improve matplotlib code

add new helper functions to get all y_values for a plot
use seperate figures for each plot
s/filepostfix/file_postfix/ to be consistent

17 files changed, 97 insertions, 114 deletions

diff --git a/src/benchmarks/blowup.py b/src/benchmarks/blowup.py
index cc85643..28692d6 100644
--- a/src/benchmarks/blowup.py
+++ b/src/benchmarks/blowup.py
@@ -58,7 +58,7 @@ class BenchmarkBlowup(Benchmark):
                                "{VmHWM}/1000",
                                ylabel="VmHWM in MB",
                                title="blowup test",
-                               filepostfix="vmhwm")
+                               file_postfix="vmhwm")
 
         plt.pgfplot(self,
                     self.iterate_args(self.results["args"]),
diff --git a/src/benchmarks/cfrac.py b/src/benchmarks/cfrac.py
index 07385eb..9b4bc64 100644
--- a/src/benchmarks/cfrac.py
+++ b/src/benchmarks/cfrac.py
@@ -80,7 +80,7 @@ class BenchmarkCfrac(Benchmark):
                                "{task-clock}/1000",
                                ylabel='"cpu-second"',
                                title='"Cfrac: runtime"',
-                               filepostfix="time")
+                               file_postfix="time")
 
         # L1 cache misses
         plt.barplot_single_arg(
@@ -88,7 +88,7 @@ class BenchmarkCfrac(Benchmark):
             "({L1-dcache-load-misses}/{L1-dcache-loads})*100",
             ylabel="L1 misses in %",
             title="Cfrac l1 cache misses",
-            filepostfix="l1misses",
+            file_postfix="l1misses",
             yerr=False)
 
         # Memusage
@@ -96,7 +96,7 @@ class BenchmarkCfrac(Benchmark):
                                "{VmHWM}",
                                ylabel="VmHWM in KB",
                                title="Cfrac VmHWM",
-                               filepostfix="vmhwm")
+                               file_postfix="vmhwm")
 
         plt.write_tex_table(self, [{
             "label": "Runtime [ms]",
@@ -107,7 +107,7 @@ class BenchmarkCfrac(Benchmark):
             "expression": "{VmHWM}",
             "sort": "<"
         }],
-                             filepostfix="table")
+                             file_postfix="table")
 
         plt.export_stats_to_dataref(self, "task-clock")
 
diff --git a/src/benchmarks/dj_trace.py b/src/benchmarks/dj_trace.py
index 4730db3..b6ae83d 100644
--- a/src/benchmarks/dj_trace.py
+++ b/src/benchmarks/dj_trace.py
@@ -200,7 +200,7 @@ class BenchmarkDJTrace(Benchmark):
                                  "{cputime}/1000",
                                  ylabel="time in ms",
                                  title="total runtime",
-                                 filepostfix="runtime")
+                                 file_postfix="runtime")
 
         # Function Times
         func_times_means = {allocator: {} for allocator in allocators}
@@ -262,12 +262,12 @@ class BenchmarkDJTrace(Benchmark):
                                  "{Max_RSS}/1000",
                                  ylabel="Max RSS in MB",
                                  title="Max RSS (VmHWM)",
-                                 filepostfix="newrss")
+                                 file_postfix="newrss")
 
         # self.barplot_fixed_arg("{Max_RSS}/1000",
         # ylabel='"Max RSS in MB"',
         # title='"Highwatermark of Vm (VmHWM)"',
-        # filepostfix="newrss")
+        # file_postfix="newrss")
 
         del allocators["Ideal_RSS"]
         del self.results["stats"]["Ideal_RSS"]
@@ -312,7 +312,7 @@ class BenchmarkDJTrace(Benchmark):
             "expression": "{Max_RSS}/1000",
             "sort": "<"
         }],
-                              filepostfix="table")
+                              file_postfix="table")
 
         # Tables
         for perm in self.iterate_args(args=args):
diff --git a/src/benchmarks/espresso.py b/src/benchmarks/espresso.py
index 5149dcb..6012d02 100644
--- a/src/benchmarks/espresso.py
+++ b/src/benchmarks/espresso.py
@@ -84,7 +84,7 @@ class BenchmarkEspresso(Benchmark):
                                "{task-clock}/1000",
                                ylabel="cpu-second",
                                title="Espresso: runtime",
-                               filepostfix="time")
+                               file_postfix="time")
 
         # L1 cache misses
         plt.barplot_single_arg(
@@ -92,7 +92,7 @@ class BenchmarkEspresso(Benchmark):
             "({L1-dcache-load-misses}/{L1-dcache-loads})*100",
             ylabel="L1 misses in %",
             title="Espresso l1 cache misses",
-            filepostfix="l1misses",
+            file_postfix="l1misses",
             yerr=False)
 
         # Memusage
@@ -100,7 +100,7 @@ class BenchmarkEspresso(Benchmark):
                                "{VmHWM}",
                                ylabel="VmHWM in KB",
                                title="Espresso VmHWM",
-                               filepostfix="vmhwm")
+                               file_postfix="vmhwm")
 
         plt.write_tex_table(self, [{
             "label": "Runtime [ms]",
@@ -111,7 +111,7 @@ class BenchmarkEspresso(Benchmark):
             "expression": "{VmHWM}",
             "sort": "<"
         }],
-                            filepostfix="table")
+                            file_postfix="table")
 
         plt.export_stats_to_dataref(self, "task-clock")
 
diff --git a/src/benchmarks/falsesharing.py b/src/benchmarks/falsesharing.py
index b9d0006..5c18367 100644
--- a/src/benchmarks/falsesharing.py
+++ b/src/benchmarks/falsesharing.py
@@ -83,7 +83,7 @@ class BenchmarkFalsesharing(Benchmark):
                            "{speedup}",
                            ylabel="Speedup",
                            title="Speedup: {arg} {arg_value}",
-                           filepostfix="speedup",
+                           file_postfix="speedup",
                            autoticks=False,
                            fixed=["bench"])
 
@@ -92,7 +92,7 @@ class BenchmarkFalsesharing(Benchmark):
             "{l1chache_misses}",
             ylabel="l1 cache misses in %",
             title="cache misses: {arg} {arg_value}",
-            filepostfix="l1-misses",
+            file_postfix="l1-misses",
             autoticks=False,
             fixed=["bench"])
 
@@ -100,7 +100,7 @@ class BenchmarkFalsesharing(Benchmark):
                            # "({LLC-load-misses}/{LLC-loads})*100",
                            # ylabel="llc cache misses in %",
                            # title="LLC misses: {arg} {arg_value}",
-                           # filepostfix="llc-misses",
+                           # file_postfix="llc-misses",
                            # autoticks=False,
                            # fixed=["bench"])
 
@@ -109,7 +109,7 @@ class BenchmarkFalsesharing(Benchmark):
             "expression": "{speedup}",
             "sort": ">"
         }],
-                            filepostfix="speedup.table")
+                            file_postfix="speedup.table")
 
         # plt.export_stats_to_csv(self, "speedup", "time")
         # plt.export_stats_to_csv(self, "l1chache_misses", "l1-misses")
diff --git a/src/benchmarks/fd.py b/src/benchmarks/fd.py
index 0d5657d..6f88878 100644
--- a/src/benchmarks/fd.py
+++ b/src/benchmarks/fd.py
@@ -71,7 +71,7 @@ class BenchmarkFd(Benchmark):
                                "{task-clock}",
                                ylabel="runtime in ms",
                                title="fd runtime",
-                               filepostfix="runtime")
+                               file_postfix="runtime")
 
         plt.export_stats_to_dataref(self, "task-clock")
 
@@ -79,7 +79,7 @@ class BenchmarkFd(Benchmark):
                                "{VmHWM}",
                                ylabel="VmHWM in KB",
                                title="fd memusage",
-                               filepostfix="memusage")
+                               file_postfix="memusage")
 
         plt.export_stats_to_dataref(self, "VmHWM")
 
diff --git a/src/benchmarks/httpd.py b/src/benchmarks/httpd.py
index 2138834..8332f27 100644
--- a/src/benchmarks/httpd.py
+++ b/src/benchmarks/httpd.py
@@ -65,7 +65,7 @@ class BenchmarkHTTPD(Benchmark):
                            xlabel="threads",
                            ylabel="requests/s",
                            autoticks=False,
-                           filepostfix="requests",
+                           file_postfix="requests",
                            title="{perm.site}: requests/s")
 
         plt.plot_fixed_arg(self,
@@ -73,7 +73,7 @@ class BenchmarkHTTPD(Benchmark):
                            xlabel="threads",
                            ylabel="VmHWM in KB",
                            title="{perm.site}: nginx memory usage",
-                           filepostfix="httpd_vmhwm",
+                           file_postfix="httpd_vmhwm",
                            autoticks=False)
 
         plt.plot_fixed_arg(self,
@@ -81,7 +81,7 @@ class BenchmarkHTTPD(Benchmark):
                            xlabel="threads",
                            ylabel="VmHWM in KB",
                            title="{perm.site}: php-fpm memory usage",
-                           filepostfix="php-fpm_vmhwm",
+                           file_postfix="php-fpm_vmhwm",
                            autoticks=False)
 
 
diff --git a/src/benchmarks/keydb.py b/src/benchmarks/keydb.py
index 3833a87..b8da614 100644
--- a/src/benchmarks/keydb.py
+++ b/src/benchmarks/keydb.py
@@ -113,11 +113,11 @@ class BenchmarkKeyDB(Benchmark):
         plt.plot_fixed_arg(self, "{totals_ops}",
                             ylabel="'OPS/second'",
                             title="KeyDB Operations: {perm}",
-                            filepostfix="total_ops")
+                            file_postfix="total_ops")
 
         plt.plot_fixed_arg(self, "{keydb_vmhwm}",
                             ylabel="'VmHWM [KB]'",
                             title="KeyDB Memusage: {perm}",
-                            filepostfix="vmhwm")
+                            file_postfix="vmhwm")
 
 keydb = BenchmarkKeyDB()
diff --git a/src/benchmarks/larson.py b/src/benchmarks/larson.py
index 5f153ea..d2e9286 100644
--- a/src/benchmarks/larson.py
+++ b/src/benchmarks/larson.py
@@ -85,14 +85,14 @@ class BenchmarkLarson(Benchmark):
                            "{throughput}/1000000",
                            ylabel="MOPS/s",
                            title="Larson: {arg} {arg_value}",
-                           filepostfix="throughput")
+                           file_postfix="throughput")
 
         plt.plot_fixed_arg(
             self,
             "({L1-dcache-load-misses}/{L1-dcache-loads})*100",
             ylabel="l1 cache misses in %",
             title="Larson cache misses: {arg} {arg_value}",
-            filepostfix="cachemisses")
+            file_postfix="cachemisses")
 
 
 larson = BenchmarkLarson()
diff --git a/src/benchmarks/lld.py b/src/benchmarks/lld.py
index a9fe26c..55b9063 100644
--- a/src/benchmarks/lld.py
+++ b/src/benchmarks/lld.py
@@ -294,7 +294,7 @@ class BenchmarkLld(Benchmark):
             "expression": "{task-clock}",
             "sort": "<"
         }],
-                                  filepostfix="table")
+                                  file_postfix="table")
 
 
 lld = BenchmarkLld()
diff --git a/src/benchmarks/loop.py b/src/benchmarks/loop.py
index 6d590c9..81d55cb 100644
--- a/src/benchmarks/loop.py
+++ b/src/benchmarks/loop.py
@@ -65,7 +65,7 @@ class BenchmarkLoop(Benchmark):
                            "{mops}",
                            ylabel="MOPS/cpu-second",
                            title="Loop: {arg} {arg_value}",
-                           filepostfix="time",
+                           file_postfix="time",
                            autoticks=False)
 
         # L1 cache misses
@@ -74,21 +74,21 @@ class BenchmarkLoop(Benchmark):
             "({L1-dcache-load-misses}/{L1-dcache-loads})*100",
             ylabel="L1 misses in %",
             title="Loop l1 cache misses: {arg} {arg_value}",
-            filepostfix="l1misses",
+            file_postfix="l1misses",
             autoticks=False)
 
         # Speed Matrix
         plt.write_best_doublearg_tex_table(
             self,
             "{mops}",
-            filepostfix="time.matrix")
+            file_postfix="time.matrix")
 
         plt.write_tex_table(self, [{
             "label": "MOPS/s",
             "expression": "{mops}",
             "sort": ">"
         }],
-                            filepostfix="mops.table")
+                            file_postfix="mops.table")
 
         # plt.export_stats_to_csv(self, "task-clock")
         # plt.export_stats_to_dataref(self, "task-clock")
diff --git a/src/benchmarks/mysql.py b/src/benchmarks/mysql.py
index 345a675..a5b215c 100644
--- a/src/benchmarks/mysql.py
+++ b/src/benchmarks/mysql.py
@@ -205,7 +205,7 @@ class BenchmarkMYSQL(Benchmark):
                              xlabel='"threads"',
                              ylabel='"transactions"',
                              title='"sysbench oltp read only"',
-                             filepostfix="l")
+                             file_postfix="l")
 
         # normalized linear plot
         ref_alloc = list(allocators)[0]
@@ -213,7 +213,7 @@ class BenchmarkMYSQL(Benchmark):
                              xlabel='"threads"',
                              ylabel='"transactions scaled at " + scale',
                              title='"sysbench oltp read only"',
-                             filepostfix="norm.l",
+                             file_postfix="norm.l",
                              scale=ref_alloc)
 
         # bar plot
@@ -221,14 +221,14 @@ class BenchmarkMYSQL(Benchmark):
                                 xlabel='"threads"',
                                 ylabel='"transactions"',
                                 title='"sysbench oltp read only"',
-                                filepostfix="b")
+                                file_postfix="b")
 
         # normalized bar plot
         plt.barplot_single_arg(self, "{transactions}",
                                 xlabel='"threads"',
                                 ylabel='"transactions scaled at " + scale',
                                 title='"sysbench oltp read only"',
-                                filepostfix="norm.b",
+                                file_postfix="norm.b",
                                 scale=ref_alloc)
 
         # Memusage
@@ -236,7 +236,7 @@ class BenchmarkMYSQL(Benchmark):
                                 xlabel='"threads"',
                                 ylabel='"VmHWM in kB"',
                                 title='"Memusage sysbench oltp read only"',
-                                filepostfix="mem")
+                                file_postfix="mem")
 
         plt.write_tex_table(self, [{
             "label": "Transactions",
@@ -247,7 +247,7 @@ class BenchmarkMYSQL(Benchmark):
             "expression": "{mysqld_vmhwm}",
             "sort": "<"
         }],
-                             filepostfix="table")
+                             file_postfix="table")
 
         # Colored latex table showing transactions count
         d = {allocator: {} for allocator in allocators}
diff --git a/src/benchmarks/raxmlng.py b/src/benchmarks/raxmlng.py
index 228c220..8c8b878 100644
--- a/src/benchmarks/raxmlng.py
+++ b/src/benchmarks/raxmlng.py
@@ -86,7 +86,7 @@ class BenchmarkRaxmlng(Benchmark):
                                "{runtime}",
                                ylabel='"runtime in s"',
                                title='"raxml-ng tree inference benchmark"',
-                               filepostfix="runtime")
+                               file_postfix="runtime")
 
         plt.export_stats_to_dataref(self, "runtime")
 
@@ -94,7 +94,7 @@ class BenchmarkRaxmlng(Benchmark):
                                "{VmHWM}",
                                ylabel='"VmHWM in KB"',
                                title='"raxml-ng memusage"',
-                               filepostfix="memusage")
+                               file_postfix="memusage")
 
         plt.export_stats_to_dataref(self, "VmHWM")
 
diff --git a/src/benchmarks/realloc.py b/src/benchmarks/realloc.py
index 251185e..bd8f801 100644
--- a/src/benchmarks/realloc.py
+++ b/src/benchmarks/realloc.py
@@ -37,7 +37,7 @@ class BenchmarkRealloc(Benchmark):
         plt.barplot_single_arg(self, "{task-clock}",
                                 ylabel='"task-clock in ms"',
                                 title='"realloc micro benchmark"',
-                                filepostfix="time")
+                                file_postfix="time")
 
         plt.export_stats_to_csv(self, "task-clock")
         plt.export_stats_to_dataref(self, "task-clock")
diff --git a/src/benchmarks/redis.py b/src/benchmarks/redis.py
index 71d9227..cfad489 100644
--- a/src/benchmarks/redis.py
+++ b/src/benchmarks/redis.py
@@ -84,12 +84,12 @@ class BenchmarkRedis(Benchmark):
         plt.barplot_single_arg(self, "{requests}",
                                 ylabel='"requests per s"',
                                 title='"redis throughput"',
-                                filepostfix="requests")
+                                file_postfix="requests")
 
         plt.barplot_single_arg(self, "{redis_vmhwm}",
                                 ylabel='"VmHWM in KB"',
                                 title='"redis memusage"',
-                                filepostfix="vmhwm")
+                                file_postfix="vmhwm")
 
         plt.export_stats_to_dataref(self, "requests")
 
diff --git a/src/benchmarks/t_test1.py b/src/benchmarks/t_test1.py
index 119f2a5..f0856f6 100644
--- a/src/benchmarks/t_test1.py
+++ b/src/benchmarks/t_test1.py
@@ -45,7 +45,7 @@ class BenchmarkTTest1(Benchmark):
         plt.plot_fixed_arg(self, yval,
                             ylabel='"Mops / CPU second"',
                             title='"T-Ttest1: " + arg + " " + str(arg_value)',
-                            filepostfix="time",
+                            file_postfix="time",
                             autoticks=False)
 
         # L1 cache misses
@@ -53,18 +53,18 @@ class BenchmarkTTest1(Benchmark):
             "({L1-dcache-load-misses}/{L1-dcache-loads})*100",
             ylabel='"L1 misses in %"',
             title='"T-Test1 l1 cache misses: " + arg + " " + str(arg_value)',
-            filepostfix="l1misses",
+            file_postfix="l1misses",
             autoticks=False)
 
         # Speed Matrix
-        plt.write_best_doublearg_tex_table(self, yval, filepostfix="mops.matrix")
+        plt.write_best_doublearg_tex_table(self, yval, file_postfix="mops.matrix")
 
         plt.write_tex_table(self, [{
             "label": "MOPS/s",
             "expression": yval,
             "sort": ">"
         }],
-                             filepostfix="mops.table")
+                             file_postfix="mops.table")
 
         plt.export_stats_to_csv(self, "task-clock")
 
diff --git a/src/plots.py b/src/plots.py
index c254741..94dcc4b 100644
--- a/src/plots.py
+++ b/src/plots.py
@@ -56,8 +56,32 @@ def _eval_with_stat(bench, evaluation, alloc, perm, stat):
         return nan
     return eval(res)
 
+def _get_y_data(bench, expression, allocator, perms, stat="mean", scale=None):
+    """Helper to get the y data of an allocator for given permutations"""
+    y_data = []
+    for perm in perms:
+        if scale:
+            if scale == allocator:
+                y_data.append(1)
+            else:
+                val = _eval_with_stat(bench, expression, allocator, perm, stat)
+                norm_val = _eval_with_stat(bench, expression, scale, perm, stat)
+                y_data.append(val / norm_val)
+        else:
+            y_data.append(_eval_with_stat(bench, expression, allocator, perm, stat))
+
+    return y_data
+
+def _save_figure(bench, fig, sumdir='', file_postfix='', file_ext=src.globalvars.summary_file_ext):
+    figname = os.path.join(sumdir, f"{bench.name}.{file_postfix}.{file_ext}")
+    if figname.endswith(".tex"):
+        import tikzplotlib
+        tikzplotlib.save(figname)
+    else:
+        fig.savefig(figname)
+
 def plot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
-                    autoticks=True, title="default title", filepostfix="",
+                    autoticks=True, title="default title", file_postfix="",
                     sumdir="", arg="", scale=None, file_ext=src.globalvars.summary_file_ext):
     """plot line graphs for each permutation of the benchmark's command arguments"""
 
@@ -66,24 +90,15 @@ def plot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
 
     arg = arg or list(args.keys())[0]
 
+    fig = plt.figure()
+
     if not autoticks:
         x_vals = list(range(1, len(args[arg]) + 1))
     else:
         x_vals = args[arg]
 
     for allocator in allocators:
-        y_vals = []
-        for perm in bench.iterate_args(args=args):
-            if scale:
-                if scale == allocator:
-                    y_vals = [1] * len(x_vals)
-                else:
-                    mean = _eval_with_stat(bench, yval, allocator, perm, "mean")
-                    norm_mean = _eval_with_stat(bench, yval, scale, perm, "mean")
-                    y_vals.append(mean / norm_mean)
-            else:
-                y_vals.append(_eval_with_stat(bench, yval, allocator, perm, "mean"))
-
+        y_vals = _get_y_data(bench, yval, allocator, bench.iterate_args(args=args), stat='mean', scale=scale)
         plt.plot(x_vals, y_vals, marker='.', linestyle='-',
                  label=allocator, color=_get_alloc_color(bench, allocator))
 
@@ -95,16 +110,14 @@ def plot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
     plt.xlabel(xlabel.format(**label_substitutions))
     plt.ylabel(ylabel.format(**label_substitutions))
     plt.title(title.format(**label_substitutions))
-    figname = os.path.join(sumdir, f"{bench.name}.{filepostfix}.{file_ext}")
-    if figname.endswith(".tex"):
-        import tikzplotlib
-        tikzplotlib.save(figname)
-    else:
-        plt.savefig(figname)
-    plt.clf()
+
+    _save_figure(bench, fig, sumdir, file_postfix, file_ext)
+    fig.close()
+
+    return fig
 
 def barplot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
-                       title="default title", filepostfix="", sumdir="",
+                       title="default title", file_postfix="", sumdir="",
                        arg="", scale=None, file_ext=src.globalvars.summary_file_ext, yerr=True):
     """plot bar plots for each permutation of the benchmark's command arguments"""
 
@@ -121,26 +134,14 @@ def barplot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
 
     narg = len(arg)
 
+    fig = plt.figure()
+
     for i, allocator in enumerate(allocators):
         x_vals = list(range(i, narg * (nallocators+1), nallocators+1))
-        y_vals = []
+        y_vals = _get_y_data(bench, yval, allocator, bench.iterate_args(args=args), stat='mean', scale=scale)
         y_errs = None
         if yerr:
-            y_errs = []
-
-        for perm in bench.iterate_args(args=args):
-            if scale:
-                if scale == allocator:
-                    y_vals = [1] * len(x_vals)
-                else:
-                    mean = _eval_with_stat(bench, yval, allocator, perm, "mean")
-                    norm_mean = _eval_with_stat(bench, yval, scale, perm, "mean")
-                    y_vals.append(mean / norm_mean)
-            else:
-                y_vals.append(_eval_with_stat(bench, yval, allocator, perm, "mean"))
-
-            if yerr:
-                y_errs.append(_eval_with_stat(bench, yval, allocator, perm, "std"))
+            y_vals = _get_y_data(bench, yval, allocator, bench.iterate_args(args=args), stat='std')
 
         plt.bar(x_vals, y_vals, width=1, label=allocator, yerr=y_errs,
                 color=_get_alloc_color(bench, allocator))
@@ -153,16 +154,12 @@ def barplot_single_arg(bench, yval, ylabel="y-label", xlabel="x-label",
     plt.xlabel(xlabel.format(**label_substitutions))
     plt.ylabel(ylabel.format(**label_substitutions))
     plt.title(title.format(**label_substitutions))
-    figname = os.path.join(sumdir, f"{bench.name}.{filepostfix}.{file_ext}")
-    if figname.endswith(".tex"):
-        import tikzplotlib
-        tikzplotlib.save(figname)
-    else:
-        plt.savefig(figname)
-    plt.clf()
+
+    _save_figure(bench, fig, sumdir, file_postfix, file_ext)
+    fig.close()
 
 def plot_fixed_arg(bench, yval, ylabel="y-label", xlabel="{loose_arg}",
-                   autoticks=True, title="default title", filepostfix="",
+                   autoticks=True, title="default title", file_postfix="",
                    sumdir="", fixed=None, file_ext=src.globalvars.summary_file_ext, scale=None):
 
     args = bench.results["args"]
@@ -177,18 +174,10 @@ def plot_fixed_arg(bench, yval, ylabel="y-label", xlabel="{loose_arg}",
             x_vals = args[loose_arg]
 
         for arg_value in args[arg]:
+            fig = plt.figure()
+
             for allocator in allocators:
-                y_vals = []
-                for perm in bench.iterate_args_fixed({arg: arg_value}, args=args):
-                    if scale:
-                        if scale == allocator:
-                            y_vals = [1] * len(x_vals)
-                        else:
-                            mean = _eval_with_stat(bench, yval, allocator, perm, "mean")
-                            norm_mean = _eval_with_stat(bench, yval, scale, perm, "mean")
-                            y_vals.append(mean / norm_mean)
-                    else:
-                        y_vals.append(_eval_with_stat(bench, yval, allocator, perm, "mean"))
+                y_vals = _get_y_data(bench, yval, allocator, bench.iterate_args_fixed({arg: arg_value}, args=args), stat='mean', scale=scale)
 
                 plt.plot(x_vals, y_vals, marker='.', linestyle='-',
                          label=allocator, color=_get_alloc_color(bench, allocator))
@@ -202,14 +191,9 @@ def plot_fixed_arg(bench, yval, ylabel="y-label", xlabel="{loose_arg}",
             plt.xlabel(xlabel.format(**label_substitutions))
             plt.ylabel(ylabel.format(**label_substitutions))
             plt.title(title.format(**label_substitutions))
-            figname = os.path.join(sumdir,
-                                   f"{bench.name}.{arg}.{arg_value}.{filepostfix}.{file_ext}")
-            if figname.endswith(".tex"):
-                import tikzplotlib
-                tikzplotlib.save(figname)
-            else:
-                plt.savefig(figname)
-            plt.clf()
+
+            _save_figure(bench, fig, sumdir, file_postfix, file_ext)
+            fig.close()
 
 def export_facts_to_file(bench, comment_symbol, output_file):
     """Write collected facts about used system and benchmark to file"""
@@ -304,8 +288,7 @@ def export_stats_to_dataref(bench, datapoint, path=None):
                     cur_line.replace("_", "-")
                     print(cur_line, file=dataref_file)
 
-def write_best_doublearg_tex_table(bench, expr, sort=">",
-                                   filepostfix="", sumdir=""):
+def write_best_doublearg_tex_table(bench, expr, sort=">", file_postfix="", sumdir=""):
     args = bench.results["args"]
     keys = list(args.keys())
     allocators = bench.results["allocators"]
@@ -358,7 +341,7 @@ def write_best_doublearg_tex_table(bench, expr, sort=">",
         print("\\end{tabular}", file=tex_file)
         print("\\end{document}", file=tex_file)
 
-def write_tex_table(bench, entries, filepostfix="", sumdir=""):
+def write_tex_table(bench, entries, file_postfix="", sumdir=""):
     """generate a latex standalone table from an list of entries dictionaries
 
     Entries must have at least the two keys: "label" and "expression".
@@ -396,7 +379,7 @@ def write_tex_table(bench, entries, filepostfix="", sumdir=""):
     entry_header_line = perm_fields_header + entry_header_line * nallocators
     entry_header_line = entry_header_line[:-1] + "\\\\"
 
-    fname = os.path.join(sumdir, ".".join([bench.name, filepostfix, "tex"]))
+    fname = os.path.join(sumdir, ".".join([bench.name, file_postfix, "tex"]))
     with open(fname, "w") as tex_file:
         print("\\documentclass{standalone}", file=tex_file)
         print("\\usepackage{booktabs}", file=tex_file)