotSimTimeVSarchVarImpZoneDist(dirName, fileName, zoneNum, archColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') paraSim = np.zeros(zoneNum, dtype = float) custSim = np.zeros(zoneNum, dtype = float) localSim = np.zeros(zoneNum, dtype = float) idenImpSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[archColNum] == "parallelization": paraSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "customization": custSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "locality": localSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "identity": idenImpSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("architectural principle is not giving parallelization or customization or locality or identity! The new type: " + row[archColNum]) # plt.figure() plotdata = pd.DataFrame({ "parallelization":paraSim, "customization":custSim, "locality":localSim, "identity":idenImpSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Architectural Principle") plt.savefig(dirName + fileName + "/simTimeVSarchVarImpZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. architectural principle in a zonal format def plotMovGenTimeVSarchVarImpZoneDist(dirName, fileName, zoneNum, archColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') paraMov = np.zeros(zoneNum, dtype = float) custMov = np.zeros(zoneNum, dtype = float) localMov = np.zeros(zoneNum, dtype = float) idenImpMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[archColNum] == "parallelization": paraMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "customization": custMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "locality": localMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "identity": idenImpMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("architectural principle is not giving parallelization or customization or locality or identity! The new type: " + row[archColNum]) # plt.figure() plotdata = pd.DataFrame({ "parallelization":paraMov, "customization":custMov, "locality":localMov, "identity":idenImpMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Tranformation Generation Time") plt.title("Tranformation Generation Time in Each Zone based on Architectural Principle") plt.savefig(dirName + fileName + "/movGenTimeVSarchVarImpZoneZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot convergence vs. iteration cnt, system block count, and routing complexity in 3d def plotBudgets3d(dirName, subDirName): newDirName = dirName + "/"+ subDirName + "/" if os.path.exists(newDirName + "/figures"): shutil.rmtree(newDirName + "/figures") resultList = os.listdir(newDirName) latBudgets = [] powBudgets = [] areaBudgets = [] itrValues = [] cntValues = [] routingValues = [] workloads = [] for j, fileName in enumerate(resultList): with open(newDirName + fileName + "/result_summary/FARSI_simple_run_0_1.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') for i, row in enumerate(resultReader): if i == 1: itrValues.append(int(row[columnNum(newDirName, fileName, "iteration cnt", "simple")])) cntValues.append(int(row[columnNum(newDirName, fileName, "system block count", "simple")])) routingValues.append(float(row[columnNum(newDirName, fileName, "routing complexity", "simple")])) powBudgets.append(float(row[columnNum(newDirName, fileName, "power_budget", "simple")])) areaBudgets.append(float(row[columnNum(newDirName, fileName, "area_budget", "simple")])) lat = row[int(columnNum(newDirName, fileName, "latency_budget", "simple"))][:-1] latDict = dict(item.split("=") for item in lat.split(";")) if j == 0: for k in range(0, len(latDict)): latBudgets.append([]) workloads.append(list(latDict.keys())[k]) latList = list(latDict.values()) for k in range(0, len(latList)): latBudgets[k].append(float(latList[k])) m = ['o', 'x', '^', 's', 'd', '+', 'v', '<', '>'] axis_font = {'size': '10'} fontSize = 10 os.mkdir(newDirName + "figures") fig_budget_itr = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_itr = fig_budget_itr.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_itr.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=itrValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(itrValues[j]) ax_itr.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_itr.set_xlabel("Power Budget") ax_itr.set_ylabel("Area Budget") ax_itr.set_zlabel("Latency Budget") ax_itr.legend() cbar_itr = fig_budget_itr.colorbar(img, aspect = 40) cbar_itr.set_label("Number of Iterations", rotation = 270) plt.title("{Power Budget, Area Budget, Latency Budget} VS Iteration Cnt: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSitr-" + subDirName + ".png") # plt.show() plt.close('all') fig_budget_blkcnt = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_blkcnt = fig_budget_blkcnt.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_blkcnt.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=cntValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(cntValues[j]) ax_blkcnt.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_blkcnt.set_xlabel("Power Budget") ax_blkcnt.set_ylabel("Area Budget") ax_blkcnt.set_zlabel("Latency Budget") ax_blkcnt.legend() cbar = fig_budget_blkcnt.colorbar(img, aspect=40) cbar.set_label("System Block Count", rotation=270) plt.title("{Power Budget, Area Budget, Latency Budget} VS System Block Count: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSblkcnt-" + subDirName + ".png") # plt.show() plt.close('all') fig_budget_routing = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_routing = fig_budget_routing.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_routing.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=cntValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(routingValues[j]) ax_routing.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_routing.set_xlabel("Power Budget") ax_routing.set_ylabel("Area Budget") ax_routing.set_zlabel("Latency Budget") ax_routing.legend() cbar = fig_budget_routing.colorbar(img, aspect=40) cbar.set_label("System Block Count", rotation=270) plt.title("{Power Budget, Area Budget, Latency Budget} VS System Block Count: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSroutingComplexity-" + subDirName + ".png") # plt.show() plt.close('all') def get_experiment_dir_list(run_folder_name): workload_set_folder_list = os.listdir(run_folder_name) experiment_full_addr_list = [] # iterate and generate plots for workload_set_folder in workload_set_folder_list: # ignore irelevant files if workload_set_folder in config_plotting.ignore_file_names: continue # get experiment folder workload_set_full_addr = os.path.join(run_folder_name,workload_set_folder) folder_list = os.listdir(workload_set_full_addr) for experiment_name_relative_addr in folder_list: if experiment_name_relative_addr in config_plotting.ignore_file_names: continue experiment_full_addr_list.append(os.path.join(workload_set_full_addr, experiment_name_relative_addr)) return experiment_full_addr_list def find_the_most_recent_directory(top_dir): dirs = [os.path.join(top_dir, el) for el in os.listdir(top_dir)] dirs = list(filter(os.path.isdir, dirs)) dirs.sort(key=lambda x: os.path.getmtime(x), reverse=True) return dirs def get_experiment_full_file_addr_list(experiment_full_dir_list): file_name = "result_summary/FARSI_simple_run_0_1.csv" results = [] for el in experiment_full_dir_list: results.append(os.path.join(el, file_name)) return results - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - visualization_utils/plotting.py [2259:2933]: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads/nav_breakdown") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.tight_layout() plt.savefig(os.path.join(output_dir,'_'.join(column_name.split(" "))+".png"), bbox_inches='tight') # plt.show() plt.close('all') column_column_value_experiment_frequency_dict[column_name] = copy.deepcopy(column_value_experiment_frequency_dict) """ # multi-stack plot here index = experiment_names plotdata = pd.DataFrame(column_column_value_experiment_frequency_dict, index=index) df_g = plotdata.groupby(["transformation_metric", "move name"]) plotdata.plot(kind='bar', stacked=True, figsize=(12, 10)) plt.rc('font', **axis_font) plt.xlabel("experiments", **axis_font) plt.ylabel(column_name, **axis_font) plt.xticks(fontsize=fontSize, rotation=45) plt.yticks(fontsize=fontSize) plt.title("experiment vs " + column_name, **axis_font) plt.legend(bbox_to_anchor=(1, 1), loc='upper left', fontsize=fontSize) # dump in the top folder output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads/nav_breakdown") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.tight_layout() plt.savefig(os.path.join(output_dir,'column____'.join(column_name.split(" "))+".png"), bbox_inches='tight') # plt.show() plt.close('all') """ return column_column_value_experiment_frequency_dict # the function to plot distance to goal vs. iteration cnt def plotDistToGoalVSitr(input_dir_names, all_res_column_name_number): itrColNum = all_res_column_name_number["iteration cnt"] distColNum = all_res_column_name_number["dist_to_goal_non_cost"] trueNum = all_res_column_name_number["move validity"] experiment_itr_dist_to_goal_dict = {} # iterate through directories, get data and store in a dictionary for dir_name in input_dir_names: itr = [] distToGoal = [] file_full_addr = os.path.join(dir_name, "result_summary/FARSI_simple_run_0_1_all_reults.csv") with open(file_full_addr, newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') experiment_name = get_experiments_name(file_full_addr, all_res_column_name_number) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i > 1: itr.append(int(row[itrColNum])) distToGoal.append(float(row[distColNum])) experiment_itr_dist_to_goal_dict[experiment_name] = (itr[:], distToGoal[:]) plt.figure() # iterate and plot for experiment_name, value in experiment_itr_dist_to_goal_dict.items(): itr, distToGoal = value[0], value[1] if len(itr) == 0 or len(distToGoal) == 0: # no valid move continue plt.plot(itr, distToGoal, label=experiment_name) plt.xlabel("Iteration Cnt") plt.ylabel("Distance to Goal") plt.title("Distance to Goal vs. Iteration Cnt") # decide on the output dir if len(input_dir_names) == 1: output_dir = input_dir_names[0] else: # dump in the top folder output_base_dir = '/'.join(input_dir_names[0].split("/")[:-2]) output_dir = os.path.join(output_base_dir, "cross_workloads") if not os.path.exists(output_dir): os.makedirs(output_dir) plt.savefig(os.path.join(output_dir, "distToGoalVSitr.png")) # plt.show() plt.close('all') # the function to plot distance to goal vs. iteration cnt def plotRefDistToGoalVSitr(dirName, fileName, itrColNum, refDistColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') itr = [] refDistToGoal = [] for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i > 1: itr.append(int(row[itrColNum])) refDistToGoal.append(float(row[refDistColNum])) plt.figure() plt.plot(itr, refDistToGoal) plt.xlabel("Iteration Cnt") plt.ylabel("Reference Design Distance to Goal") plt.title("Reference Design Distance to Goal vs. Iteration Cnt") plt.savefig(dirName + fileName + "/refDistToGoalVSitr-" + fileName + ".png") # plt.show() plt.close('all') # the function to do the zonal partitioning def zonalPartition(comparedValue, zoneNum, maxValue): unit = maxValue / zoneNum if comparedValue > maxValue: return zoneNum - 1 if comparedValue < 0: return 0 for i in range(0, zoneNum): if comparedValue <= unit * (i + 1): return i raise Exception("zonalPartition is fed by a strange value! maxValue: " + str(maxValue) + "; comparedValue: " + str(comparedValue)) # the function to plot simulation time vs. move name in a zonal format def plotSimTimeVSmoveNameZoneDist(dirName, fileName, zoneNum, moveColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') splitSwapSim = np.zeros(zoneNum, dtype = float) splitSim = np.zeros(zoneNum, dtype = float) migrateSim = np.zeros(zoneNum, dtype = float) swapSim = np.zeros(zoneNum, dtype = float) tranSim = np.zeros(zoneNum, dtype = float) routeSim = np.zeros(zoneNum, dtype = float) identitySim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): # print('"' + row[trueNum] + '"\t"' + row[moveColNum] + '"\t"' + row[distColNum] + '"\t"' + row[simColNum] + '"') if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[moveColNum] == "split_swap": splitSwapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "split": splitSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "migrate": migrateSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "swap": swapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "transfer": tranSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "routing": routeSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[moveColNum] == "identity": identitySim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("move name is not split_swap or split or migrate or swap or transfer or routing or identity! The new type: " + row[moveColNum]) # plt.figure() plotdata = pd.DataFrame({ "split_swap":splitSwapSim, "split":splitSim, "migrate":migrateSim, "swap":swapSim, "transfer":tranSim, "routing":routeSim, "identity":identitySim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Move Name") plt.savefig(dirName + fileName + "/simTimeVSmoveNameZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot move generation time vs. move name in a zonal format def plotMovGenTimeVSmoveNameZoneDist(dirName, fileName, zoneNum, moveColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') splitSwapMov = np.zeros(zoneNum, dtype = float) splitMov = np.zeros(zoneNum, dtype = float) migrateMov = np.zeros(zoneNum, dtype = float) swapMov = np.zeros(zoneNum, dtype = float) tranMov = np.zeros(zoneNum, dtype = float) routeMov = np.zeros(zoneNum, dtype = float) identityMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): # print('"' + row[trueNum] + '"\t"' + row[moveColNum] + '"\t"' + row[distColNum] + '"\t"' + row[movGenColNum] + '"') if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[moveColNum] == "split_swap": splitSwapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "split": splitMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "migrate": migrateMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "swap": swapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "transfer": tranMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "routing": routeMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[moveColNum] == "identity": identityMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("move name is not split_swap or split or migrate or swap or transfer of routing or identity! The new type: " + row[moveColNum]) # plt.figure() plotdata = pd.DataFrame({ "split_swap":splitSwapMov, "split":splitMov, "migrate":migrateMov, "swap":swapMov, "transfer":tranMov, "routing":routeMov, "identity":identityMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Move Generation Time") plt.title("Move Generation Time in Each Zone based on Move Name") plt.savefig(dirName + fileName + "/movGenTimeVSmoveNameZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. comm_comp in a zonal format def plotSimTimeVScommCompZoneDist(dirName, fileName, zoneNum, commcompColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') commSim = np.zeros(zoneNum, dtype = float) compSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[commcompColNum] == "comm": commSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[commcompColNum] == "comp": compSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("comm_comp is not giving comm or comp! The new type: " + row[colNum]) # plt.figure() plotdata = pd.DataFrame({ "comm":commSim, "comp":compSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on comm_comp") plt.savefig(dirName + fileName + "/simTimeVScommCompZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. comm_comp in a zonal format def plotMovGenTimeVScommCompZoneDist(dirName, fileName, zoneNum, commcompColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') commMov = np.zeros(zoneNum, dtype = float) compMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[commcompColNum] == "comm": commMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[commcompColNum] == "comp": compMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("comm_comp is not giving comm or comp! The new type: " + row[colNum]) # plt.figure() plotdata = pd.DataFrame({ "comm":commMov, "comp":compMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Move Generation Time") plt.title("Move Generation Time in Each Zone based on comm_comp") plt.savefig(dirName + fileName + "/movGenTimeVScommCompZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. high level optimization name in a zonal format def plotSimTimeVShighLevelOptZoneDist(dirName, fileName, zoneNum, optColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') topoSim = np.zeros(zoneNum, dtype = float) tunSim = np.zeros(zoneNum, dtype = float) mapSim = np.zeros(zoneNum, dtype = float) idenOptSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[optColNum] == "topology": topoSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "customization": tunSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "mapping": mapSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[optColNum] == "identity": idenOptSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("high level optimization name is not giving topology or customization or mapping or identity! The new type: " + row[optColNum]) # plt.figure() plotdata = pd.DataFrame({ "topology":topoSim, "customization":tunSim, "mapping":mapSim, "identity":idenOptSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Optimation Name") plt.savefig(dirName + fileName + "/simTimeVShighLevelOptZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. high level optimization name in a zonal format def plotMovGenTimeVShighLevelOptZoneDist(dirName, fileName, zoneNum, optColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') topoMov = np.zeros(zoneNum, dtype = float) tunMov = np.zeros(zoneNum, dtype = float) mapMov = np.zeros(zoneNum, dtype = float) idenOptMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[optColNum] == "topology": topoMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "customization": tunMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "mapping": mapMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[optColNum] == "identity": idenOptMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("high level optimization name is not giving topology or customization or mapping or identity! The new type: " + row[optColNum]) # plt.figure() plotdata = pd.DataFrame({ "topology":topoMov, "customization":tunMov, "mapping":mapMov, "identity":idenOptMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Transformation Generation Time") plt.title("Transformation Generation Time in Each Zone based on Optimization Name") plt.savefig(dirName + fileName + "/movGenTimeVShighLevelOptZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. architectural principle in a zonal format def plotSimTimeVSarchVarImpZoneDist(dirName, fileName, zoneNum, archColNum, distColNum, simColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') paraSim = np.zeros(zoneNum, dtype = float) custSim = np.zeros(zoneNum, dtype = float) localSim = np.zeros(zoneNum, dtype = float) idenImpSim = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[archColNum] == "parallelization": paraSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "customization": custSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "locality": localSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) elif row[archColNum] == "identity": idenImpSim[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[simColNum]) else: raise Exception("architectural principle is not giving parallelization or customization or locality or identity! The new type: " + row[archColNum]) # plt.figure() plotdata = pd.DataFrame({ "parallelization":paraSim, "customization":custSim, "locality":localSim, "identity":idenImpSim }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Simulation Time") plt.title("Simulation Time in Each Zone based on Architectural Principle") plt.savefig(dirName + fileName + "/simTimeVSarchVarImpZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot simulation time vs. architectural principle in a zonal format def plotMovGenTimeVSarchVarImpZoneDist(dirName, fileName, zoneNum, archColNum, distColNum, movGenColNum, trueNum): with open(dirName + fileName + "/result_summary/FARSI_simple_run_0_1_all_reults.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') paraMov = np.zeros(zoneNum, dtype = float) custMov = np.zeros(zoneNum, dtype = float) localMov = np.zeros(zoneNum, dtype = float) idenImpMov = np.zeros(zoneNum, dtype = float) maxDist = 0 index = [] for i in range(0, zoneNum): index.append(i) for i, row in enumerate(resultReader): if row[trueNum] != "True": continue if i == 2: maxDist = float(row[distColNum]) if i > 1: if row[archColNum] == "parallelization": paraMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "customization": custMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "locality": localMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) elif row[archColNum] == "identity": idenImpMov[zonalPartition(float(row[distColNum]), zoneNum, maxDist)] += float(row[movGenColNum]) else: raise Exception("architectural principle is not giving parallelization or customization or locality or identity! The new type: " + row[archColNum]) # plt.figure() plotdata = pd.DataFrame({ "parallelization":paraMov, "customization":custMov, "locality":localMov, "identity":idenImpMov }, index = index ) plotdata.plot(kind = 'bar', stacked = True) plt.xlabel("Zone decided by the max distance to goal") plt.ylabel("Tranformation Generation Time") plt.title("Tranformation Generation Time in Each Zone based on Architectural Principle") plt.savefig(dirName + fileName + "/movGenTimeVSarchVarImpZoneZoneDist-" + fileName + ".png") # plt.show() plt.close('all') # the function to plot convergence vs. iteration cnt, system block count, and routing complexity in 3d def plotBudgets3d(dirName, subDirName): newDirName = dirName + "/"+ subDirName + "/" if os.path.exists(newDirName + "/figures"): shutil.rmtree(newDirName + "/figures") resultList = os.listdir(newDirName) latBudgets = [] powBudgets = [] areaBudgets = [] itrValues = [] cntValues = [] routingValues = [] workloads = [] for j, fileName in enumerate(resultList): with open(newDirName + fileName + "/result_summary/FARSI_simple_run_0_1.csv", newline='') as csvfile: resultReader = csv.reader(csvfile, delimiter=',', quotechar='|') for i, row in enumerate(resultReader): if i == 1: itrValues.append(int(row[columnNum(newDirName, fileName, "iteration cnt", "simple")])) cntValues.append(int(row[columnNum(newDirName, fileName, "system block count", "simple")])) routingValues.append(float(row[columnNum(newDirName, fileName, "routing complexity", "simple")])) powBudgets.append(float(row[columnNum(newDirName, fileName, "power_budget", "simple")])) areaBudgets.append(float(row[columnNum(newDirName, fileName, "area_budget", "simple")])) lat = row[int(columnNum(newDirName, fileName, "latency_budget", "simple"))][:-1] latDict = dict(item.split("=") for item in lat.split(";")) if j == 0: for k in range(0, len(latDict)): latBudgets.append([]) workloads.append(list(latDict.keys())[k]) latList = list(latDict.values()) for k in range(0, len(latList)): latBudgets[k].append(float(latList[k])) m = ['o', 'x', '^', 's', 'd', '+', 'v', '<', '>'] axis_font = {'size': '10'} fontSize = 10 os.mkdir(newDirName + "figures") fig_budget_itr = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_itr = fig_budget_itr.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_itr.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=itrValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(itrValues[j]) ax_itr.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_itr.set_xlabel("Power Budget") ax_itr.set_ylabel("Area Budget") ax_itr.set_zlabel("Latency Budget") ax_itr.legend() cbar_itr = fig_budget_itr.colorbar(img, aspect = 40) cbar_itr.set_label("Number of Iterations", rotation = 270) plt.title("{Power Budget, Area Budget, Latency Budget} VS Iteration Cnt: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSitr-" + subDirName + ".png") # plt.show() plt.close('all') fig_budget_blkcnt = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_blkcnt = fig_budget_blkcnt.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_blkcnt.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=cntValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(cntValues[j]) ax_blkcnt.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_blkcnt.set_xlabel("Power Budget") ax_blkcnt.set_ylabel("Area Budget") ax_blkcnt.set_zlabel("Latency Budget") ax_blkcnt.legend() cbar = fig_budget_blkcnt.colorbar(img, aspect=40) cbar.set_label("System Block Count", rotation=270) plt.title("{Power Budget, Area Budget, Latency Budget} VS System Block Count: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSblkcnt-" + subDirName + ".png") # plt.show() plt.close('all') fig_budget_routing = plt.figure(figsize=(12, 12)) plt.rc('font', **axis_font) ax_routing = fig_budget_routing.add_subplot(projection='3d') for i in range(0, len(latBudgets)): img = ax_routing.scatter3D(powBudgets, areaBudgets, latBudgets[i], c=cntValues, cmap="bwr", marker=m[i], s=80, label='{0}'.format(workloads[i])) for j in range(0, len(latBudgets[i])): coordinate = str(routingValues[j]) ax_routing.text(powBudgets[j], areaBudgets[j], latBudgets[i][j], '%s' % coordinate, size=fontSize) break ax_routing.set_xlabel("Power Budget") ax_routing.set_ylabel("Area Budget") ax_routing.set_zlabel("Latency Budget") ax_routing.legend() cbar = fig_budget_routing.colorbar(img, aspect=40) cbar.set_label("System Block Count", rotation=270) plt.title("{Power Budget, Area Budget, Latency Budget} VS System Block Count: " + subDirName) plt.tight_layout() plt.savefig(newDirName + "figures/budgetVSroutingComplexity-" + subDirName + ".png") # plt.show() plt.close('all') def get_experiment_dir_list(run_folder_name): workload_set_folder_list = os.listdir(run_folder_name) experiment_full_addr_list = [] # iterate and generate plots for workload_set_folder in workload_set_folder_list: # ignore irelevant files if workload_set_folder in config_plotting.ignore_file_names: continue # get experiment folder workload_set_full_addr = os.path.join(run_folder_name,workload_set_folder) folder_list = os.listdir(workload_set_full_addr) for experiment_name_relative_addr in folder_list: if experiment_name_relative_addr in config_plotting.ignore_file_names: continue experiment_full_addr_list.append(os.path.join(workload_set_full_addr, experiment_name_relative_addr)) return experiment_full_addr_list def find_the_most_recent_directory(top_dir): dirs = [os.path.join(top_dir, el) for el in os.listdir(top_dir)] dirs = list(filter(os.path.isdir, dirs)) dirs.sort(key=lambda x: os.path.getmtime(x), reverse=True) return dirs def get_experiment_full_file_addr_list(experiment_full_dir_list): file_name = "result_summary/FARSI_simple_run_0_1.csv" results = [] for el in experiment_full_dir_list: results.append(os.path.join(el, file_name)) return results - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -