Index: trunk/platforms/tsar_generic_xbar/scripts/counter_defs.py =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/counter_defs.py (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/counter_defs.py (revision 1048) @@ -11,9 +11,17 @@ m_prot_name['hmesi'] = "HMESI" m_prot_name['wtidl'] = "WT IDEAL" +m_prot_name['th0'] = "TH. 0" m_app_name = {} -m_app_name['mandel'] = "Mandelbrot" +m_app_name['barnes'] = "Barnes" +m_app_name['blackscholes'] = "Blackscholes" +m_app_name['cholesky'] = "Cholesky" +m_app_name['fal_sh_1'] = "False Sharing 1" +m_app_name['fal_sh_2'] = "False Sharing 2" +m_app_name['fft'] = "FFT" +m_app_name['fft_ga'] = "FFT" m_app_name['filter'] = "Convol" m_app_name['filt_ga'] = "Convol" +m_app_name['fmm'] = "FMM" m_app_name['histogram'] = "Histogram" m_app_name['histogram-pad'] = "Histogram" @@ -22,21 +30,19 @@ m_app_name['kmeans'] = "Kmeans" m_app_name['kmeans-opt'] = "Kmeans" +m_app_name['linear_regression'] = "Linear Reg." +m_app_name['lu'] = "LU" +m_app_name['mandel'] = "Mandelbrot" +m_app_name['mat_mult'] = "Matrix Mult." +m_app_name['mat_mult-opt'] = "Matrix Mult." m_app_name['pca'] = "PCA" m_app_name['pca-opt'] = "PCA" -m_app_name['mat_mult'] = "Matrix Mult." -m_app_name['mat_mult-opt'] = "Matrix Mult." -m_app_name['barnes'] = "Barnes" -m_app_name['fmm'] = "FMM" m_app_name['ocean'] = "Ocean" +m_app_name['radix'] = "Radix" +m_app_name['radix_ga'] = "Radix" m_app_name['raytrace'] = "Raytrace" m_app_name['radiosity'] = "Radiosity" +m_app_name['string_match'] = "String Match" m_app_name['waters'] = "Water Sp." m_app_name['watern'] = "Water Nsq." -m_app_name['cholesky'] = "Cholesky" -m_app_name['lu'] = "LU" -m_app_name['fft'] = "FFT" -m_app_name['radix'] = "Radix" -m_app_name['fft_ga'] = "FFT" -m_app_name['radix_ga'] = "Radix" m_metric_id = {} @@ -237,5 +243,5 @@ #colors = [ "\"#008000\"", "\"#000080\"", "\"#BADC98\"", "\"#BA98DC\"", "\"#98DCBA\"", "\"#98BADC\"", "\"#BA9876\"", "\"#BA7698\"", "\"#98BA76\"", "\"#9876BA\"", "\"#76BA98\"", "\"#7698BA\"" ] -colors = [ "\"#008000\"", "\"#FEDCBA\"", "\"#BADC98\"", "\"#000080\"", "\"#98DCBA\"", "\"#98BADC\"", "\"#BA9876\"", "\"#BA7698\"", "\"#98BA76\"", "\"#9876BA\"", "\"#76BA98\"", "\"#7698BA\"" ] - - +colors = [ "\"#008000\"", "\"#FEDCBA\"", "\"#9876BA\"", "\"#000080\"", "\"#98DCBA\"", "\"#98BADC\"", "\"#BA9876\"", "\"#BA7698\"", "\"#98BA76\"", "\"#BADC98\"", "\"#76BA98\"", "\"#7698BA\"" ] + + Index: trunk/platforms/tsar_generic_xbar/scripts/create_graphs.py =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/create_graphs.py (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/create_graphs.py (revision 1048) @@ -7,12 +7,13 @@ +apps = [ 'blackscholes', 'filt_ga', 'fft_ga', 'histo-opt', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'pca-opt', 'radix_ga', 'string_match' ] +#apps = [ 'blackscholes', 'fft_ga', 'filt_ga', 'histo-opt', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'pca-opt', 'radix_ga' ] #apps = [ 'histo-opt', 'mandel', 'filt_ga', 'radix_ga', 'fft_ga', 'pca-opt', 'fft', 'radix', 'filter', 'kmeans-opt' ] -apps = [ 'fft_ga', 'filt_ga', 'lu', 'radix_ga', 'histo-opt', 'mandel', 'pca-opt', 'kmeans-opt' ] #apps = [ 'histogram', 'mandel', 'filter', 'fft', 'fft_ga', 'filt_ga', 'pca', 'lu' ] # radix radix_ga kmeans -#apps = [ 'histo-opt', 'histogram2', 'histo-opt2' ] -#nb_procs = [ 1, 4, 8, 16, 32, 64, 128, 256 ] +#apps = [ 'fal_sh_2' ] nb_procs = [ 1, 4, 8, 16, 32, 64, 128, 256 ] -single_protocols = ['dhccp', 'hmesi'] -joint_protocols = ['dhccp', 'hmesi'] +#nb_procs = [ 256 ] +single_protocols = ['dhccp', 'th0', 'hmesi', 'wtidl'] +joint_protocols = ['dhccp', 'th0', 'hmesi', 'wtidl' ] #joint_protocols = [] @@ -31,8 +32,9 @@ log_term_name = '_term_' -coherence_tmpl = os.path.join(scripts_path, template_dir, 'coherence_template.gp') # 1 graph per appli -speedup_tmpl = os.path.join(scripts_path, template_dir, 'speedup_template.gp') -metric_tmpl = os.path.join(scripts_path, template_dir, 'metric_template.gp') # 1 graph per metric -stacked_tmpl = os.path.join(scripts_path, template_dir, 'stacked_template.gp') +coherence_tmpl = os.path.join(scripts_path, template_dir, 'coherence_template.gp') # 1 graph per appli +speedup_tmpl = os.path.join(scripts_path, template_dir, 'speedup_template.gp') +metric_tmpl = os.path.join(scripts_path, template_dir, 'metric_template.gp') # 1 graph per metric +stacked_tmpl = os.path.join(scripts_path, template_dir, 'stacked_template.gp') +cst_exec_time_tmpl = os.path.join(scripts_path, template_dir, 'cst_exec_time_template.gp') @@ -140,5 +142,6 @@ metrics_val[prot][app][i]['direct_cost'] = metrics_val[prot][app][i]['read_cost'] + metrics_val[prot][app][i]['write_cost'] metrics_val[prot][app][i]['broadcast_cost'] = metrics_val[prot][app][i]['broadcast'] * 2 * (x * y - 1) - metrics_val[prot][app][i]['coherence_cost'] = metrics_val[prot][app][i]['broadcast_cost'] + metrics_val[prot][app][i]['m_inv_cost'] + metrics_val[prot][app][i]['update_cost'] + #metrics_val[prot][app][i]['coherence_cost'] = metrics_val[prot][app][i]['broadcast_cost'] + metrics_val[prot][app][i]['m_inv_cost'] + metrics_val[prot][app][i]['update_cost'] + metrics_val[prot][app][i]['coherence_cost'] = metrics_val[prot][app][i]['broadcast_cost'] + metrics_val[prot][app][i]['m_inv_cost'] + metrics_val[prot][app][i]['update_cost'] + metrics_val[prot][app][i]['cleanup_cost'] + metrics_val[prot][app][i]['cleanup_d_cost'] + metrics_val[prot][app][i]['inval_ro_cost'] if metrics_val[prot][app][i]['broadcast'] < metrics_val[prot][app][i]['write_broadcast']: # test to patch a bug in mem_cache @@ -159,312 +162,315 @@ subprocess.call([ 'mkdir', '-p', os.path.join(scripts_path, graph_dir) ]) -############################################################ -### Graph 1 : Coherence traffic Cost per application ### -############################################################ - -for prot in single_protocols: - for app in apps: - data_coherence_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_coherence.dat') - gp_coherence_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_coherence.gp') - - # Creating the data file - width = 15 - content = "" - - for metric in [ '#nb_procs' ] + grouped_metrics: - content += metric + " " - nb_spaces = width - len(metric) - content += nb_spaces * ' ' - content += "\n" - - for i in nb_procs: - content += "%-15d " % i - for metric in grouped_metrics: - val = float(metrics_val[prot][app][i][metric]) / exec_time[prot][app][i] * 1000 - content += "%-15f " % val - content += "\n" - - create_file(data_coherence_name, content) - - # Creating the gp file - template_file = open(coherence_tmpl, 'r') - template = template_file.read() - - plot_str = "" - col = 2 - for metric in grouped_metrics: - if metric != grouped_metrics[0]: - plot_str += ", \\\n " - plot_str += "\"" + data_coherence_name + "\" using ($1):($" + str(col) + ") lc rgb " + colors[col - 2] + " title \"" + m_metric_name[metric] + "\" with linespoint" - col += 1 - gp_commands = template % dict(app_name = m_app_name[app], nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + app + '_coherence')) - - create_file(gp_coherence_name, gp_commands) - - # Calling gnuplot - print "gnuplot", gp_coherence_name - subprocess.call([ 'gnuplot', gp_coherence_name ]) - - -############################################################ -### Graph 2 : Speedup per Application ### -############################################################ - -for prot in single_protocols: - for app in apps: - - data_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_speedup.dat') - gp_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_speedup.gp') - - # Creating data file - width = 15 - content = "#nb_procs" - nb_spaces = width - len(content) - content += nb_spaces * ' ' - content += "speedup\n" - - for i in nb_procs: - content += "%-15d " % i - val = exec_time[prot][app][i] - content += "%-15f\n" % (exec_time[prot][app][1] / float(val)) - - plot_str = "\"" + data_speedup_name + "\" using ($1):($2) lc rgb \"#654387\" title \"Speedup\" with linespoint" - - create_file(data_speedup_name, content) - - # Creating the gp file - template_file = open(speedup_tmpl, 'r') - template = template_file.read() - - gp_commands = template % dict(appli = m_app_name[app], nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + app + '_speedup')) - - create_file(gp_speedup_name, gp_commands) - - # Calling gnuplot - print "gnuplot", gp_speedup_name - subprocess.call([ 'gnuplot', gp_speedup_name ]) - - -############################################################ -### Graph 3 : All speedups on the same Graph ### -############################################################ - -for prot in single_protocols: - # This graph uses the same template as the graph 2 - data_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_all_speedup.dat') - gp_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_all_speedup.gp') - - # Creating data file - width = 15 - content = "#nb_procs" - nb_spaces = width - len(content) - content += (nb_spaces + 1) * ' ' - for app in apps: - content += app + " " - content += (width - len(app)) * " " - content += "\n" - - for i in nb_procs: - content += "%-15d " % i - for app in apps: - val = exec_time[prot][app][i] - content += "%-15f " % (exec_time[prot][app][1] / float(val)) - content += "\n" - - create_file(data_speedup_name, content) - - # Creating gp file - template_file = open(speedup_tmpl, 'r') - template = template_file.read() - - plot_str = "" - col = 2 - for app in apps: - if app != apps[0]: - plot_str += ", \\\n " - plot_str += "\"" + data_speedup_name + "\" using ($1):($" + str(col) + ") lc rgb %s title \"" % (colors[col - 2]) + m_app_name[app] + "\" with linespoint" - col += 1 - - gp_commands = template % dict(appli = "All Applications", nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_all_speedup')) - - create_file(gp_speedup_name, gp_commands) - - # Calling gnuplot - print "gnuplot", gp_speedup_name - subprocess.call([ 'gnuplot', gp_speedup_name ]) - - -############################################################ -### Graph 4 : Graph per metric ### -############################################################ - -# The following section creates the graphs grouped by measure (e.g. #broadcasts) -# The template file cannot be easily created otherwise it would not be generic -# in many ways. This is why it is mainly created here. -# Graphs are created for metric in the "individual_metrics" list - -for prot in single_protocols: - for metric in individual_metrics: - data_metric_name = os.path.join(scripts_path, gen_dir, prot + '_' + metric + '.dat') - gp_metric_name = os.path.join(scripts_path, gen_dir, prot + '_' + metric + '.gp') - - # Creating the gp file - # Setting xtics, i.e. number of procs for each application - xtics_str = "(" - first = True - xpos = 1 - app_labels = "" - for num_appli in range(0, len(apps)): - for i in nb_procs: - if not first: - xtics_str += ", " - first = False - if i == nb_procs[0]: - xpos_first = xpos - xtics_str += "\"%d\" %.1f" % (i, xpos) - xpos_last = xpos - xpos += 1.5 - xpos += 0.5 - app_name_xpos = float((xpos_first + xpos_last)) / 2 - app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) - xtics_str += ")" - - xmax_val = float(xpos - 1) - - # Writing the lines of "plot" - plot_str = "" - xpos = 0 - first = True - column = 2 - for i in range(0, len(nb_procs)): - if not first: - plot_str += ", \\\n " - first = False - plot_str += "\"%s\" using ($1+%.1f):($%d) lc rgb %s notitle with boxes" % (data_metric_name, xpos, column, colors[i]) - column += 1 - xpos += 1.5 - - template_file = open(metric_tmpl, 'r') - template = template_file.read() - - gp_commands = template % dict(xtics_str = xtics_str, app_labels = app_labels, ylabel_str = m_metric_name[metric], norm_factor_str = m_norm_factor_name[m_metric_norm[metric]], xmax_val = xmax_val, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + metric)) - - create_file(gp_metric_name, gp_commands) - - # Creating the data file - width = 15 - content = "#x_pos" - nb_spaces = width - len(content) - content += nb_spaces * ' ' - for i in nb_procs: - content += "%-15d" % i - content += "\n" - - x_pos = 1 - for app in apps: - # Computation of x_pos - content += "%-15f" % x_pos - x_pos += len(nb_procs) * 1.5 + 0.5 - for i in nb_procs: - if m_metric_norm[metric] == "N": - content += "%-15d" % (metrics_val[prot][app][i][metric]) - elif m_metric_norm[metric] == "P": - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / i) - elif m_metric_norm[metric] == "C": - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / exec_time[prot][app][i] * 1000) - elif m_metric_norm[metric] == "W": - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_write'])) # Number of writes - elif m_metric_norm[metric] == "R": - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_read'])) # Number of reads - elif m_metric_norm[metric] == "D": - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_direct'])) # Number of req. - elif is_numeric(m_metric_norm[metric]): - content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][int(m_metric_norm[metric])][metric])) - else: - assert(False) - - app_name = m_app_name[app] - content += "#" + app_name + "\n" - - create_file(data_metric_name, content) - - # Calling gnuplot - print "gnuplot", gp_metric_name - subprocess.call([ 'gnuplot', gp_metric_name ]) - - -############################################################ -### Graph 5 : Stacked histogram with counters ### -############################################################ - -# The following section creates a stacked histogram containing -# the metrics in the "stacked_metric" list -# It is normalized per application w.r.t the values on 256 procs - -for prot in single_protocols: - data_stacked_name = os.path.join(scripts_path, gen_dir, prot + '_stacked.dat') - gp_stacked_name = os.path.join(scripts_path, gen_dir, prot + '_stacked.gp') - - norm_factor_value = nb_procs[-1] - - # Creating the gp file - template_file = open(stacked_tmpl, 'r') - template = template_file.read() - - xtics_str = "(" - first = True - xpos = 1 - app_labels = "" - for num_appli in range(0, len(apps)): - for i in nb_procs[1:len(nb_procs)]: # skipping values for 1 proc - if not first: - xtics_str += ", " - first = False - if i == nb_procs[1]: - xpos_first = xpos - xtics_str += "\"%d\" %d -1" % (i, xpos) - xpos_last = xpos - xpos += 1 - xpos += 1 - app_name_xpos = float((xpos_first + xpos_last)) / 2 - app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) - xtics_str += ")" - - plot_str = "newhistogram \"\"" - n = 1 - for stacked_metric in stacked_metrics: - plot_str += ", \\\n " + "'" + data_stacked_name + "'" + " using " + str(n) + " lc rgb " + colors[n] + " title \"" + m_metric_name[stacked_metric] + "\"" - n += 1 - - ylabel_str = "Breakdown of Coherence Traffic Normalized w.r.t. \\nthe Values on %d Processors" % norm_factor_value - content = template % dict(svg_name = os.path.join(graph_dir, prot + '_stacked'), xtics_str = xtics_str, plot_str = plot_str, ylabel_str = ylabel_str, app_labels = app_labels, prot_labels = "") - - create_file(gp_stacked_name, content) - - # Creating the data file - # Values are normalized by application, w.r.t. the number of requests for a given number of procs - content = "#" - for stacked_metric in stacked_metrics: - content += stacked_metric - content += ' ' + ' ' * (15 - len(stacked_metric)) - content += "\n" - for app in apps: - if app != apps[0]: - for i in range(0, len(stacked_metrics)): - content += "%-15f" % 0.0 - content += "\n" - for i in nb_procs[1:len(nb_procs)]: - for stacked_metric in stacked_metrics: - metric_val = metrics_val[prot][app][norm_factor_value]['total_stacked'] # Normalisation - if metric_val != 0: - content += "%-15f" % (float(metrics_val[prot][app][i][stacked_metric]) / metric_val) - else: - content += "%-15f" % 0 - content += "\n" - - create_file(data_stacked_name, content) - # Calling gnuplot - print "gnuplot", gp_stacked_name - subprocess.call([ 'gnuplot', gp_stacked_name ]) +############################################################# +#### Graph 1 : Coherence traffic Cost per application ### +############################################################# +# +#for prot in single_protocols: +# for app in apps: +# data_coherence_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_coherence.dat') +# gp_coherence_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_coherence.gp') +# +# # Creating the data file +# width = 15 +# content = "" +# +# for metric in [ '#nb_procs' ] + grouped_metrics: +# content += metric + " " +# nb_spaces = width - len(metric) +# content += nb_spaces * ' ' +# content += "\n" +# +# for i in nb_procs: +# content += "%-15d " % i +# for metric in grouped_metrics: +# #print "app : %s - prot : %s - i : %d" % (app, prot, i) +# val = float(metrics_val[prot][app][i][metric]) / exec_time[prot][app][i] * 1000 +# content += "%-15f " % val +# content += "\n" +# +# create_file(data_coherence_name, content) +# +# # Creating the gp file +# template_file = open(coherence_tmpl, 'r') +# template = template_file.read() +# +# plot_str = "" +# col = 2 +# for metric in grouped_metrics: +# if metric != grouped_metrics[0]: +# plot_str += ", \\\n " +# plot_str += "\"" + data_coherence_name + "\" using ($1):($" + str(col) + ") lc rgb " + colors[col - 2] + " title \"" + m_metric_name[metric] + "\" with linespoint" +# col += 1 +# gp_commands = template % dict(app_name = m_app_name[app], nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + app + '_coherence')) +# +# create_file(gp_coherence_name, gp_commands) +# +# # Calling gnuplot +# print "gnuplot", gp_coherence_name +# subprocess.call([ 'gnuplot', gp_coherence_name ]) +# +# +############################################################# +#### Graph 2 : Speedup per Application ### +############################################################# +# +#if 1 in nb_procs: +# for prot in single_protocols: +# for app in apps: +# +# data_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_speedup.dat') +# gp_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_' + app + '_speedup.gp') +# +# # Creating data file +# width = 15 +# content = "#nb_procs" +# nb_spaces = width - len(content) +# content += nb_spaces * ' ' +# content += "speedup\n" +# +# for i in nb_procs: +# content += "%-15d " % i +# val = exec_time[prot][app][i] +# content += "%-15f\n" % (exec_time[prot][app][1] / float(val)) +# +# plot_str = "\"" + data_speedup_name + "\" using ($1):($2) lc rgb \"#654387\" title \"Speedup\" with linespoint" +# +# create_file(data_speedup_name, content) +# +# # Creating the gp file +# template_file = open(speedup_tmpl, 'r') +# template = template_file.read() +# +# gp_commands = template % dict(appli = m_app_name[app], nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + app + '_speedup')) +# +# create_file(gp_speedup_name, gp_commands) +# +# # Calling gnuplot +# print "gnuplot", gp_speedup_name +# subprocess.call([ 'gnuplot', gp_speedup_name ]) +# +# +############################################################# +#### Graph 3 : All speedups on the same Graph ### +############################################################# +# +#if 1 in nb_procs: +# for prot in single_protocols: +# # This graph uses the same template as the graph 2 +# data_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_all_speedup.dat') +# gp_speedup_name = os.path.join(scripts_path, gen_dir, prot + '_all_speedup.gp') +# +# # Creating data file +# width = 15 +# content = "#nb_procs" +# nb_spaces = width - len(content) +# content += (nb_spaces + 1) * ' ' +# for app in apps: +# content += app + " " +# content += (width - len(app)) * " " +# content += "\n" +# +# for i in nb_procs: +# content += "%-15d " % i +# for app in apps: +# val = exec_time[prot][app][i] +# content += "%-15f " % (exec_time[prot][app][1] / float(val)) +# content += "\n" +# +# create_file(data_speedup_name, content) +# +# # Creating gp file +# template_file = open(speedup_tmpl, 'r') +# template = template_file.read() +# +# plot_str = "" +# col = 2 +# for app in apps: +# if app != apps[0]: +# plot_str += ", \\\n " +# plot_str += "\"" + data_speedup_name + "\" using ($1):($" + str(col) + ") lc rgb %s title \"" % (colors[col - 2]) + m_app_name[app] + "\" with linespoint" +# col += 1 +# +# gp_commands = template % dict(appli = "All Applications", nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_all_speedup')) +# +# create_file(gp_speedup_name, gp_commands) +# +# # Calling gnuplot +# print "gnuplot", gp_speedup_name +# subprocess.call([ 'gnuplot', gp_speedup_name ]) +# +# +############################################################# +#### Graph 4 : Graph per metric ### +############################################################# +# +## The following section creates the graphs grouped by measure (e.g. #broadcasts) +## The template file cannot be easily created otherwise it would not be generic +## in many ways. This is why it is mainly created here. +## Graphs are created for metric in the "individual_metrics" list +# +#for prot in single_protocols: +# for metric in individual_metrics: +# data_metric_name = os.path.join(scripts_path, gen_dir, prot + '_' + metric + '.dat') +# gp_metric_name = os.path.join(scripts_path, gen_dir, prot + '_' + metric + '.gp') +# +# # Creating the gp file +# # Setting xtics, i.e. number of procs for each application +# xtics_str = "(" +# first = True +# xpos = 1 +# app_labels = "" +# for num_appli in range(0, len(apps)): +# for i in nb_procs: +# if not first: +# xtics_str += ", " +# first = False +# if i == nb_procs[0]: +# xpos_first = xpos +# xtics_str += "\"%d\" %.1f" % (i, xpos) +# xpos_last = xpos +# xpos += 1.5 +# xpos += 0.5 +# app_name_xpos = float((xpos_first + xpos_last)) / 2 +# app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) +# xtics_str += ")" +# +# xmax_val = float(xpos - 1) +# +# # Writing the lines of "plot" +# plot_str = "" +# xpos = 0 +# first = True +# column = 2 +# for i in range(0, len(nb_procs)): +# if not first: +# plot_str += ", \\\n " +# first = False +# plot_str += "\"%s\" using ($1+%.1f):($%d) lc rgb %s notitle with boxes" % (data_metric_name, xpos, column, colors[i]) +# column += 1 +# xpos += 1.5 +# +# template_file = open(metric_tmpl, 'r') +# template = template_file.read() +# +# gp_commands = template % dict(xtics_str = xtics_str, app_labels = app_labels, ylabel_str = m_metric_name[metric], norm_factor_str = m_norm_factor_name[m_metric_norm[metric]], xmax_val = xmax_val, plot_str = plot_str, svg_name = os.path.join(graph_dir, prot + '_' + metric)) +# +# create_file(gp_metric_name, gp_commands) +# +# # Creating the data file +# width = 15 +# content = "#x_pos" +# nb_spaces = width - len(content) +# content += nb_spaces * ' ' +# for i in nb_procs: +# content += "%-15d" % i +# content += "\n" +# +# x_pos = 1 +# for app in apps: +# # Computation of x_pos +# content += "%-15f" % x_pos +# x_pos += len(nb_procs) * 1.5 + 0.5 +# for i in nb_procs: +# if m_metric_norm[metric] == "N": +# content += "%-15d" % (metrics_val[prot][app][i][metric]) +# elif m_metric_norm[metric] == "P": +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / i) +# elif m_metric_norm[metric] == "C": +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / exec_time[prot][app][i] * 1000) +# elif m_metric_norm[metric] == "W": +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_write'])) # Number of writes +# elif m_metric_norm[metric] == "R": +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_read'])) # Number of reads +# elif m_metric_norm[metric] == "D": +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][i]['total_direct'])) # Number of req. +# elif is_numeric(m_metric_norm[metric]): +# content += "%-15f" % (float(metrics_val[prot][app][i][metric]) / float(metrics_val[prot][app][int(m_metric_norm[metric])][metric])) +# else: +# assert(False) +# +# app_name = m_app_name[app] +# content += "#" + app_name + "\n" +# +# create_file(data_metric_name, content) +# +# # Calling gnuplot +# print "gnuplot", gp_metric_name +# subprocess.call([ 'gnuplot', gp_metric_name ]) +# +# +############################################################# +#### Graph 5 : Stacked histogram with counters ### +############################################################# +# +## The following section creates a stacked histogram containing +## the metrics in the "stacked_metric" list +## It is normalized per application w.r.t the values on 256 procs +# +#for prot in single_protocols: +# data_stacked_name = os.path.join(scripts_path, gen_dir, prot + '_stacked.dat') +# gp_stacked_name = os.path.join(scripts_path, gen_dir, prot + '_stacked.gp') +# +# norm_factor_value = nb_procs[-1] +# +# # Creating the gp file +# template_file = open(stacked_tmpl, 'r') +# template = template_file.read() +# +# xtics_str = "(" +# first = True +# xpos = 1 +# app_labels = "" +# for num_appli in range(0, len(apps)): +# for i in nb_procs[1:len(nb_procs)]: # skipping values for 1 proc +# if not first: +# xtics_str += ", " +# first = False +# if i == nb_procs[1]: +# xpos_first = xpos +# xtics_str += "\"%d\" %d -1" % (i, xpos) +# xpos_last = xpos +# xpos += 1 +# xpos += 1 +# app_name_xpos = float((xpos_first + xpos_last)) / 2 +# app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) +# xtics_str += ")" +# +# plot_str = "newhistogram \"\"" +# n = 1 +# for stacked_metric in stacked_metrics: +# plot_str += ", \\\n " + "'" + data_stacked_name + "'" + " using " + str(n) + " lc rgb " + colors[n] + " title \"" + m_metric_name[stacked_metric] + "\"" +# n += 1 +# +# ylabel_str = "Breakdown of Coherence Traffic Normalized w.r.t. \\nthe Values on %d Processors" % norm_factor_value +# content = template % dict(svg_name = os.path.join(graph_dir, prot + '_stacked'), xtics_str = xtics_str, plot_str = plot_str, ylabel_str = ylabel_str, app_labels = app_labels, prot_labels = "") +# +# create_file(gp_stacked_name, content) +# +# # Creating the data file +# # Values are normalized by application, w.r.t. the number of requests for a given number of procs +# content = "#" +# for stacked_metric in stacked_metrics: +# content += stacked_metric +# content += ' ' + ' ' * (15 - len(stacked_metric)) +# content += "\n" +# for app in apps: +# if app != apps[0]: +# for i in range(0, len(stacked_metrics)): +# content += "%-15f" % 0.0 +# content += "\n" +# for i in nb_procs[1:len(nb_procs)]: +# for stacked_metric in stacked_metrics: +# metric_val = metrics_val[prot][app][norm_factor_value]['total_stacked'] # Normalisation +# if metric_val != 0: +# content += "%-15f" % (float(metrics_val[prot][app][i][stacked_metric]) / metric_val) +# else: +# content += "%-15f" % 0 +# content += "\n" +# +# create_file(data_stacked_name, content) +# # Calling gnuplot +# print "gnuplot", gp_stacked_name +# subprocess.call([ 'gnuplot', gp_stacked_name ]) @@ -508,5 +514,5 @@ xpos += 3 app_name_xpos = float((xpos_first + xpos_last)) / 2 - app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) + app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,28\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) #xpos += 1 xtics_str += ")" @@ -518,5 +524,5 @@ n += 1 - ylabel_str = "Coherence Cost Compared to Direct Requests Cost,\\nNormalized per Application for each Number of Processors" + ylabel_str = "Coherence Cost Vs. Direct Requests Cost,\\nNormalized w.r.t. Direct Requests Cost" content = template % dict(svg_name = os.path.join(graph_dir, prot + '_rel_cost'), xtics_str = xtics_str, plot_str = plot_str, ylabel_str = ylabel_str, app_labels = app_labels, prot_labels = "") @@ -581,44 +587,45 @@ -for app in apps: - - data_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_speedup.dat') - gp_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_speedup.gp') - - # Creating data file - width = 15 - content = "#nb_procs" - nb_spaces = width - len(content) - content += nb_spaces * ' ' - content += "speedup\n" - - for i in nb_procs: - content += "%-15d " % i - for prot in joint_protocols: - val = exec_time[prot][app][i] - content += "%-15f " % (exec_time[joint_protocols[0]][app][1] / float(val)) - content += "\n" - - create_file(data_speedup_name, content) - - # Creating the gp file - template_file = open(speedup_tmpl, 'r') - template = template_file.read() - - plot_str = "" - col = 2 - for prot in joint_protocols: - if prot != joint_protocols[0]: - plot_str += ", \\\n " - plot_str += "\"" + data_speedup_name + "\" using ($1):($" + str(col) + ") lc rgb %s title \"" % (colors[col - 2]) + m_prot_name[prot] + "\" with linespoint" - col += 1 - - gp_commands = template % dict(appli = m_app_name[app] + " Normalized w.r.t. " + m_prot_name[joint_protocols[0]] + " on 1 Processor", nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, 'joint_' + app + '_speedup')) - - create_file(gp_speedup_name, gp_commands) - - # Calling gnuplot - print "gnuplot", gp_speedup_name - subprocess.call([ 'gnuplot', gp_speedup_name ]) +#if 1 in nb_procs: +# for app in apps: +# +# data_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_speedup.dat') +# gp_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_speedup.gp') +# +# # Creating data file +# width = 15 +# content = "#nb_procs" +# nb_spaces = width - len(content) +# content += nb_spaces * ' ' +# content += "speedup\n" +# +# for i in nb_procs: +# content += "%-15d " % i +# for prot in joint_protocols: +# val = exec_time[prot][app][i] +# content += "%-15f " % (exec_time[joint_protocols[0]][app][1] / float(val)) +# content += "\n" +# +# create_file(data_speedup_name, content) +# +# # Creating the gp file +# template_file = open(speedup_tmpl, 'r') +# template = template_file.read() +# +# plot_str = "" +# col = 2 +# for prot in joint_protocols: +# if prot != joint_protocols[0]: +# plot_str += ", \\\n " +# plot_str += "\"" + data_speedup_name + "\" using ($1):($" + str(col) + ") lc rgb %s title \"" % (colors[col - 2]) + m_prot_name[prot] + "\" with linespoint" +# col += 1 +# +# gp_commands = template % dict(appli = m_app_name[app] + " Normalized w.r.t. " + m_prot_name[joint_protocols[0]] + " on 1 Processor", nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, 'joint_' + app + '_speedup')) +# +# create_file(gp_speedup_name, gp_commands) +# +# # Calling gnuplot +# print "gnuplot", gp_speedup_name +# subprocess.call([ 'gnuplot', gp_speedup_name ]) @@ -638,8 +645,9 @@ template_file = open(stacked_tmpl, 'r') template = template_file.read() - + xtics_str = "(" first = True -xpos = 1 # successive x position of the center of the first bar in a application +barwidth = 1 +xpos = 1 # successive x position of the center of the first bar in an application app_labels = "" prot_labels = "" @@ -650,5 +658,5 @@ x = 0 # local var for computing position of protocol names for prot in joint_protocols: - prot_labels += "set label \"%s\" at first %f, character 2 center font \"Times,10\"\n" % (m_prot_name[prot], float((xpos - 0.5)) + x) # -0.5 instead of +0.5, don't know why... (bug gnuplot?) + prot_labels += "set label \"%s\" at first %f, character 2 center font \"Times,10\" rotate by 45\n" % (m_prot_name[prot], float((xpos - 0.5)) + x - 1) # -0.5 instead of +0.5, don't know why... (bug gnuplot?) x += 2 @@ -660,9 +668,10 @@ xpos_first = xpos xtics_str += "\"%d\" %f -1" % (i, float(xpos - 0.5 + len(joint_protocols))) + #xtics_str += "\"\" %f -1" % (float(xpos - 0.5 + len(joint_protocols))) # not to have proc number displayed xpos_last = xpos - xpos += 1 + len(joint_protocols) * 2 - app_name_xpos = float((xpos_first + xpos_last)) / 2 + xpos += 1 + len(joint_protocols) * 2 * barwidth + app_name_xpos = float((xpos_first + xpos_last)) / 2 + 2 app_labels += "set label \"%s\" at first %f,character 1 center font \"Times,12\"\n" % (m_app_name[apps[num_appli]], app_name_xpos) - xpos += 1 + #xpos += 1 xtics_str += ")" @@ -673,5 +682,5 @@ n += 1 -ylabel_str = "Coherence Cost and Direct Requests Cost,\\nNormalized per Application for each Number of Processors" +ylabel_str = "Coherence Cost vs. Direct Requests Cost,\\nNormalized per Application for each Number of Processors" content = template % dict(svg_name = os.path.join(graph_dir, 'joint_rel_cost'), xtics_str = xtics_str, plot_str = plot_str, ylabel_str = ylabel_str, app_labels = app_labels, prot_labels = prot_labels) @@ -800,4 +809,51 @@ - - +######################################################################################################### +### Graph 10: Normalized exec times with no log scale: when scalability is achieved by constant time ### +######################################################################################################### + + +if 1 in nb_procs: + for app in apps: + + data_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_cst_exec_time.dat') + gp_speedup_name = os.path.join(scripts_path, gen_dir, 'joint_' + app + '_cst_exec_time.gp') + + # Creating data file + width = 15 + content = "#nb_procs" + nb_spaces = width - len(content) + content += nb_spaces * ' ' + content += "norm_exec_time\n" + + for i in nb_procs: + content += "%-15d " % i + for prot in joint_protocols: + val = exec_time[prot][app][i] + content += "%-15f " % (float(val) / exec_time[joint_protocols[0]][app][1]) + content += "\n" + + create_file(data_speedup_name, content) + + # Creating the gp file + template_file = open(cst_exec_time_tmpl, 'r') + template = template_file.read() + + plot_str = "" + col = 2 + for prot in joint_protocols: + if prot != joint_protocols[0]: + plot_str += ", \\\n " + plot_str += "\"" + data_speedup_name + "\" using ($1):($" + str(col) + ") lc rgb %s title \"" % (colors[col - 2]) + m_prot_name[prot] + "\" with linespoint" + col += 1 + + gp_commands = template % dict(appli = m_app_name[app] + " Normalized w.r.t. " + m_prot_name[joint_protocols[0]] + " on 1 Processor", nb_procs = nb_procs[-1] + 1, plot_str = plot_str, svg_name = os.path.join(graph_dir, 'joint_' + app + '_cst_exec_time')) + + create_file(gp_speedup_name, gp_commands) + + # Calling gnuplot + print "gnuplot", gp_speedup_name + subprocess.call([ 'gnuplot', gp_speedup_name ]) + + + Index: trunk/platforms/tsar_generic_xbar/scripts/gen_arch_info.py =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/gen_arch_info.py (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/gen_arch_info.py (revision 1048) @@ -29,4 +29,5 @@ print " BSTTY=0x%x" % bstty print " BSDMA=0x%x" % bsdma + print " BSHCRYPT=0" # Not supported print " " print " " Index: trunk/platforms/tsar_generic_xbar/scripts/run_simus.py =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/run_simus.py (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/run_simus.py (revision 1048) @@ -17,8 +17,8 @@ bscpu = 0 # bootstrap CPU nb_procs = [ 4 ] -#nb_procs = [ 1, 4, 8, 16, 32, 64, 128, 256 ] +#nb_procs = [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ] rerun_stats = False -use_omp = False -protocol = 'rwt' +use_omp = True +protocol = 'dhccp' cpu_per_cluster = 4 # mode must be one of 'test' and 'simu' @@ -27,19 +27,24 @@ #apps = [ 'cholesky', 'fft', 'fft_ga', 'filter', 'filt_ga', 'histogram', 'kmeans', 'lu', 'mandel', 'mat_mult', 'pca', 'radix_ga' ] #apps = [ 'histogram', 'mandel', 'filter', 'radix_ga', 'fft_ga', 'kmeans' ] -#apps = [ 'blackscholes', 'linear_regression', 'string_match', 'swaptions', 'fluidanimate' ] -apps = [ 'hello', 'taquin', '2048' ] - - -# Variables which could be changed but ought not to because they are reflected in the create_graphs.py script +apps = [ 'fal_sh_2' ] + +#apps = [ 'pca-opt', 'blackscholes', 'fft_ga', 'filt_ga', 'histo-opt', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'radix_ga', 'string_match' ] +#apps = [ 'pca-opt', 'blackscholes', 'fft_ga', 'filt_ga', 'histo-opt', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'mat_mult-opt', 'radix_ga', 'string_match' ] + if mode == 'test': data_dir = 'data_test' else: data_dir = 'data' -log_init_name = protocol + '_stdo_' + +# Variables which could be changed but ought not to because they are reflected in the create_graphs.py script + +log_stdo_name = protocol + '_stdo_' log_term_name = protocol + '_term_' + # Global Variables -all_apps = [ '2048', 'blackscholes', 'boot_only', 'cholesky', 'fft', 'fft_ga', 'filter', 'filt_ga', 'fluidanimate', 'hello', 'histogram', 'histo-opt', 'kmeans', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'mat_mult', 'mat_mult-opt', 'pca', 'pca-opt', 'radix', 'radix_ga', 'showimg', 'string_match', 'swaptions', 'taquin'] +all_modes = ['simu', 'test'] +all_apps = [ 'blackscholes', 'boot_only', 'cholesky', 'fal_sh_1', 'fal_sh_2', 'fft', 'fft_ga', 'filter', 'filt_ga', 'fluidanimate', 'hello', 'histogram', 'histo-opt', 'kmeans', 'kmeans-opt', 'linear_regression', 'lu', 'mandel', 'mat_mult', 'mat_mult-opt', 'pca', 'pca-opt', 'radix', 'radix_ga', 'showimg', 'string_match', 'swaptions', ] # to come: 'barnes', 'fmm', 'ocean', 'raytrace', 'radiosity', 'waters', 'watern' @@ -60,5 +65,5 @@ hard_config_name = os.path.join(almos_path, "hard_config.h") bootloader_file_name = os.path.join(almos_path, "bootloader-tsar-mipsel.bin") -preloader_file_name = os.path.join(almos_path, "preloader.elf") +#preloader_file_name = os.path.join(almos_path, "preloader.elf") preloader_build_path = os.path.join(almos_path, "build_preloader") bootloader_build_path = os.path.join(almos_path, "build_bootloader") @@ -227,5 +232,4 @@ for common_module in common_modules: f.write("config.addDescPath(\"%s/%s\")\n" % (tsar_dir, common_module)) - #f.write("\n") if protocol == "dhccp": @@ -243,5 +247,4 @@ f.write("config.addDescPath(\"%s/%s\")\n" % (arch_dir, specific_module)) - #f.write("\n") f.close() @@ -330,5 +333,5 @@ print "cd", preloader_src_dir os.chdir(preloader_src_dir) - cmd = ['make', hard_conf_path_set, bscpu_set, 'USE_DT=0', 'BUILD_DIR=%s' % (preloader_build_path)] + cmd = ['make', hard_conf_path_set, bscpu_set, 'BLOCK_SIZE=4096', 'USE_32BIT=1', 'USE_DT=0', 'BUILD_DIR=%s' % (preloader_build_path)] retval = print_and_call(cmd) if retval != 0: @@ -406,4 +409,8 @@ elif (app_name == "boot_only"): shrc = "exec -p 0 /bin/boot_onl\n" + elif (app_name == "fal_sh_1"): + shrc = "exec -p 0 /bin/fal_sh_1 -n %(nproc)d -p %(niter)d\n" % dict(nproc = nprocs, niter = 100) + elif (app_name == "fal_sh_2"): + shrc = "exec -p 0 /bin/fal_sh_2 -n %(nproc)d -p %(niter)d\n" % dict(nproc = nprocs, niter = 100) elif (app_name == "filter"): shrc = "exec -p 0 /bin/filter -l 128 -c 128 -n %(nproc)d -i /etc/img128.raw\n" % dict(nproc = nprocs) @@ -460,4 +467,8 @@ elif (app_name == "cholesky"): shrc = "exec -p 0 /bin/cholesky -n%(nproc)d /etc/tk14.O\n" % dict(nproc = nprocs) + elif (app_name == "fal_sh_1"): + shrc = "exec -p 0 /bin/fal_sh_1 -n %(nproc)d -p %(niter)d\n" % dict(nproc = nprocs, niter = 10000) + elif (app_name == "fal_sh_2"): + shrc = "exec -p 0 /bin/fal_sh_2 -n %(nproc)d -p %(niter)d\n" % dict(nproc = nprocs, niter = 10000) elif (app_name == "fft"): shrc = "exec -p 0 /bin/fft -n%(nproc)d -m18\n" % dict(nproc = nprocs) @@ -533,5 +544,4 @@ print_and_call(cmd) -#gen_sym_links() gen_soclib_conf() compile_almos() @@ -580,5 +590,5 @@ else: cmd = ['./simul.x'] - filename = os.path.join(scripts_path, data_dir, app + '_' + log_init_name + str(i)) + filename = os.path.join(scripts_path, data_dir, app + '_' + log_stdo_name + str(i)) output = print_and_popen(cmd, filename) @@ -613,5 +623,5 @@ else: cmd = ['./simul.x', '--reset-counters', start2, '--dump-counters', end] - filename = os.path.join(scripts_path, data_dir, app + '_' + log_init_name + str(i)) + filename = os.path.join(scripts_path, data_dir, app + '_' + log_stdo_name + str(i)) output = print_and_popen(cmd, filename) Index: trunk/platforms/tsar_generic_xbar/scripts/templates/metric_template.gp =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/templates/metric_template.gp (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/templates/metric_template.gp (revision 1048) @@ -1,4 +1,4 @@ -set terminal svg size 1000 300 fixed +set terminal svg size 1200 300 fixed set output "%(svg_name)s.svg" Index: trunk/platforms/tsar_generic_xbar/scripts/templates/speedup_template.gp =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/templates/speedup_template.gp (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/templates/speedup_template.gp (revision 1048) @@ -3,6 +3,6 @@ set output "%(svg_name)s.svg" -set ylabel "Speedup for %(appli)s" font "Times,12" -set xlabel "Number of Cores" font "Times,12" +#set ylabel "Speedup for %(appli)s" font "Times,12" +#set xlabel "Number of Cores" font "Times,12" set xrange [1:%(nb_procs)d] @@ -15,4 +15,7 @@ set grid mytics +set ytics font "Times,16" +set xtics font "Times,16" + set logscale x 2 set logscale y 2 Index: trunk/platforms/tsar_generic_xbar/scripts/templates/stacked_template.gp =================================================================== --- trunk/platforms/tsar_generic_xbar/scripts/templates/stacked_template.gp (revision 1047) +++ trunk/platforms/tsar_generic_xbar/scripts/templates/stacked_template.gp (revision 1048) @@ -1,4 +1,4 @@ -set terminal svg size 1500 500 fixed +set terminal svg size 2500 500 fixed set output "%(svg_name)s.svg" @@ -24,5 +24,5 @@ # border : ne croise pas l'axe ; in : a l'interieur ; scale : taille (major,minor) ; offset : offset du texte -set xtics border in scale 0.0,0.0 nomirror rotate by 0 offset character 0,0 font "Times,10" +set xtics border in scale 0.0,0.0 nomirror rotate by 0 offset character 0,0 font "Times,19" set ytics border out scale 1,0.5 nomirror norotate offset character 0,0 @@ -30,5 +30,5 @@ set xlabel " " -set ylabel "%(ylabel_str)s" font "Times,14" +set ylabel "%(ylabel_str)s" font "Times,24" %(app_labels)s Index: trunk/platforms/tsar_generic_xbar/soclib.conf =================================================================== --- trunk/platforms/tsar_generic_xbar/soclib.conf (revision 1047) +++ (revision ) @@ -1,19 +1,0 @@ - -# DescPath modified by a script - - -config.default = config.mysystemcass - -config.addDescPath("/users/cao/meunier/src/tsar/lib/generic_llsc_global_table") -config.addDescPath("/users/cao/meunier/src/tsar/modules/dspin_router_tsar") -config.addDescPath("/users/cao/meunier/src/tsar/modules/sdmmc") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_block_device_tsar") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_ethernet_tsar") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_io_bridge") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_iox_network") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_spi") -config.addDescPath("/users/cao/meunier/src/tsar/platforms/tsar_generic_xbar/tsar_xbar_cluster") -config.addDescPath("/users/cao/meunier/src/tsar/communication") -config.addDescPath("/users/cao/meunier/src/tsar/lib/generic_cache_tsar") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_cc_vcache_wrapper") -config.addDescPath("/users/cao/meunier/src/tsar/modules/vci_mem_cache") Index: trunk/platforms/tsar_generic_xbar/top.cpp =================================================================== --- trunk/platforms/tsar_generic_xbar/top.cpp (revision 1047) +++ trunk/platforms/tsar_generic_xbar/top.cpp (revision 1048) @@ -95,17 +95,5 @@ #define USE_ALMOS 1 -//#define USE_GIET - -#ifdef USE_ALMOS -#ifdef USE_GIET -#error "Can't use Two different OS" -#endif -#endif - -#ifndef USE_ALMOS -#ifndef USE_GIET -#error "You need to specify one OS" -#endif -#endif + #ifdef USE_ALMOS @@ -113,7 +101,4 @@ #include "almos/hard_config.h" #endif -#ifdef USE_GIET - #define PREFIX_OS "giet_vm/" -#endif /////////////////////////////////////////////////// @@ -122,6 +107,6 @@ -#if USE_OPENMP -#include +#ifdef USE_OPENMP + #include #endif @@ -152,7 +137,4 @@ #ifdef USE_ALMOS #define vci_address_width 32 -#endif -#ifdef USE_GIET -#define vci_address_width 40 #endif #define vci_plen_width 8 @@ -185,13 +167,5 @@ #define FBUF_Y_SIZE 1024 #endif -#ifdef USE_GIET -#define FBUF_X_SIZE 128 -#define FBUF_Y_SIZE 128 -#endif - -#ifdef USE_GIET -#define BDEV_SECTOR_SIZE 512 -#define BDEV_IMAGE_NAME PREFIX_OS"display/images.raw" -#endif + #ifdef USE_ALMOS #define BDEV_SECTOR_SIZE 4096 @@ -214,7 +188,4 @@ #ifdef USE_ALMOS #define soft_name PREFIX_OS"preloader.elf" -#endif -#ifdef USE_GIET -#define soft_pathname PREFIX_OS"soft.elf" #endif @@ -244,121 +215,113 @@ #endif -#ifdef USE_ALMOS - // 2^19 is the offset for the local id (8 bits for global ID : - // 1 bit for Memcache or Peripheral, 4 for local peripheral id) - // (Almos supports 32 bits physical addresses) -#endif - bool stop_called = false; + +using namespace sc_core; +using namespace soclib::caba; +using namespace soclib::common; ///////////////////////////////// int _main(int argc, char *argv[]) { - using namespace sc_core; - using namespace soclib::caba; - using namespace soclib::common; - - const int64_t max_cycles = 5000000; // Maximum number of cycles simulated in one sc_start call - int64_t ncycles = 0x7FFFFFFFFFFFFFFF; // simulated cycles - char disk_name[256] = BDEV_IMAGE_NAME; // pathname to the disk image - char nic_rx_name[256] = NIC_RX_NAME; // pathname to the rx packets file - char nic_tx_name[256] = NIC_TX_NAME; // pathname to the tx packets file - ssize_t threads_nr = 1; // simulator's threads number - bool debug_ok = false; // trace activated - size_t debug_period = 1; // trace period - size_t debug_memc_id = 0; // index of memc to be traced - size_t debug_proc_id = 0; // index of proc to be traced - int64_t debug_from = 0; // trace start cycle - int64_t frozen_cycles = MAX_FROZEN_CYCLES; // monitoring frozen processor - int64_t reset_counters = -1; - int64_t dump_counters = -1; - bool do_reset_counters = false; - bool do_dump_counters = false; - struct timeval t1, t2; - uint64_t ms1, ms2; - - ////////////// command line arguments ////////////////////// - if (argc > 1) { - for (int n = 1; n < argc; n = n + 2) { - if ((strcmp(argv[n], "-NCYCLES") == 0) && (n + 1 < argc)) { - ncycles = (int64_t) strtol(argv[n + 1], NULL, 0); - } - else if ((strcmp(argv[n], "-SOFT") == 0) && (n + 1 < argc)) { + + const int64_t max_cycles = 5000000; // Maximum number of cycles simulated in one sc_start call + int64_t ncycles = 0x7FFFFFFFFFFFFFFF; // simulated cycles + char disk_name[256] = BDEV_IMAGE_NAME; // pathname to the disk image + char nic_rx_name[256] = NIC_RX_NAME; // pathname to the rx packets file + char nic_tx_name[256] = NIC_TX_NAME; // pathname to the tx packets file + ssize_t threads_nr = 1; // simulator's threads number + bool debug_ok = false; // trace activated + size_t debug_period = 1; // trace period + size_t debug_memc_id = 0; // index of memc to be traced + size_t debug_proc_id = 0; // index of proc to be traced + int64_t debug_from = 0; // trace start cycle + int64_t frozen_cycles = MAX_FROZEN_CYCLES; // monitoring frozen processor + int64_t reset_counters = -1; + int64_t dump_counters = -1; + bool do_reset_counters = false; + bool do_dump_counters = false; + struct timeval t1, t2; + uint64_t ms1, ms2; + + ////////////// command line arguments ////////////////////// + if (argc > 1) { + for (int n = 1; n < argc; n = n + 2) { + if ((strcmp(argv[n], "-NCYCLES") == 0) && (n + 1 < argc)) { + ncycles = (int64_t) strtol(argv[n + 1], NULL, 0); + } + else if ((strcmp(argv[n], "-SOFT") == 0) && (n + 1 < argc)) { #ifdef USE_ALMOS - assert( 0 && "Can't define almos soft name" ); -#endif -#ifdef USE_GIET - strcpy(soft_name, argv[n + 1]); -#endif - } - else if ((strcmp(argv[n],"-DISK") == 0) && (n + 1 < argc)) { - strcpy(disk_name, argv[n + 1]); - } - else if ((strcmp(argv[n],"-DEBUG") == 0) && (n + 1 < argc)) { - debug_ok = true; - debug_from = (int64_t) strtol(argv[n + 1], NULL, 0); - } - else if ((strcmp(argv[n], "-MEMCID") == 0) && (n + 1 < argc)) { - debug_memc_id = (size_t) strtol(argv[n + 1], NULL, 0); + assert( 0 && "Can't define almos soft name" ); +#endif + } + else if ((strcmp(argv[n],"-DISK") == 0) && (n + 1 < argc)) { + strcpy(disk_name, argv[n + 1]); + } + else if ((strcmp(argv[n],"-DEBUG") == 0) && (n + 1 < argc)) { + debug_ok = true; + debug_from = (int64_t) strtol(argv[n + 1], NULL, 0); + } + else if ((strcmp(argv[n], "-MEMCID") == 0) && (n + 1 < argc)) { + debug_memc_id = (size_t) strtol(argv[n + 1], NULL, 0); #ifdef USE_ALMOS - assert((debug_memc_id < (X_SIZE * Y_SIZE)) && - "debug_memc_id larger than X_SIZE * Y_SIZE" ); + assert((debug_memc_id < (X_SIZE * Y_SIZE)) && + "debug_memc_id larger than X_SIZE * Y_SIZE" ); #else - size_t x = debug_memc_id >> Y_WIDTH; - size_t y = debug_memc_id & ((1 << Y_WIDTH) - 1); - - assert( (x <= X_SIZE) and (y <= Y_SIZE) && - "MEMCID parameter refers a not valid memory cache"); -#endif - } - else if ((strcmp(argv[n], "-PROCID") == 0) && (n + 1 < argc)) { - debug_proc_id = (size_t) strtol(argv[n + 1], NULL, 0); + size_t x = debug_memc_id >> Y_WIDTH; + size_t y = debug_memc_id & ((1 << Y_WIDTH) - 1); + + assert( (x <= X_SIZE) and (y <= Y_SIZE) && + "MEMCID parameter refers a not valid memory cache"); +#endif + } + else if ((strcmp(argv[n], "-PROCID") == 0) && (n + 1 < argc)) { + debug_proc_id = (size_t) strtol(argv[n + 1], NULL, 0); #ifdef USE_ALMOS - assert((debug_proc_id < (X_SIZE * Y_SIZE * NB_PROCS_MAX)) && - "debug_proc_id larger than X_SIZE * Y_SIZE * NB_PROCS"); + assert((debug_proc_id < (X_SIZE * Y_SIZE * NB_PROCS_MAX)) && + "debug_proc_id larger than X_SIZE * Y_SIZE * NB_PROCS"); #else - size_t cluster_xy = debug_proc_id / NB_PROCS_MAX ; - size_t x = cluster_xy >> Y_WIDTH; - size_t y = cluster_xy & ((1 << Y_WIDTH) - 1); - - assert( (x <= X_SIZE) and (y <= Y_SIZE) && - "PROCID parameter refers a not valid processor"); -#endif - } - else if ((strcmp(argv[n], "-THREADS") == 0) && ((n + 1) < argc)) { - threads_nr = (ssize_t) strtol(argv[n + 1], NULL, 0); - threads_nr = (threads_nr < 1) ? 1 : threads_nr; - } - else if ((strcmp(argv[n], "-FROZEN") == 0) && (n + 1 < argc)) { - frozen_cycles = (int64_t) strtol(argv[n + 1], NULL, 0); - } - else if ((strcmp(argv[n], "-PERIOD") == 0) && (n + 1 < argc)) { - debug_period = (size_t) strtol(argv[n + 1], NULL, 0); - } - else if ((strcmp(argv[n], "--reset-counters") == 0) && (n + 1 < argc)) { - reset_counters = (int64_t) strtol(argv[n + 1], NULL, 0); - do_reset_counters = true; - } - else if ((strcmp(argv[n], "--dump-counters") == 0) && (n + 1 < argc)) { - dump_counters = (int64_t) strtol(argv[n + 1], NULL, 0); - do_dump_counters = true; - } - else { - std::cout << " Arguments are (key,value) couples." << std::endl; - std::cout << " The order is not important." << std::endl; - std::cout << " Accepted arguments are :" << std::endl << std::endl; - std::cout << " -SOFT pathname_for_embedded_soft" << std::endl; - std::cout << " -DISK pathname_for_disk_image" << std::endl; - std::cout << " -NCYCLES number_of_simulated_cycles" << std::endl; - std::cout << " -DEBUG debug_start_cycle" << std::endl; - std::cout << " -THREADS simulator's threads number" << std::endl; - std::cout << " -FROZEN max_number_of_lines" << std::endl; - std::cout << " -PERIOD number_of_cycles between trace" << std::endl; - std::cout << " -MEMCID index_memc_to_be_traced" << std::endl; - std::cout << " -PROCID index_proc_to_be_traced" << std::endl; - exit(0); - } - } - } + size_t cluster_xy = debug_proc_id / NB_PROCS_MAX ; + size_t x = cluster_xy >> Y_WIDTH; + size_t y = cluster_xy & ((1 << Y_WIDTH) - 1); + + assert( (x <= X_SIZE) and (y <= Y_SIZE) && + "PROCID parameter refers a not valid processor"); +#endif + } + else if ((strcmp(argv[n], "-THREADS") == 0) && ((n + 1) < argc)) { + threads_nr = (ssize_t) strtol(argv[n + 1], NULL, 0); + threads_nr = (threads_nr < 1) ? 1 : threads_nr; + } + else if ((strcmp(argv[n], "-FROZEN") == 0) && (n + 1 < argc)) { + frozen_cycles = (int64_t) strtol(argv[n + 1], NULL, 0); + } + else if ((strcmp(argv[n], "-PERIOD") == 0) && (n + 1 < argc)) { + debug_period = (size_t) strtol(argv[n + 1], NULL, 0); + } + else if ((strcmp(argv[n], "--reset-counters") == 0) && (n + 1 < argc)) { + reset_counters = (int64_t) strtol(argv[n + 1], NULL, 0); + do_reset_counters = true; + } + else if ((strcmp(argv[n], "--dump-counters") == 0) && (n + 1 < argc)) { + dump_counters = (int64_t) strtol(argv[n + 1], NULL, 0); + do_dump_counters = true; + } + else { + std::cout << " Arguments are (key,value) couples." << std::endl; + std::cout << " The order is not important." << std::endl; + std::cout << " Accepted arguments are :" << std::endl << std::endl; + std::cout << " -SOFT pathname_for_embedded_soft" << std::endl; + std::cout << " -DISK pathname_for_disk_image" << std::endl; + std::cout << " -NCYCLES number_of_simulated_cycles" << std::endl; + std::cout << " -DEBUG debug_start_cycle" << std::endl; + std::cout << " -THREADS simulator's threads number" << std::endl; + std::cout << " -FROZEN max_number_of_lines" << std::endl; + std::cout << " -PERIOD number_of_cycles between trace" << std::endl; + std::cout << " -MEMCID index_memc_to_be_traced" << std::endl; + std::cout << " -PROCID index_proc_to_be_traced" << std::endl; + exit(0); + } + } + } // checking hardware parameters @@ -383,9 +346,4 @@ assert( (NB_NIC_CHANNELS < 9) and "The NB_NIC_CHANNELS parameter must be smaller than 9" ); - -#ifdef USE_GIET - assert( (vci_address_width == 40) and - "VCI address width with the GIET must be 40 bits" ); -#endif #ifdef USE_ALMOS @@ -431,229 +389,232 @@ vci_wrplen_width> vci_param_ext; -#if USE_OPENMP - omp_set_dynamic(false); - omp_set_num_threads(threads_nr); - std::cerr << "Built with openmp version " << _OPENMP << std::endl; -#endif - - // Define parameters depending on mesh size - size_t x_width; - size_t y_width; +#ifdef USE_OPENMP + omp_set_dynamic(false); + omp_set_num_threads(threads_nr); + std::cerr << "Built with openmp version " << _OPENMP << std::endl; + std::cerr << "Run with " << threads_nr << " threads" << std::endl; +#endif + + // Define parameters depending on mesh size + size_t x_width; + size_t y_width; #ifdef USE_ALMOS - if (X_SIZE == 1) x_width = 0; - else if (X_SIZE == 2) x_width = 1; - else if (X_SIZE <= 4) x_width = 2; - else if (X_SIZE <= 8) x_width = 3; - else x_width = 4; - - if (Y_SIZE == 1) y_width = 0; - else if (Y_SIZE == 2) y_width = 1; - else if (Y_SIZE <= 4) y_width = 2; - else if (Y_SIZE <= 8) y_width = 3; - else y_width = 4; + if (X_SIZE == 1) x_width = 0; + else if (X_SIZE == 2) x_width = 1; + else if (X_SIZE <= 4) x_width = 2; + else if (X_SIZE <= 8) x_width = 3; + else x_width = 4; + + if (Y_SIZE == 1) y_width = 0; + else if (Y_SIZE == 2) y_width = 1; + else if (Y_SIZE <= 4) y_width = 2; + else if (Y_SIZE <= 8) y_width = 3; + else y_width = 4; #else - size_t x_width = X_WIDTH; - size_t y_width = Y_WIDTH; - - assert((X_WIDTH <= 4) and (Y_WIDTH <= 4) and - "Up to 256 clusters"); - - assert((X_SIZE <= (1 << X_WIDTH)) and (Y_SIZE <= (1 << Y_WIDTH)) and - "The X_WIDTH and Y_WIDTH parameter are insufficient"); - -#endif - - ///////////////////// - // Mapping Tables - ///////////////////// - - // internal network - MappingTable maptabd(vci_address_width, - IntTab(x_width + y_width, 16 - x_width - y_width), - IntTab(x_width + y_width, vci_srcid_width - x_width - y_width), - 0x00FF800000); - - for (size_t x = 0; x < X_SIZE; x++) { - for (size_t y = 0; y < Y_SIZE; y++) { - sc_uint offset; - offset = (sc_uint) cluster(x,y) - << (vci_address_width - x_width - y_width); - - std::ostringstream si; - si << "seg_xicu_" << x << "_" << y; - maptabd.add(Segment(si.str(), SEG_XCU_BASE + offset, SEG_XCU_SIZE, - IntTab(cluster(x,y), XCU_TGTID), false)); - - std::ostringstream sd; - sd << "seg_mdma_" << x << "_" << y; - maptabd.add(Segment(sd.str(), SEG_DMA_BASE + offset, SEG_DMA_SIZE, - IntTab(cluster(x,y), DMA_TGTID), false)); - - std::ostringstream sh; - sh << "seg_memc_" << x << "_" << y; - maptabd.add(Segment(sh.str(), SEG_RAM_BASE + offset, SEG_RAM_SIZE, - IntTab(cluster(x,y), RAM_TGTID), true)); - - if (x == X_IO && y == Y_IO) { - maptabd.add(Segment("seg_mtty", SEG_TTY_BASE, SEG_TTY_SIZE, - IntTab(cluster(x,y),TTY_TGTID), false)); - maptabd.add(Segment("seg_fbuf", SEG_FBF_BASE, SEG_FBF_SIZE, - IntTab(cluster(x,y),FBF_TGTID), false)); - maptabd.add(Segment("seg_bdev", SEG_IOC_BASE, SEG_IOC_SIZE, - IntTab(cluster(x,y),IOC_TGTID), false)); - maptabd.add(Segment("seg_brom", SEG_ROM_BASE, SEG_ROM_SIZE, - IntTab(cluster(x,y),ROM_TGTID), true)); - maptabd.add(Segment("seg_mnic", SEG_NIC_BASE, SEG_NIC_SIZE, - IntTab(cluster(x,y),NIC_TGTID), false)); - maptabd.add(Segment("seg_cdma", SEG_CMA_BASE, SEG_CMA_SIZE, - IntTab(cluster(x,y),CMA_TGTID), false)); - maptabd.add(Segment("seg_simh", SEG_SIM_BASE, SEG_SIM_SIZE, - IntTab(cluster(x,y),SIM_TGTID), false)); - } - } - } - std::cout << maptabd << std::endl; - - // external network - MappingTable maptabx(vci_address_width, - IntTab(x_width + y_width), - IntTab(x_width + y_width), - 0xFFFF000000ULL); - - for (size_t x = 0; x < X_SIZE; x++) { - for (size_t y = 0; y < Y_SIZE ; y++) { - - sc_uint offset; - offset = (sc_uint) cluster(x,y) - << (vci_address_width - x_width - y_width); - - std::ostringstream sh; - sh << "x_seg_memc_" << x << "_" << y; - - maptabx.add(Segment(sh.str(), SEG_RAM_BASE + offset, - SEG_RAM_SIZE, IntTab(cluster(x,y)), false)); - } - } - std::cout << maptabx << std::endl; - - //////////////////// - // Signals - /////////////////// - - sc_clock signal_clk("clk"); - sc_signal signal_resetn("resetn"); - - // Horizontal inter-clusters DSPIN signals - DspinSignals** signal_dspin_h_cmd_inc = - alloc_elems >("signal_dspin_h_cmd_inc", X_SIZE - 1, Y_SIZE); - DspinSignals** signal_dspin_h_cmd_dec = - alloc_elems >("signal_dspin_h_cmd_dec", X_SIZE - 1, Y_SIZE); - - DspinSignals** signal_dspin_h_rsp_inc = - alloc_elems >("signal_dspin_h_rsp_inc", X_SIZE - 1, Y_SIZE); - DspinSignals** signal_dspin_h_rsp_dec = - alloc_elems >("signal_dspin_h_rsp_dec", X_SIZE - 1, Y_SIZE); - - DspinSignals** signal_dspin_h_m2p_inc = - alloc_elems >("signal_dspin_h_m2p_inc", X_SIZE- 1 , Y_SIZE); - DspinSignals** signal_dspin_h_m2p_dec = - alloc_elems >("signal_dspin_h_m2p_dec", X_SIZE - 1, Y_SIZE); - - DspinSignals** signal_dspin_h_p2m_inc = - alloc_elems >("signal_dspin_h_p2m_inc", X_SIZE - 1, Y_SIZE); - DspinSignals** signal_dspin_h_p2m_dec = - alloc_elems >("signal_dspin_h_p2m_dec", X_SIZE - 1, Y_SIZE); - - DspinSignals** signal_dspin_h_cla_inc = - alloc_elems >("signal_dspin_h_cla_inc", X_SIZE - 1, Y_SIZE); - DspinSignals** signal_dspin_h_cla_dec = - alloc_elems >("signal_dspin_h_cla_dec", X_SIZE - 1, Y_SIZE); - - // Vertical inter-clusters DSPIN signals - DspinSignals** signal_dspin_v_cmd_inc = - alloc_elems >("signal_dspin_v_cmd_inc", X_SIZE, Y_SIZE - 1); - DspinSignals** signal_dspin_v_cmd_dec = - alloc_elems >("signal_dspin_v_cmd_dec", X_SIZE, Y_SIZE - 1); - - DspinSignals** signal_dspin_v_rsp_inc = - alloc_elems >("signal_dspin_v_rsp_inc", X_SIZE, Y_SIZE - 1); - DspinSignals** signal_dspin_v_rsp_dec = - alloc_elems >("signal_dspin_v_rsp_dec", X_SIZE, Y_SIZE - 1); - - DspinSignals** signal_dspin_v_m2p_inc = - alloc_elems >("signal_dspin_v_m2p_inc", X_SIZE, Y_SIZE - 1); - DspinSignals** signal_dspin_v_m2p_dec = - alloc_elems >("signal_dspin_v_m2p_dec", X_SIZE, Y_SIZE - 1); - - DspinSignals** signal_dspin_v_p2m_inc = - alloc_elems >("signal_dspin_v_p2m_inc", X_SIZE, Y_SIZE - 1); - DspinSignals** signal_dspin_v_p2m_dec = - alloc_elems >("signal_dspin_v_p2m_dec", X_SIZE, Y_SIZE - 1); - - DspinSignals** signal_dspin_v_cla_inc = - alloc_elems >("signal_dspin_v_cla_inc", X_SIZE, Y_SIZE - 1); - DspinSignals** signal_dspin_v_cla_dec = - alloc_elems >("signal_dspin_v_cla_dec", X_SIZE, Y_SIZE - 1); - - // Mesh boundaries DSPIN signals (Most of those signals are not used...) - DspinSignals*** signal_dspin_bound_cmd_in = - alloc_elems >("signal_dspin_bound_cmd_in" , X_SIZE, Y_SIZE, 4); - DspinSignals*** signal_dspin_bound_cmd_out = - alloc_elems >("signal_dspin_bound_cmd_out", X_SIZE, Y_SIZE, 4); - - DspinSignals*** signal_dspin_bound_rsp_in = - alloc_elems >("signal_dspin_bound_rsp_in" , X_SIZE, Y_SIZE, 4); - DspinSignals*** signal_dspin_bound_rsp_out = - alloc_elems >("signal_dspin_bound_rsp_out", X_SIZE, Y_SIZE, 4); - - DspinSignals*** signal_dspin_bound_m2p_in = - alloc_elems >("signal_dspin_bound_m2p_in" , X_SIZE, Y_SIZE, 4); - DspinSignals*** signal_dspin_bound_m2p_out = - alloc_elems >("signal_dspin_bound_m2p_out", X_SIZE, Y_SIZE, 4); - - DspinSignals*** signal_dspin_bound_p2m_in = - alloc_elems >("signal_dspin_bound_p2m_in" , X_SIZE, Y_SIZE, 4); - DspinSignals*** signal_dspin_bound_p2m_out = - alloc_elems >("signal_dspin_bound_p2m_out", X_SIZE, Y_SIZE, 4); - - DspinSignals*** signal_dspin_bound_cla_in = - alloc_elems >("signal_dspin_bound_cla_in" , X_SIZE, Y_SIZE, 4); - DspinSignals*** signal_dspin_bound_cla_out = - alloc_elems >("signal_dspin_bound_cla_out", X_SIZE, Y_SIZE, 4); - - - //////////////////////////// - // Loader - //////////////////////////// - - soclib::common::Loader loader(soft_name); - - typedef soclib::common::GdbServer proc_iss; - proc_iss::set_loader(loader); - - //////////////////////////// - // Clusters construction - //////////////////////////// - - TsarXbarCluster * clusters[X_SIZE][Y_SIZE]; - -#if USE_OPENMP + size_t x_width = X_WIDTH; + size_t y_width = Y_WIDTH; + + assert((X_WIDTH <= 4) and (Y_WIDTH <= 4) and + "Up to 256 clusters"); + + assert((X_SIZE <= (1 << X_WIDTH)) and (Y_SIZE <= (1 << Y_WIDTH)) and + "The X_WIDTH and Y_WIDTH parameter are insufficient"); + +#endif + + ///////////////////// + // Mapping Tables + ///////////////////// + + // internal network + MappingTable maptabd(vci_address_width, + IntTab(x_width + y_width, 16 - x_width - y_width), + IntTab(x_width + y_width, vci_srcid_width - x_width - y_width), + 0x00FF800000); + + for (size_t x = 0; x < X_SIZE; x++) { + for (size_t y = 0; y < Y_SIZE; y++) { + sc_uint offset; + offset = (sc_uint) cluster(x,y) + << (vci_address_width - x_width - y_width); + + std::ostringstream si; + si << "seg_xicu_" << x << "_" << y; + maptabd.add(Segment(si.str(), SEG_XCU_BASE + offset, SEG_XCU_SIZE, + IntTab(cluster(x,y), XCU_TGTID), false)); + + std::ostringstream sd; + sd << "seg_mdma_" << x << "_" << y; + maptabd.add(Segment(sd.str(), SEG_DMA_BASE + offset, SEG_DMA_SIZE, + IntTab(cluster(x,y), DMA_TGTID), false)); + + std::ostringstream sh; + sh << "seg_memc_" << x << "_" << y; + maptabd.add(Segment(sh.str(), SEG_RAM_BASE + offset, SEG_RAM_SIZE, + IntTab(cluster(x,y), RAM_TGTID), true)); + + if (x == X_IO && y == Y_IO) { + maptabd.add(Segment("seg_mtty", SEG_TTY_BASE, SEG_TTY_SIZE, + IntTab(cluster(x,y),TTY_TGTID), false)); + maptabd.add(Segment("seg_fbuf", SEG_FBF_BASE, SEG_FBF_SIZE, + IntTab(cluster(x,y),FBF_TGTID), false)); + maptabd.add(Segment("seg_bdev", SEG_IOC_BASE, SEG_IOC_SIZE, + IntTab(cluster(x,y),IOC_TGTID), false)); + maptabd.add(Segment("seg_brom", SEG_ROM_BASE, SEG_ROM_SIZE, + IntTab(cluster(x,y),ROM_TGTID), true)); + maptabd.add(Segment("seg_mnic", SEG_NIC_BASE, SEG_NIC_SIZE, + IntTab(cluster(x,y),NIC_TGTID), false)); + maptabd.add(Segment("seg_cdma", SEG_CMA_BASE, SEG_CMA_SIZE, + IntTab(cluster(x,y),CMA_TGTID), false)); + maptabd.add(Segment("seg_simh", SEG_SIM_BASE, SEG_SIM_SIZE, + IntTab(cluster(x,y),SIM_TGTID), false)); + } + } + } + std::cout << maptabd << std::endl; + + // external network + MappingTable maptabx(vci_address_width, + IntTab(x_width + y_width), + IntTab(x_width + y_width), + 0xFFFF000000ULL); + + for (size_t x = 0; x < X_SIZE; x++) { + for (size_t y = 0; y < Y_SIZE ; y++) { + + sc_uint offset; + offset = (sc_uint) cluster(x,y) + << (vci_address_width - x_width - y_width); + + std::ostringstream sh; + sh << "x_seg_memc_" << x << "_" << y; + + maptabx.add(Segment(sh.str(), SEG_RAM_BASE + offset, + SEG_RAM_SIZE, IntTab(cluster(x,y)), false)); + } + } + std::cout << maptabx << std::endl; + + //////////////////// + // Signals + /////////////////// + + sc_clock signal_clk("clk"); + sc_signal signal_resetn("resetn"); + + // Horizontal inter-clusters DSPIN signals + DspinSignals** signal_dspin_h_cmd_inc = + alloc_elems >("signal_dspin_h_cmd_inc", X_SIZE - 1, Y_SIZE); + DspinSignals** signal_dspin_h_cmd_dec = + alloc_elems >("signal_dspin_h_cmd_dec", X_SIZE - 1, Y_SIZE); + + DspinSignals** signal_dspin_h_rsp_inc = + alloc_elems >("signal_dspin_h_rsp_inc", X_SIZE - 1, Y_SIZE); + DspinSignals** signal_dspin_h_rsp_dec = + alloc_elems >("signal_dspin_h_rsp_dec", X_SIZE - 1, Y_SIZE); + + DspinSignals** signal_dspin_h_m2p_inc = + alloc_elems >("signal_dspin_h_m2p_inc", X_SIZE- 1 , Y_SIZE); + DspinSignals** signal_dspin_h_m2p_dec = + alloc_elems >("signal_dspin_h_m2p_dec", X_SIZE - 1, Y_SIZE); + + DspinSignals** signal_dspin_h_p2m_inc = + alloc_elems >("signal_dspin_h_p2m_inc", X_SIZE - 1, Y_SIZE); + DspinSignals** signal_dspin_h_p2m_dec = + alloc_elems >("signal_dspin_h_p2m_dec", X_SIZE - 1, Y_SIZE); + + DspinSignals** signal_dspin_h_cla_inc = + alloc_elems >("signal_dspin_h_cla_inc", X_SIZE - 1, Y_SIZE); + DspinSignals** signal_dspin_h_cla_dec = + alloc_elems >("signal_dspin_h_cla_dec", X_SIZE - 1, Y_SIZE); + + // Vertical inter-clusters DSPIN signals + DspinSignals** signal_dspin_v_cmd_inc = + alloc_elems >("signal_dspin_v_cmd_inc", X_SIZE, Y_SIZE - 1); + DspinSignals** signal_dspin_v_cmd_dec = + alloc_elems >("signal_dspin_v_cmd_dec", X_SIZE, Y_SIZE - 1); + + DspinSignals** signal_dspin_v_rsp_inc = + alloc_elems >("signal_dspin_v_rsp_inc", X_SIZE, Y_SIZE - 1); + DspinSignals** signal_dspin_v_rsp_dec = + alloc_elems >("signal_dspin_v_rsp_dec", X_SIZE, Y_SIZE - 1); + + DspinSignals** signal_dspin_v_m2p_inc = + alloc_elems >("signal_dspin_v_m2p_inc", X_SIZE, Y_SIZE - 1); + DspinSignals** signal_dspin_v_m2p_dec = + alloc_elems >("signal_dspin_v_m2p_dec", X_SIZE, Y_SIZE - 1); + + DspinSignals** signal_dspin_v_p2m_inc = + alloc_elems >("signal_dspin_v_p2m_inc", X_SIZE, Y_SIZE - 1); + DspinSignals** signal_dspin_v_p2m_dec = + alloc_elems >("signal_dspin_v_p2m_dec", X_SIZE, Y_SIZE - 1); + + DspinSignals** signal_dspin_v_cla_inc = + alloc_elems >("signal_dspin_v_cla_inc", X_SIZE, Y_SIZE - 1); + DspinSignals** signal_dspin_v_cla_dec = + alloc_elems >("signal_dspin_v_cla_dec", X_SIZE, Y_SIZE - 1); + + // Mesh boundaries DSPIN signals (Most of those signals are not used...) + DspinSignals*** signal_dspin_bound_cmd_in = + alloc_elems >("signal_dspin_bound_cmd_in" , X_SIZE, Y_SIZE, 4); + DspinSignals*** signal_dspin_bound_cmd_out = + alloc_elems >("signal_dspin_bound_cmd_out", X_SIZE, Y_SIZE, 4); + + DspinSignals*** signal_dspin_bound_rsp_in = + alloc_elems >("signal_dspin_bound_rsp_in" , X_SIZE, Y_SIZE, 4); + DspinSignals*** signal_dspin_bound_rsp_out = + alloc_elems >("signal_dspin_bound_rsp_out", X_SIZE, Y_SIZE, 4); + + DspinSignals*** signal_dspin_bound_m2p_in = + alloc_elems >("signal_dspin_bound_m2p_in" , X_SIZE, Y_SIZE, 4); + DspinSignals*** signal_dspin_bound_m2p_out = + alloc_elems >("signal_dspin_bound_m2p_out", X_SIZE, Y_SIZE, 4); + + DspinSignals*** signal_dspin_bound_p2m_in = + alloc_elems >("signal_dspin_bound_p2m_in" , X_SIZE, Y_SIZE, 4); + DspinSignals*** signal_dspin_bound_p2m_out = + alloc_elems >("signal_dspin_bound_p2m_out", X_SIZE, Y_SIZE, 4); + + DspinSignals*** signal_dspin_bound_cla_in = + alloc_elems >("signal_dspin_bound_cla_in" , X_SIZE, Y_SIZE, 4); + DspinSignals*** signal_dspin_bound_cla_out = + alloc_elems >("signal_dspin_bound_cla_out", X_SIZE, Y_SIZE, 4); + + + //////////////////////////// + // Loader + //////////////////////////// + + soclib::common::Loader loader(soft_name); + + typedef soclib::common::GdbServer proc_iss; + proc_iss::set_loader(loader); + + //////////////////////////// + // Clusters construction + //////////////////////////// + + TsarXbarCluster * clusters[X_SIZE][Y_SIZE]; + +#ifdef USE_OPENMP #pragma omp parallel +#endif { +#ifdef USE_OPENMP #pragma omp for #endif - for (size_t i = 0; i < (X_SIZE * Y_SIZE); i++) { - size_t x = i / Y_SIZE; - size_t y = i % Y_SIZE; - -#if USE_OPENMP + for (size_t i = 0; i < (X_SIZE * Y_SIZE); i++) { + size_t x = i / Y_SIZE; + size_t y = i % Y_SIZE; + +#ifdef USE_OPENMP #pragma omp critical - { -#endif +#endif + { std::cout << std::endl; std::cout << "Cluster_" << x << "_" << y << std::endl; @@ -663,198 +624,194 @@ sc << "cluster_" << x << "_" << y; clusters[x][y] = new TsarXbarCluster - ( - sc.str().c_str(), - NB_PROCS_MAX, - NB_TTY_CHANNELS, - NB_DMA_CHANNELS, - x, - y, - cluster(x,y), - maptabd, - maptabx, - x_width, - y_width, - vci_srcid_width - x_width - y_width, // l_id width, - P_WIDTH, - RAM_TGTID, - XCU_TGTID, - DMA_TGTID, - FBF_TGTID, - TTY_TGTID, - ROM_TGTID, - NIC_TGTID, - CMA_TGTID, - IOC_TGTID, - SIM_TGTID, - MEMC_WAYS, - MEMC_SETS, - L1_IWAYS, - L1_ISETS, - L1_DWAYS, - L1_DSETS, - IRQ_PER_PROCESSOR, - XRAM_LATENCY, - x == X_IO && y == Y_IO, - FBF_X_SIZE, - FBF_Y_SIZE, - disk_name, - BDEV_SECTOR_SIZE, - NB_NIC_CHANNELS, - nic_rx_name, - nic_tx_name, - NIC_TIMEOUT, - NB_CMA_CHANNELS, - loader, - frozen_cycles, - debug_from, - debug_ok, - debug_ok - ); - -#if USE_OPENMP - } // end critical -#endif - } // end for -#if USE_OPENMP - } -#endif - - /////////////////////////////////////////////////////////////// - // Net-list - /////////////////////////////////////////////////////////////// - - // Clock & RESET - for (int x = 0; x < X_SIZE; x++) { - for (int y = 0; y < Y_SIZE; y++) { - clusters[x][y]->p_clk (signal_clk); - clusters[x][y]->p_resetn (signal_resetn); - } - } - - // Inter Clusters horizontal connections - for (int x = 0; x < X_SIZE - 1; x++) { - for (int y = 0; y < Y_SIZE; y++) { - clusters[x][y]->p_cmd_out[EAST] (signal_dspin_h_cmd_inc[x][y]); - clusters[x + 1][y]->p_cmd_in[WEST] (signal_dspin_h_cmd_inc[x][y]); - clusters[x][y]->p_cmd_in[EAST] (signal_dspin_h_cmd_dec[x][y]); - clusters[x + 1][y]->p_cmd_out[WEST] (signal_dspin_h_cmd_dec[x][y]); - - clusters[x][y]->p_rsp_out[EAST] (signal_dspin_h_rsp_inc[x][y]); - clusters[x + 1][y]->p_rsp_in[WEST] (signal_dspin_h_rsp_inc[x][y]); - clusters[x][y]->p_rsp_in[EAST] (signal_dspin_h_rsp_dec[x][y]); - clusters[x + 1][y]->p_rsp_out[WEST] (signal_dspin_h_rsp_dec[x][y]); - - clusters[x][y]->p_m2p_out[EAST] (signal_dspin_h_m2p_inc[x][y]); - clusters[x + 1][y]->p_m2p_in[WEST] (signal_dspin_h_m2p_inc[x][y]); - clusters[x][y]->p_m2p_in[EAST] (signal_dspin_h_m2p_dec[x][y]); - clusters[x + 1][y]->p_m2p_out[WEST] (signal_dspin_h_m2p_dec[x][y]); - - clusters[x][y]->p_p2m_out[EAST] (signal_dspin_h_p2m_inc[x][y]); - clusters[x + 1][y]->p_p2m_in[WEST] (signal_dspin_h_p2m_inc[x][y]); - clusters[x][y]->p_p2m_in[EAST] (signal_dspin_h_p2m_dec[x][y]); - clusters[x + 1][y]->p_p2m_out[WEST] (signal_dspin_h_p2m_dec[x][y]); - - clusters[x][y]->p_cla_out[EAST] (signal_dspin_h_cla_inc[x][y]); - clusters[x + 1][y]->p_cla_in[WEST] (signal_dspin_h_cla_inc[x][y]); - clusters[x][y]->p_cla_in[EAST] (signal_dspin_h_cla_dec[x][y]); - clusters[x + 1][y]->p_cla_out[WEST] (signal_dspin_h_cla_dec[x][y]); - } - } - std::cout << std::endl << "Horizontal connections done" << std::endl; - - // Inter Clusters vertical connections - for (int y = 0; y < Y_SIZE - 1; y++) { - for (int x = 0; x < X_SIZE; x++) { - clusters[x][y]->p_cmd_out[NORTH] (signal_dspin_v_cmd_inc[x][y]); - clusters[x][y + 1]->p_cmd_in[SOUTH] (signal_dspin_v_cmd_inc[x][y]); - clusters[x][y]->p_cmd_in[NORTH] (signal_dspin_v_cmd_dec[x][y]); - clusters[x][y + 1]->p_cmd_out[SOUTH] (signal_dspin_v_cmd_dec[x][y]); - - clusters[x][y]->p_rsp_out[NORTH] (signal_dspin_v_rsp_inc[x][y]); - clusters[x][y + 1]->p_rsp_in[SOUTH] (signal_dspin_v_rsp_inc[x][y]); - clusters[x][y]->p_rsp_in[NORTH] (signal_dspin_v_rsp_dec[x][y]); - clusters[x][y + 1]->p_rsp_out[SOUTH] (signal_dspin_v_rsp_dec[x][y]); - - clusters[x][y]->p_m2p_out[NORTH] (signal_dspin_v_m2p_inc[x][y]); - clusters[x][y + 1]->p_m2p_in[SOUTH] (signal_dspin_v_m2p_inc[x][y]); - clusters[x][y]->p_m2p_in[NORTH] (signal_dspin_v_m2p_dec[x][y]); - clusters[x][y + 1]->p_m2p_out[SOUTH] (signal_dspin_v_m2p_dec[x][y]); - - clusters[x][y]->p_p2m_out[NORTH] (signal_dspin_v_p2m_inc[x][y]); - clusters[x][y + 1]->p_p2m_in[SOUTH] (signal_dspin_v_p2m_inc[x][y]); - clusters[x][y]->p_p2m_in[NORTH] (signal_dspin_v_p2m_dec[x][y]); - clusters[x][y + 1]->p_p2m_out[SOUTH] (signal_dspin_v_p2m_dec[x][y]); - - clusters[x][y]->p_cla_out[NORTH] (signal_dspin_v_cla_inc[x][y]); - clusters[x][y + 1]->p_cla_in[SOUTH] (signal_dspin_v_cla_inc[x][y]); - clusters[x][y]->p_cla_in[NORTH] (signal_dspin_v_cla_dec[x][y]); - clusters[x][y + 1]->p_cla_out[SOUTH] (signal_dspin_v_cla_dec[x][y]); - } - } - std::cout << std::endl << "Vertical connections done" << std::endl; - - // East & West boundary cluster connections - for (size_t y = 0; y < Y_SIZE; y++) { - clusters[0][y]->p_cmd_in[WEST] (signal_dspin_bound_cmd_in[0][y][WEST]); - clusters[0][y]->p_cmd_out[WEST] (signal_dspin_bound_cmd_out[0][y][WEST]); - clusters[X_SIZE - 1][y]->p_cmd_in[EAST] (signal_dspin_bound_cmd_in[X_SIZE - 1][y][EAST]); - clusters[X_SIZE - 1][y]->p_cmd_out[EAST] (signal_dspin_bound_cmd_out[X_SIZE - 1][y][EAST]); - - clusters[0][y]->p_rsp_in[WEST] (signal_dspin_bound_rsp_in[0][y][WEST]); - clusters[0][y]->p_rsp_out[WEST] (signal_dspin_bound_rsp_out[0][y][WEST]); - clusters[X_SIZE - 1][y]->p_rsp_in[EAST] (signal_dspin_bound_rsp_in[X_SIZE - 1][y][EAST]); - clusters[X_SIZE - 1][y]->p_rsp_out[EAST] (signal_dspin_bound_rsp_out[X_SIZE - 1][y][EAST]); - - clusters[0][y]->p_m2p_in[WEST] (signal_dspin_bound_m2p_in[0][y][WEST]); - clusters[0][y]->p_m2p_out[WEST] (signal_dspin_bound_m2p_out[0][y][WEST]); - clusters[X_SIZE - 1][y]->p_m2p_in[EAST] (signal_dspin_bound_m2p_in[X_SIZE - 1][y][EAST]); - clusters[X_SIZE - 1][y]->p_m2p_out[EAST] (signal_dspin_bound_m2p_out[X_SIZE - 1][y][EAST]); - - clusters[0][y]->p_p2m_in[WEST] (signal_dspin_bound_p2m_in[0][y][WEST]); - clusters[0][y]->p_p2m_out[WEST] (signal_dspin_bound_p2m_out[0][y][WEST]); - clusters[X_SIZE - 1][y]->p_p2m_in[EAST] (signal_dspin_bound_p2m_in[X_SIZE - 1][y][EAST]); - clusters[X_SIZE - 1][y]->p_p2m_out[EAST] (signal_dspin_bound_p2m_out[X_SIZE - 1][y][EAST]); - - clusters[0][y]->p_cla_in[WEST] (signal_dspin_bound_cla_in[0][y][WEST]); - clusters[0][y]->p_cla_out[WEST] (signal_dspin_bound_cla_out[0][y][WEST]); - clusters[X_SIZE - 1][y]->p_cla_in[EAST] (signal_dspin_bound_cla_in[X_SIZE - 1][y][EAST]); - clusters[X_SIZE - 1][y]->p_cla_out[EAST] (signal_dspin_bound_cla_out[X_SIZE - 1][y][EAST]); - } - - std::cout << std::endl << "West & East boundaries connections done" << std::endl; - - // North & South boundary clusters connections - for (size_t x = 0; x < X_SIZE; x++) { - clusters[x][0]->p_cmd_in[SOUTH] (signal_dspin_bound_cmd_in[x][0][SOUTH]); - clusters[x][0]->p_cmd_out[SOUTH] (signal_dspin_bound_cmd_out[x][0][SOUTH]); - clusters[x][Y_SIZE - 1]->p_cmd_in[NORTH] (signal_dspin_bound_cmd_in[x][Y_SIZE - 1][NORTH]); - clusters[x][Y_SIZE - 1]->p_cmd_out[NORTH](signal_dspin_bound_cmd_out[x][Y_SIZE - 1][NORTH]); - - clusters[x][0]->p_rsp_in[SOUTH] (signal_dspin_bound_rsp_in[x][0][SOUTH]); - clusters[x][0]->p_rsp_out[SOUTH] (signal_dspin_bound_rsp_out[x][0][SOUTH]); - clusters[x][Y_SIZE - 1]->p_rsp_in[NORTH] (signal_dspin_bound_rsp_in[x][Y_SIZE - 1][NORTH]); - clusters[x][Y_SIZE - 1]->p_rsp_out[NORTH](signal_dspin_bound_rsp_out[x][Y_SIZE - 1][NORTH]); - - clusters[x][0]->p_m2p_in[SOUTH] (signal_dspin_bound_m2p_in[x][0][SOUTH]); - clusters[x][0]->p_m2p_out[SOUTH] (signal_dspin_bound_m2p_out[x][0][SOUTH]); - clusters[x][Y_SIZE - 1]->p_m2p_in[NORTH] (signal_dspin_bound_m2p_in[x][Y_SIZE - 1][NORTH]); - clusters[x][Y_SIZE - 1]->p_m2p_out[NORTH](signal_dspin_bound_m2p_out[x][Y_SIZE - 1][NORTH]); - - clusters[x][0]->p_p2m_in[SOUTH] (signal_dspin_bound_p2m_in[x][0][SOUTH]); - clusters[x][0]->p_p2m_out[SOUTH] (signal_dspin_bound_p2m_out[x][0][SOUTH]); - clusters[x][Y_SIZE - 1]->p_p2m_in[NORTH] (signal_dspin_bound_p2m_in[x][Y_SIZE - 1][NORTH]); - clusters[x][Y_SIZE - 1]->p_p2m_out[NORTH](signal_dspin_bound_p2m_out[x][Y_SIZE - 1][NORTH]); - - clusters[x][0]->p_cla_in[SOUTH] (signal_dspin_bound_cla_in[x][0][SOUTH]); - clusters[x][0]->p_cla_out[SOUTH] (signal_dspin_bound_cla_out[x][0][SOUTH]); - clusters[x][Y_SIZE - 1]->p_cla_in[NORTH] (signal_dspin_bound_cla_in[x][Y_SIZE - 1][NORTH]); - clusters[x][Y_SIZE - 1]->p_cla_out[NORTH](signal_dspin_bound_cla_out[x][Y_SIZE - 1][NORTH]); - } - - std::cout << std::endl << "North & South boundaries connections done" << std::endl; - std::cout << std::endl; + dspin_rsp_width, + vci_param_int, + vci_param_ext> + ( + sc.str().c_str(), + NB_PROCS_MAX, + NB_TTY_CHANNELS, + NB_DMA_CHANNELS, + x, + y, + cluster(x,y), + maptabd, + maptabx, + x_width, + y_width, + vci_srcid_width - x_width - y_width, // l_id width, + P_WIDTH, + RAM_TGTID, + XCU_TGTID, + DMA_TGTID, + FBF_TGTID, + TTY_TGTID, + ROM_TGTID, + NIC_TGTID, + CMA_TGTID, + IOC_TGTID, + SIM_TGTID, + MEMC_WAYS, + MEMC_SETS, + L1_IWAYS, + L1_ISETS, + L1_DWAYS, + L1_DSETS, + IRQ_PER_PROCESSOR, + XRAM_LATENCY, + x == X_IO && y == Y_IO, + FBF_X_SIZE, + FBF_Y_SIZE, + disk_name, + BDEV_SECTOR_SIZE, + NB_NIC_CHANNELS, + nic_rx_name, + nic_tx_name, + NIC_TIMEOUT, + NB_CMA_CHANNELS, + loader, + frozen_cycles, + debug_from, + debug_ok, + debug_ok + ); + + } + } + } + + /////////////////////////////////////////////////////////////// + // Net-list + /////////////////////////////////////////////////////////////// + + // Clock & RESET + for (int x = 0; x < X_SIZE; x++) { + for (int y = 0; y < Y_SIZE; y++) { + clusters[x][y]->p_clk (signal_clk); + clusters[x][y]->p_resetn (signal_resetn); + } + } + + // Inter Clusters horizontal connections + for (int x = 0; x < X_SIZE - 1; x++) { + for (int y = 0; y < Y_SIZE; y++) { + clusters[x][y]->p_cmd_out[EAST] (signal_dspin_h_cmd_inc[x][y]); + clusters[x + 1][y]->p_cmd_in[WEST] (signal_dspin_h_cmd_inc[x][y]); + clusters[x][y]->p_cmd_in[EAST] (signal_dspin_h_cmd_dec[x][y]); + clusters[x + 1][y]->p_cmd_out[WEST] (signal_dspin_h_cmd_dec[x][y]); + + clusters[x][y]->p_rsp_out[EAST] (signal_dspin_h_rsp_inc[x][y]); + clusters[x + 1][y]->p_rsp_in[WEST] (signal_dspin_h_rsp_inc[x][y]); + clusters[x][y]->p_rsp_in[EAST] (signal_dspin_h_rsp_dec[x][y]); + clusters[x + 1][y]->p_rsp_out[WEST] (signal_dspin_h_rsp_dec[x][y]); + + clusters[x][y]->p_m2p_out[EAST] (signal_dspin_h_m2p_inc[x][y]); + clusters[x + 1][y]->p_m2p_in[WEST] (signal_dspin_h_m2p_inc[x][y]); + clusters[x][y]->p_m2p_in[EAST] (signal_dspin_h_m2p_dec[x][y]); + clusters[x + 1][y]->p_m2p_out[WEST] (signal_dspin_h_m2p_dec[x][y]); + + clusters[x][y]->p_p2m_out[EAST] (signal_dspin_h_p2m_inc[x][y]); + clusters[x + 1][y]->p_p2m_in[WEST] (signal_dspin_h_p2m_inc[x][y]); + clusters[x][y]->p_p2m_in[EAST] (signal_dspin_h_p2m_dec[x][y]); + clusters[x + 1][y]->p_p2m_out[WEST] (signal_dspin_h_p2m_dec[x][y]); + + clusters[x][y]->p_cla_out[EAST] (signal_dspin_h_cla_inc[x][y]); + clusters[x + 1][y]->p_cla_in[WEST] (signal_dspin_h_cla_inc[x][y]); + clusters[x][y]->p_cla_in[EAST] (signal_dspin_h_cla_dec[x][y]); + clusters[x + 1][y]->p_cla_out[WEST] (signal_dspin_h_cla_dec[x][y]); + } + } + std::cout << std::endl << "Horizontal connections done" << std::endl; + + // Inter Clusters vertical connections + for (int y = 0; y < Y_SIZE - 1; y++) { + for (int x = 0; x < X_SIZE; x++) { + clusters[x][y]->p_cmd_out[NORTH] (signal_dspin_v_cmd_inc[x][y]); + clusters[x][y + 1]->p_cmd_in[SOUTH] (signal_dspin_v_cmd_inc[x][y]); + clusters[x][y]->p_cmd_in[NORTH] (signal_dspin_v_cmd_dec[x][y]); + clusters[x][y + 1]->p_cmd_out[SOUTH] (signal_dspin_v_cmd_dec[x][y]); + + clusters[x][y]->p_rsp_out[NORTH] (signal_dspin_v_rsp_inc[x][y]); + clusters[x][y + 1]->p_rsp_in[SOUTH] (signal_dspin_v_rsp_inc[x][y]); + clusters[x][y]->p_rsp_in[NORTH] (signal_dspin_v_rsp_dec[x][y]); + clusters[x][y + 1]->p_rsp_out[SOUTH] (signal_dspin_v_rsp_dec[x][y]); + + clusters[x][y]->p_m2p_out[NORTH] (signal_dspin_v_m2p_inc[x][y]); + clusters[x][y + 1]->p_m2p_in[SOUTH] (signal_dspin_v_m2p_inc[x][y]); + clusters[x][y]->p_m2p_in[NORTH] (signal_dspin_v_m2p_dec[x][y]); + clusters[x][y + 1]->p_m2p_out[SOUTH] (signal_dspin_v_m2p_dec[x][y]); + + clusters[x][y]->p_p2m_out[NORTH] (signal_dspin_v_p2m_inc[x][y]); + clusters[x][y + 1]->p_p2m_in[SOUTH] (signal_dspin_v_p2m_inc[x][y]); + clusters[x][y]->p_p2m_in[NORTH] (signal_dspin_v_p2m_dec[x][y]); + clusters[x][y + 1]->p_p2m_out[SOUTH] (signal_dspin_v_p2m_dec[x][y]); + + clusters[x][y]->p_cla_out[NORTH] (signal_dspin_v_cla_inc[x][y]); + clusters[x][y + 1]->p_cla_in[SOUTH] (signal_dspin_v_cla_inc[x][y]); + clusters[x][y]->p_cla_in[NORTH] (signal_dspin_v_cla_dec[x][y]); + clusters[x][y + 1]->p_cla_out[SOUTH] (signal_dspin_v_cla_dec[x][y]); + } + } + std::cout << std::endl << "Vertical connections done" << std::endl; + + // East & West boundary cluster connections + for (size_t y = 0; y < Y_SIZE; y++) { + clusters[0][y]->p_cmd_in[WEST] (signal_dspin_bound_cmd_in[0][y][WEST]); + clusters[0][y]->p_cmd_out[WEST] (signal_dspin_bound_cmd_out[0][y][WEST]); + clusters[X_SIZE - 1][y]->p_cmd_in[EAST] (signal_dspin_bound_cmd_in[X_SIZE - 1][y][EAST]); + clusters[X_SIZE - 1][y]->p_cmd_out[EAST] (signal_dspin_bound_cmd_out[X_SIZE - 1][y][EAST]); + + clusters[0][y]->p_rsp_in[WEST] (signal_dspin_bound_rsp_in[0][y][WEST]); + clusters[0][y]->p_rsp_out[WEST] (signal_dspin_bound_rsp_out[0][y][WEST]); + clusters[X_SIZE - 1][y]->p_rsp_in[EAST] (signal_dspin_bound_rsp_in[X_SIZE - 1][y][EAST]); + clusters[X_SIZE - 1][y]->p_rsp_out[EAST] (signal_dspin_bound_rsp_out[X_SIZE - 1][y][EAST]); + + clusters[0][y]->p_m2p_in[WEST] (signal_dspin_bound_m2p_in[0][y][WEST]); + clusters[0][y]->p_m2p_out[WEST] (signal_dspin_bound_m2p_out[0][y][WEST]); + clusters[X_SIZE - 1][y]->p_m2p_in[EAST] (signal_dspin_bound_m2p_in[X_SIZE - 1][y][EAST]); + clusters[X_SIZE - 1][y]->p_m2p_out[EAST] (signal_dspin_bound_m2p_out[X_SIZE - 1][y][EAST]); + + clusters[0][y]->p_p2m_in[WEST] (signal_dspin_bound_p2m_in[0][y][WEST]); + clusters[0][y]->p_p2m_out[WEST] (signal_dspin_bound_p2m_out[0][y][WEST]); + clusters[X_SIZE - 1][y]->p_p2m_in[EAST] (signal_dspin_bound_p2m_in[X_SIZE - 1][y][EAST]); + clusters[X_SIZE - 1][y]->p_p2m_out[EAST] (signal_dspin_bound_p2m_out[X_SIZE - 1][y][EAST]); + + clusters[0][y]->p_cla_in[WEST] (signal_dspin_bound_cla_in[0][y][WEST]); + clusters[0][y]->p_cla_out[WEST] (signal_dspin_bound_cla_out[0][y][WEST]); + clusters[X_SIZE - 1][y]->p_cla_in[EAST] (signal_dspin_bound_cla_in[X_SIZE - 1][y][EAST]); + clusters[X_SIZE - 1][y]->p_cla_out[EAST] (signal_dspin_bound_cla_out[X_SIZE - 1][y][EAST]); + } + + std::cout << std::endl << "West & East boundaries connections done" << std::endl; + + // North & South boundary clusters connections + for (size_t x = 0; x < X_SIZE; x++) { + clusters[x][0]->p_cmd_in[SOUTH] (signal_dspin_bound_cmd_in[x][0][SOUTH]); + clusters[x][0]->p_cmd_out[SOUTH] (signal_dspin_bound_cmd_out[x][0][SOUTH]); + clusters[x][Y_SIZE - 1]->p_cmd_in[NORTH] (signal_dspin_bound_cmd_in[x][Y_SIZE - 1][NORTH]); + clusters[x][Y_SIZE - 1]->p_cmd_out[NORTH](signal_dspin_bound_cmd_out[x][Y_SIZE - 1][NORTH]); + + clusters[x][0]->p_rsp_in[SOUTH] (signal_dspin_bound_rsp_in[x][0][SOUTH]); + clusters[x][0]->p_rsp_out[SOUTH] (signal_dspin_bound_rsp_out[x][0][SOUTH]); + clusters[x][Y_SIZE - 1]->p_rsp_in[NORTH] (signal_dspin_bound_rsp_in[x][Y_SIZE - 1][NORTH]); + clusters[x][Y_SIZE - 1]->p_rsp_out[NORTH](signal_dspin_bound_rsp_out[x][Y_SIZE - 1][NORTH]); + + clusters[x][0]->p_m2p_in[SOUTH] (signal_dspin_bound_m2p_in[x][0][SOUTH]); + clusters[x][0]->p_m2p_out[SOUTH] (signal_dspin_bound_m2p_out[x][0][SOUTH]); + clusters[x][Y_SIZE - 1]->p_m2p_in[NORTH] (signal_dspin_bound_m2p_in[x][Y_SIZE - 1][NORTH]); + clusters[x][Y_SIZE - 1]->p_m2p_out[NORTH](signal_dspin_bound_m2p_out[x][Y_SIZE - 1][NORTH]); + + clusters[x][0]->p_p2m_in[SOUTH] (signal_dspin_bound_p2m_in[x][0][SOUTH]); + clusters[x][0]->p_p2m_out[SOUTH] (signal_dspin_bound_p2m_out[x][0][SOUTH]); + clusters[x][Y_SIZE - 1]->p_p2m_in[NORTH] (signal_dspin_bound_p2m_in[x][Y_SIZE - 1][NORTH]); + clusters[x][Y_SIZE - 1]->p_p2m_out[NORTH](signal_dspin_bound_p2m_out[x][Y_SIZE - 1][NORTH]); + + clusters[x][0]->p_cla_in[SOUTH] (signal_dspin_bound_cla_in[x][0][SOUTH]); + clusters[x][0]->p_cla_out[SOUTH] (signal_dspin_bound_cla_out[x][0][SOUTH]); + clusters[x][Y_SIZE - 1]->p_cla_in[NORTH] (signal_dspin_bound_cla_in[x][Y_SIZE - 1][NORTH]); + clusters[x][Y_SIZE - 1]->p_cla_out[NORTH](signal_dspin_bound_cla_out[x][Y_SIZE - 1][NORTH]); + } + + std::cout << std::endl << "North & South boundaries connections done" << std::endl; + std::cout << std::endl; @@ -865,17 +822,17 @@ GdbServer > * > l1_caches; - for (int x = 0; x < X_SIZE; x++) { - for (int y = 0; y < Y_SIZE; y++) { - for (int proc = 0; proc < NB_PROCS_MAX; proc++) { - l1_caches.push_back(clusters[x][y]->proc[proc]); - } - } - } - - for (int x = 0; x < X_SIZE; x++) { - for (int y = 0; y < Y_SIZE; y++) { - clusters[x][y]->memc->set_vcache_list(l1_caches); - } - } + for (int x = 0; x < X_SIZE; x++) { + for (int y = 0; y < Y_SIZE; y++) { + for (int proc = 0; proc < NB_PROCS_MAX; proc++) { + l1_caches.push_back(clusters[x][y]->proc[proc]); + } + } + } + + for (int x = 0; x < X_SIZE; x++) { + for (int y = 0; y < Y_SIZE; y++) { + clusters[x][y]->memc->set_vcache_list(l1_caches); + } + } #endif @@ -883,289 +840,285 @@ //#define SC_TRACE #ifdef SC_TRACE - sc_trace_file * tf = sc_create_vcd_trace_file("my_trace_file"); - - for (int x = 0; x < X_SIZE - 1; x++) { - for (int y = 0; y < Y_SIZE; y++) { - for (int k = 0; k < 3; k++) { - signal_dspin_h_cmd_inc[x][y][k].trace(tf, "dspin_h_cmd_inc"); - signal_dspin_h_cmd_dec[x][y][k].trace(tf, "dspin_h_cmd_dec"); - } - - for (int k = 0; k < 2; k++) { - signal_dspin_h_rsp_inc[x][y][k].trace(tf, "dspin_h_rsp_inc"); - signal_dspin_h_rsp_dec[x][y][k].trace(tf, "dspin_h_rsp_dec"); - } - } - } - - for (int y = 0; y < Y_SIZE - 1; y++) { - for (int x = 0; x < X_SIZE; x++) { - for (int k = 0; k < 3; k++) { - signal_dspin_v_cmd_inc[x][y][k].trace(tf, "dspin_v_cmd_inc"); - signal_dspin_v_cmd_dec[x][y][k].trace(tf, "dspin_v_cmd_dec"); - } - - for (int k = 0; k < 2; k++) { - signal_dspin_v_rsp_inc[x][y][k].trace(tf, "dspin_v_rsp_inc"); - signal_dspin_v_rsp_dec[x][y][k].trace(tf, "dspin_v_rsp_dec"); - } - } - } - - for (int x = 0; x < (X_SIZE); x++) { - for (int y = 0; y < Y_SIZE; y++) { - std::ostringstream signame; - signame << "cluster" << x << "_" << y; - clusters[x][y]->trace(tf, signame.str()); - } - } -#endif - - - //////////////////////////////////////////////////////// - // Simulation - /////////////////////////////////////////////////////// - - sc_start(sc_core::sc_time(0, SC_NS)); - signal_resetn = false; - - // set network boundaries signals default values - // for all boundary clusters - for (size_t x = 0; x < X_SIZE ; x++) { - for (size_t y = 0; y < Y_SIZE ; y++) { - for (size_t face = 0; face < 4; face++) { - signal_dspin_bound_cmd_in [x][y][face].write = false; - signal_dspin_bound_cmd_in [x][y][face].read = true; - signal_dspin_bound_cmd_out[x][y][face].write = false; - signal_dspin_bound_cmd_out[x][y][face].read = true; - - signal_dspin_bound_rsp_in [x][y][face].write = false; - signal_dspin_bound_rsp_in [x][y][face].read = true; - signal_dspin_bound_rsp_out[x][y][face].write = false; - signal_dspin_bound_rsp_out[x][y][face].read = true; - - signal_dspin_bound_m2p_in [x][y][face].write = false; - signal_dspin_bound_m2p_in [x][y][face].read = true; - signal_dspin_bound_m2p_out[x][y][face].write = false; - signal_dspin_bound_m2p_out[x][y][face].read = true; - - signal_dspin_bound_p2m_in [x][y][face].write = false; - signal_dspin_bound_p2m_in [x][y][face].read = true; - signal_dspin_bound_p2m_out[x][y][face].write = false; - signal_dspin_bound_p2m_out[x][y][face].read = true; - - signal_dspin_bound_cla_in [x][y][face].write = false; - signal_dspin_bound_cla_in [x][y][face].read = true; - signal_dspin_bound_cla_out[x][y][face].write = false; - signal_dspin_bound_cla_out[x][y][face].read = true; - } - } - } - // @M debug fu** - clusters[0][0]->signal_dspin_m2p_proc[2].read = true; - - sc_start(sc_core::sc_time(1, SC_NS)); - signal_resetn = true; - - if (debug_ok) { - #if USE_OPENMP - assert(false && "OPEN MP should not be used with debug because of its traces"); - #endif - - if (gettimeofday(&t1, NULL) != 0) { - perror("gettimeofday"); - return EXIT_FAILURE; - } - - for (int64_t n = 1; n < ncycles && !stop_called; n++) { - if ((n % max_cycles) == 0) { + sc_trace_file * tf = sc_create_vcd_trace_file("my_trace_file"); + + for (int x = 0; x < X_SIZE - 1; x++) { + for (int y = 0; y < Y_SIZE; y++) { + for (int k = 0; k < 3; k++) { + signal_dspin_h_cmd_inc[x][y][k].trace(tf, "dspin_h_cmd_inc"); + signal_dspin_h_cmd_dec[x][y][k].trace(tf, "dspin_h_cmd_dec"); + } + + for (int k = 0; k < 2; k++) { + signal_dspin_h_rsp_inc[x][y][k].trace(tf, "dspin_h_rsp_inc"); + signal_dspin_h_rsp_dec[x][y][k].trace(tf, "dspin_h_rsp_dec"); + } + } + } + + for (int y = 0; y < Y_SIZE - 1; y++) { + for (int x = 0; x < X_SIZE; x++) { + for (int k = 0; k < 3; k++) { + signal_dspin_v_cmd_inc[x][y][k].trace(tf, "dspin_v_cmd_inc"); + signal_dspin_v_cmd_dec[x][y][k].trace(tf, "dspin_v_cmd_dec"); + } + + for (int k = 0; k < 2; k++) { + signal_dspin_v_rsp_inc[x][y][k].trace(tf, "dspin_v_rsp_inc"); + signal_dspin_v_rsp_dec[x][y][k].trace(tf, "dspin_v_rsp_dec"); + } + } + } + + for (int x = 0; x < (X_SIZE); x++) { + for (int y = 0; y < Y_SIZE; y++) { + std::ostringstream signame; + signame << "cluster" << x << "_" << y; + clusters[x][y]->trace(tf, signame.str()); + } + } +#endif + + + //////////////////////////////////////////////////////// + // Simulation + /////////////////////////////////////////////////////// + + sc_start(sc_core::sc_time(0, SC_NS)); + signal_resetn = false; + + // set network boundaries signals default values + // for all boundary clusters + for (size_t x = 0; x < X_SIZE ; x++) { + for (size_t y = 0; y < Y_SIZE ; y++) { + for (size_t face = 0; face < 4; face++) { + signal_dspin_bound_cmd_in [x][y][face].write = false; + signal_dspin_bound_cmd_in [x][y][face].read = true; + signal_dspin_bound_cmd_out[x][y][face].write = false; + signal_dspin_bound_cmd_out[x][y][face].read = true; + + signal_dspin_bound_rsp_in [x][y][face].write = false; + signal_dspin_bound_rsp_in [x][y][face].read = true; + signal_dspin_bound_rsp_out[x][y][face].write = false; + signal_dspin_bound_rsp_out[x][y][face].read = true; + + signal_dspin_bound_m2p_in [x][y][face].write = false; + signal_dspin_bound_m2p_in [x][y][face].read = true; + signal_dspin_bound_m2p_out[x][y][face].write = false; + signal_dspin_bound_m2p_out[x][y][face].read = true; + + signal_dspin_bound_p2m_in [x][y][face].write = false; + signal_dspin_bound_p2m_in [x][y][face].read = true; + signal_dspin_bound_p2m_out[x][y][face].write = false; + signal_dspin_bound_p2m_out[x][y][face].read = true; + + signal_dspin_bound_cla_in [x][y][face].write = false; + signal_dspin_bound_cla_in [x][y][face].read = true; + signal_dspin_bound_cla_out[x][y][face].write = false; + signal_dspin_bound_cla_out[x][y][face].read = true; + } + } + } + // @QM : what is the following line? + //clusters[0][0]->signal_dspin_m2p_proc[2].read = true; + + sc_start(sc_core::sc_time(1, SC_NS)); + signal_resetn = true; + + if (debug_ok) { +#ifdef USE_OPENMP + assert(false && "OPEN MP should not be used with debug because of its traces"); +#endif + + if (gettimeofday(&t1, NULL) != 0) { + perror("gettimeofday"); + return EXIT_FAILURE; + } + + for (int64_t n = 1; n < ncycles && !stop_called; n++) { + if ((n % max_cycles) == 0) { + + if (gettimeofday(&t2, NULL) != 0) { + perror("gettimeofday"); + return EXIT_FAILURE; + } + + ms1 = (uint64_t) t1.tv_sec * 1000ULL + (uint64_t) t1.tv_usec / 1000; + ms2 = (uint64_t) t2.tv_sec * 1000ULL + (uint64_t) t2.tv_usec / 1000; + std::cerr << "platform clock frequency " << (double) 5000000 / (double) (ms2 - ms1) << "Khz" << std::endl; + + if (gettimeofday(&t1, NULL) != 0) { + perror("gettimeofday"); + return EXIT_FAILURE; + } + } + + + if (n == reset_counters) { + for (size_t x = 0; x < (X_SIZE); x++) { + for (size_t y = 0; y < Y_SIZE; y++) { + clusters[x][y]->memc->reset_counters(); + } + } + } + + if (n == dump_counters) { + for (size_t x = 0; x < (X_SIZE); x++) { + for (size_t y = 0; y < Y_SIZE; y++) { + clusters[x][y]->memc->print_stats(true, false); + } + } + } + + if ((n > debug_from) and (n % debug_period == 0)) { + std::cout << "****************** cycle " << std::dec << n ; + std::cout << "************************************************" << std::endl; + + for (size_t x = 0; x < X_SIZE ; x++) { + for (size_t y = 0; y < Y_SIZE ; y++) { + for (int proc = 0; proc < NB_PROCS_MAX; proc++) { + clusters[x][y]->proc[proc]->print_trace(); + std::ostringstream proc_signame; + proc_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc ; + std::ostringstream p2m_signame; + p2m_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc << " P2M"; + std::ostringstream m2p_signame; + m2p_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc << " M2P"; + + clusters[x][y]->signal_vci_ini_proc[proc].print_trace(proc_signame.str()); + clusters[x][y]->signal_dspin_p2m_proc[proc].print_trace(p2m_signame.str()); + clusters[x][y]->signal_dspin_m2p_proc[proc].print_trace(m2p_signame.str()); + } + + clusters[x][y]->memc->print_trace(); + + std::ostringstream smemc; + smemc << "[SIG]MEMC_" << x << "_" << y; + std::ostringstream sxram; + sxram << "[SIG]XRAM_" << x << "_" << y; + std::ostringstream sm2p; + sm2p << "[SIG]MEMC_" << x << "_" << y << " M2P"; + std::ostringstream sp2m; + sp2m << "[SIG]MEMC_" << x << "_" << y << " P2M"; + + clusters[x][y]->signal_vci_tgt_memc.print_trace(smemc.str()); + clusters[x][y]->signal_vci_xram.print_trace(sxram.str()); + clusters[x][y]->signal_dspin_p2m_memc.print_trace(sp2m.str()); + clusters[x][y]->signal_dspin_m2p_memc.print_trace(sm2p.str()); + } + } + } + + sc_start(sc_core::sc_time(1, SC_NS)); + } + } + else { + int64_t n = 0; + while (!stop_called && n != ncycles) { + if (gettimeofday(&t1, NULL) != 0) { + perror("gettimeofday"); + return EXIT_FAILURE; + } + int64_t nb_cycles = min(max_cycles, ncycles - n); + if (do_reset_counters) { + nb_cycles = min(nb_cycles, reset_counters - n); + } + if (do_dump_counters) { + nb_cycles = min(nb_cycles, dump_counters - n); + } + + sc_start(sc_core::sc_time(nb_cycles, SC_NS)); + n += nb_cycles; + + if (do_reset_counters && n == reset_counters) { + // Reseting counters + for (size_t x = 0; x < (X_SIZE); x++) { + for (size_t y = 0; y < Y_SIZE; y++) { + clusters[x][y]->memc->reset_counters(); + } + } + do_reset_counters = false; + } + + if (do_dump_counters && n == dump_counters) { + // Dumping counters + for (size_t x = 0; x < (X_SIZE); x++) { + for (size_t y = 0; y < Y_SIZE; y++) { + clusters[x][y]->memc->print_stats(true, false); + } + } + do_dump_counters = false; + } + if (gettimeofday(&t2, NULL) != 0) { - perror("gettimeofday"); - return EXIT_FAILURE; - } - + perror("gettimeofday"); + return EXIT_FAILURE; + } ms1 = (uint64_t) t1.tv_sec * 1000ULL + (uint64_t) t1.tv_usec / 1000; ms2 = (uint64_t) t2.tv_sec * 1000ULL + (uint64_t) t2.tv_usec / 1000; - std::cerr << "platform clock frequency " << (double) 5000000 / (double) (ms2 - ms1) << "Khz" << std::endl; - - if (gettimeofday(&t1, NULL) != 0) - { - perror("gettimeofday"); - return EXIT_FAILURE; - } - } - - - if (n == reset_counters) { - for (size_t x = 0; x < (X_SIZE); x++) { - for (size_t y = 0; y < Y_SIZE; y++) { - clusters[x][y]->memc->reset_counters(); - } - } - } - - if (n == dump_counters) { - for (size_t x = 0; x < (X_SIZE); x++) { - for (size_t y = 0; y < Y_SIZE; y++) { - clusters[x][y]->memc->print_stats(true, false); - } - } - } - - if ((n > debug_from) and (n % debug_period == 0)) { - std::cout << "****************** cycle " << std::dec << n ; - std::cout << "************************************************" << std::endl; - - for (size_t x = 0; x < X_SIZE ; x++) { - for (size_t y = 0; y < Y_SIZE ; y++) { - for (int proc = 0; proc < NB_PROCS_MAX; proc++) { - if (x == 0 && y == 0 && proc == 2) { - continue; - } - clusters[x][y]->proc[proc]->print_trace(); - std::ostringstream proc_signame; - proc_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc ; - std::ostringstream p2m_signame; - p2m_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc << " P2M"; - std::ostringstream m2p_signame; - m2p_signame << "[SIG]PROC_" << x << "_" << y << "_" << proc << " M2P"; - - clusters[x][y]->signal_vci_ini_proc[proc].print_trace(proc_signame.str()); - clusters[x][y]->signal_dspin_p2m_proc[proc].print_trace(p2m_signame.str()); - clusters[x][y]->signal_dspin_m2p_proc[proc].print_trace(m2p_signame.str()); - } - - clusters[x][y]->memc->print_trace(); - - std::ostringstream smemc; - smemc << "[SIG]MEMC_" << x << "_" << y; - std::ostringstream sxram; - sxram << "[SIG]XRAM_" << x << "_" << y; - std::ostringstream sm2p; - sm2p << "[SIG]MEMC_" << x << "_" << y << " M2P"; - std::ostringstream sp2m; - sp2m << "[SIG]MEMC_" << x << "_" << y << " P2M"; - - clusters[x][y]->signal_vci_tgt_memc.print_trace(smemc.str()); - clusters[x][y]->signal_vci_xram.print_trace(sxram.str()); - clusters[x][y]->signal_dspin_p2m_memc.print_trace(sp2m.str()); - clusters[x][y]->signal_dspin_m2p_memc.print_trace(sm2p.str()); - } - } - } - - sc_start(sc_core::sc_time(1, SC_NS)); - } - } - else { - int64_t n = 0; - while (!stop_called && n != ncycles) { - if (gettimeofday(&t1, NULL) != 0) { - perror("gettimeofday"); - return EXIT_FAILURE; - } - int64_t nb_cycles = min(max_cycles, ncycles - n); - if (do_reset_counters) { - nb_cycles = min(nb_cycles, reset_counters - n); - } - if (do_dump_counters) { - nb_cycles = min(nb_cycles, dump_counters - n); - } - - sc_start(sc_core::sc_time(nb_cycles, SC_NS)); - n += nb_cycles; - - if (do_reset_counters && n == reset_counters) { - // Reseting counters - for (size_t x = 0; x < (X_SIZE); x++) { - for (size_t y = 0; y < Y_SIZE; y++) { - clusters[x][y]->memc->reset_counters(); - } - } - do_reset_counters = false; - } - - if (do_dump_counters && n == dump_counters) { - // Dumping counters - for (size_t x = 0; x < (X_SIZE); x++) { - for (size_t y = 0; y < Y_SIZE; y++) { - clusters[x][y]->memc->print_stats(true, false); - } - } - do_dump_counters = false; - } - - - if (gettimeofday(&t2, NULL) != 0) { - perror("gettimeofday"); - return EXIT_FAILURE; - } - ms1 = (uint64_t) t1.tv_sec * 1000ULL + (uint64_t) t1.tv_usec / 1000; - ms2 = (uint64_t) t2.tv_sec * 1000ULL + (uint64_t) t2.tv_usec / 1000; - std::cerr << std::dec << "cycle " << n << " platform clock frequency " << (double) nb_cycles / (double) (ms2 - ms1) << "Khz" << std::endl; - } - } - - - // Free memory - for (size_t i = 0; i < (X_SIZE * Y_SIZE); i++) { - size_t x = i / Y_SIZE; - size_t y = i % Y_SIZE; - delete clusters[x][y]; - } - - dealloc_elems >(signal_dspin_h_cmd_inc, X_SIZE - 1, Y_SIZE); - dealloc_elems >(signal_dspin_h_cmd_dec, X_SIZE - 1, Y_SIZE); - - dealloc_elems >(signal_dspin_h_rsp_inc, X_SIZE - 1, Y_SIZE); - dealloc_elems >(signal_dspin_h_rsp_dec, X_SIZE - 1, Y_SIZE); - - dealloc_elems >(signal_dspin_h_m2p_inc, X_SIZE - 1, Y_SIZE); - dealloc_elems >(signal_dspin_h_m2p_dec, X_SIZE - 1, Y_SIZE); - - dealloc_elems >(signal_dspin_h_p2m_inc, X_SIZE - 1, Y_SIZE); - dealloc_elems >(signal_dspin_h_p2m_dec, X_SIZE - 1, Y_SIZE); - - dealloc_elems >(signal_dspin_h_cla_inc, X_SIZE - 1, Y_SIZE); - dealloc_elems >(signal_dspin_h_cla_dec, X_SIZE - 1, Y_SIZE); - - dealloc_elems >(signal_dspin_v_cmd_inc, X_SIZE, Y_SIZE - 1); - dealloc_elems >(signal_dspin_v_cmd_dec, X_SIZE, Y_SIZE - 1); - - dealloc_elems >(signal_dspin_v_rsp_inc, X_SIZE, Y_SIZE - 1); - dealloc_elems >(signal_dspin_v_rsp_dec, X_SIZE, Y_SIZE - 1); - - dealloc_elems >(signal_dspin_v_m2p_inc, X_SIZE, Y_SIZE - 1); - dealloc_elems >(signal_dspin_v_m2p_dec, X_SIZE, Y_SIZE - 1); - - dealloc_elems >(signal_dspin_v_p2m_inc, X_SIZE, Y_SIZE - 1); - dealloc_elems >(signal_dspin_v_p2m_dec, X_SIZE, Y_SIZE - 1); - - dealloc_elems >(signal_dspin_v_cla_inc, X_SIZE, Y_SIZE - 1); - dealloc_elems >(signal_dspin_v_cla_dec, X_SIZE, Y_SIZE - 1); - - dealloc_elems >(signal_dspin_bound_cmd_in, X_SIZE, Y_SIZE, 4); - dealloc_elems >(signal_dspin_bound_cmd_out, X_SIZE, Y_SIZE, 4); - - dealloc_elems >(signal_dspin_bound_rsp_in, X_SIZE, Y_SIZE, 4); - dealloc_elems >(signal_dspin_bound_rsp_out, X_SIZE, Y_SIZE, 4); - - dealloc_elems >(signal_dspin_bound_m2p_in, X_SIZE, Y_SIZE, 4); - dealloc_elems >(signal_dspin_bound_m2p_out, X_SIZE, Y_SIZE, 4); - - dealloc_elems >(signal_dspin_bound_p2m_in, X_SIZE, Y_SIZE, 4); - dealloc_elems >(signal_dspin_bound_p2m_out, X_SIZE, Y_SIZE, 4); - - dealloc_elems >(signal_dspin_bound_cla_in, X_SIZE, Y_SIZE, 4); - dealloc_elems >(signal_dspin_bound_cla_out, X_SIZE, Y_SIZE, 4); - - return EXIT_SUCCESS; + std::cerr << std::dec << "cycle " << n << " platform clock frequency " << (double) nb_cycles / (double) (ms2 - ms1) << "Khz" << std::endl; + } + } + + + // Free memory + for (size_t i = 0; i < (X_SIZE * Y_SIZE); i++) { + size_t x = i / Y_SIZE; + size_t y = i % Y_SIZE; + delete clusters[x][y]; + } + + dealloc_elems >(signal_dspin_h_cmd_inc, X_SIZE - 1, Y_SIZE); + dealloc_elems >(signal_dspin_h_cmd_dec, X_SIZE - 1, Y_SIZE); + + dealloc_elems >(signal_dspin_h_rsp_inc, X_SIZE - 1, Y_SIZE); + dealloc_elems >(signal_dspin_h_rsp_dec, X_SIZE - 1, Y_SIZE); + + dealloc_elems >(signal_dspin_h_m2p_inc, X_SIZE - 1, Y_SIZE); + dealloc_elems >(signal_dspin_h_m2p_dec, X_SIZE - 1, Y_SIZE); + + dealloc_elems >(signal_dspin_h_p2m_inc, X_SIZE - 1, Y_SIZE); + dealloc_elems >(signal_dspin_h_p2m_dec, X_SIZE - 1, Y_SIZE); + + dealloc_elems >(signal_dspin_h_cla_inc, X_SIZE - 1, Y_SIZE); + dealloc_elems >(signal_dspin_h_cla_dec, X_SIZE - 1, Y_SIZE); + + dealloc_elems >(signal_dspin_v_cmd_inc, X_SIZE, Y_SIZE - 1); + dealloc_elems >(signal_dspin_v_cmd_dec, X_SIZE, Y_SIZE - 1); + + dealloc_elems >(signal_dspin_v_rsp_inc, X_SIZE, Y_SIZE - 1); + dealloc_elems >(signal_dspin_v_rsp_dec, X_SIZE, Y_SIZE - 1); + + dealloc_elems >(signal_dspin_v_m2p_inc, X_SIZE, Y_SIZE - 1); + dealloc_elems >(signal_dspin_v_m2p_dec, X_SIZE, Y_SIZE - 1); + + dealloc_elems >(signal_dspin_v_p2m_inc, X_SIZE, Y_SIZE - 1); + dealloc_elems >(signal_dspin_v_p2m_dec, X_SIZE, Y_SIZE - 1); + + dealloc_elems >(signal_dspin_v_cla_inc, X_SIZE, Y_SIZE - 1); + dealloc_elems >(signal_dspin_v_cla_dec, X_SIZE, Y_SIZE - 1); + + dealloc_elems >(signal_dspin_bound_cmd_in, X_SIZE, Y_SIZE, 4); + dealloc_elems >(signal_dspin_bound_cmd_out, X_SIZE, Y_SIZE, 4); + + dealloc_elems >(signal_dspin_bound_rsp_in, X_SIZE, Y_SIZE, 4); + dealloc_elems >(signal_dspin_bound_rsp_out, X_SIZE, Y_SIZE, 4); + + dealloc_elems >(signal_dspin_bound_m2p_in, X_SIZE, Y_SIZE, 4); + dealloc_elems >(signal_dspin_bound_m2p_out, X_SIZE, Y_SIZE, 4); + + dealloc_elems >(signal_dspin_bound_p2m_in, X_SIZE, Y_SIZE, 4); + dealloc_elems >(signal_dspin_bound_p2m_out, X_SIZE, Y_SIZE, 4); + + dealloc_elems >(signal_dspin_bound_cla_in, X_SIZE, Y_SIZE, 4); + dealloc_elems >(signal_dspin_bound_cla_out, X_SIZE, Y_SIZE, 4); + + return EXIT_SUCCESS; } void handler(int dummy = 0) { - stop_called = true; - sc_stop(); + stop_called = true; + sc_stop(); } @@ -1173,26 +1126,31 @@ int sc_main (int argc, char *argv[]) { - signal(SIGINT, handler); - signal(SIGPIPE, voidhandler); - - try { - return _main(argc, argv); - } catch (std::exception &e) { - std::cout << e.what() << std::endl; - } - catch (...) { - std::cout << "Unknown exception occured" << std::endl; - throw; - } - return 1; + signal(SIGINT, handler); + signal(SIGPIPE, voidhandler); + + try { + int ret =_main(argc, argv); + if (!stop_called) { + sc_stop(); + sc_start(sc_core::sc_time(0, SC_NS)); + } + return ret; + } catch (std::exception &e) { + std::cout << e.what() << std::endl; + } + catch (...) { + std::cout << "Unknown exception occured" << std::endl; + throw; + } + return 1; } // Local Variables: -// tab-width: 3 -// c-basic-offset: 3 +// tab-width: 4 +// c-basic-offset: 4 // c-file-offsets:((innamespace . 0)(inline-open . 0)) // indent-tabs-mode: nil // End: -// vim: filetype=cpp:expandtab:shiftwidth=3:tabstop=3:softtabstop=3 +// vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=4:softtabstop=4 Index: trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/metadata/tsar_xbar_cluster.sd =================================================================== --- trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/metadata/tsar_xbar_cluster.sd (revision 1047) +++ trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/metadata/tsar_xbar_cluster.sd (revision 1048) @@ -35,9 +35,9 @@ memc_dspin_out_width = parameter.Reference('dspin_cmd_width')), + Uses('caba:vci_simple_ram', + cell_size = parameter.Reference('vci_data_width_ext')), + Uses('caba:vci_simple_rom', cell_size = parameter.Reference('vci_data_width_int')), - - Uses('caba:vci_simple_ram', - cell_size = parameter.Reference('vci_data_width_ext')), Uses('caba:vci_simple_ram', Index: trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/source/src/tsar_xbar_cluster.cpp =================================================================== --- trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/source/src/tsar_xbar_cluster.cpp (revision 1047) +++ trunk/platforms/tsar_generic_xbar/tsar_xbar_cluster/caba/source/src/tsar_xbar_cluster.cpp (revision 1048) @@ -156,5 +156,5 @@ y_width, // Number of y bits in platform memc_ways, memc_sets, 16, // CACHE SIZE - 3, // MAX NUMBER OF COPIES + 4, // MAX NUMBER OF COPIES 4096, // HEAP SIZE 8, // TRANSACTION TABLE DEPTH @@ -343,5 +343,4 @@ 64); // burst size - int mac = 0xBEEF0000; mnic = new VciMultiNic( "mnic", @@ -349,8 +348,8 @@ mtd, nic_channels, - mac, // mac_4 address - 0xBABE, // mac_2 address - nic_rx_name, - nic_tx_name); + 0, // mac_4 address + 0, // mac_2 address + 1, // NIC_MODE_SYNTHESIS + 12); // INTER_FRAME_GAP @@ -709,4 +708,7 @@ delete router_cmd; delete router_rsp; + delete router_m2p; + delete router_p2m; + delete router_cla; if (brom != NULL) { delete brom;