Generalize table row highlighting for the ovm-build benchmark.

Also, start merging TSVs that record file sizes.
oilshell · Feb 17, 2018 · bc2de38b1b71e25d85524748edbd07b97e78b574 · bc2de38
1 parent b2fe31d
commit bc2de38b1b71e25d85524748edbd07b97e78b574
Unified Split

Showing with 67 additions and 32 deletions.

+7 −2 benchmarks/common.sh

+7 −7 benchmarks/osh-parser.sh

+24 −8 benchmarks/ovm-build.sh

+6 −5 benchmarks/report.R

+1 −1 benchmarks/report.sh

+1 −1 web/benchmarks.css

+21 −8 web/table/csv2html.py
diff --git a/benchmarks/common.sh b/benchmarks/common.sh
@@ -18,9 +18,14 @@ csv-concat() { tools/csv_concat.py "$@"; }
 # TSV and CSV concatenation are actually the same.  Just use the same script.
 tsv-concat() { tools/csv_concat.py "$@"; }

-csv2html() { web/table/csv2html.py "$@"; }
+# For compatibility, if cell starts with 'osh', apply the 'special' CSS class.
+csv2html() {
+  web/table/csv2html.py --css-class-pattern 'special ^osh' "$@";
+}

-tsv2html() { web/table/csv2html.py --tsv "$@"; }
+tsv2html() {
+  web/table/csv2html.py --tsv "$@";
+}

 # Need an absolute path here.
 readonly _time_tool=$PWD/benchmarks/time.py 
diff --git a/benchmarks/osh-parser.sh b/benchmarks/osh-parser.sh
@@ -219,39 +219,39 @@ print-report() {

    <h3>Parse Time Summary</h3>
 EOF
-  web/table/csv2html.py $in_dir/summary.csv
+  csv2html $in_dir/summary.csv

  cat <<EOF
    <h3>Memory Used to Parse</h3>

    <p>Running under <code>osh-ovm</code>.  Memory usage is measured in MB
    (powers of 10), not MiB (powers of 2).</p>
 EOF
-  web/table/csv2html.py $in_dir/virtual-memory.csv
+  csv2html $in_dir/virtual-memory.csv

  cat <<EOF

    <h3>Shell and Host Details</h3>
 EOF
-  web/table/csv2html.py $in_dir/shells.csv
-  web/table/csv2html.py $in_dir/hosts.csv
+  csv2html $in_dir/shells.csv
+  csv2html $in_dir/hosts.csv

 cat <<EOF
    <h3>Raw Data</h3>
 EOF
-  web/table/csv2html.py $in_dir/raw-data.csv
+  csv2html $in_dir/raw-data.csv

 cat <<EOF
    <h3>Parse Time Breakdown by File</h3>

    <h4>Elasped Time in milliseconds</h4>
 EOF
-  web/table/csv2html.py $in_dir/elapsed.csv
+  csv2html $in_dir/elapsed.csv
  cat <<EOF

    <h4>Parsing Rate in lines/millisecond</h4>
 EOF
-  web/table/csv2html.py $in_dir/rate.csv
+  csv2html $in_dir/rate.csv
  cat <<EOF
  </body>
 </html>
diff --git a/benchmarks/ovm-build.sh b/benchmarks/ovm-build.sh
@@ -88,8 +88,10 @@ extract() {
 # - zipfile.py to look inside bytecode.zip

 sizes-tsv() {
-  echo $'num_bytes\tpath'
-  find "$@" -maxdepth 0 -printf $'%s\t%p\n'
+  # host_label matches the times.tsv file output by report.R
+  echo $'host_label\tnum_bytes\tpath'
+  local host=$(hostname)
+  find "$@" -maxdepth 0 -printf "$host\t%s\t%p\n"
 }

 # NOTE: This should be the same on all x64 machines.  But I want to run it on
@@ -292,11 +294,24 @@ stage1() {
  local out=$BASE_DIR/stage1
  mkdir -p $out

-  local times_tsv=$out/times.tsv
+  local x
+  local -a a b
+
  # Globs are in lexicographical order, which works for our dates.
-  local -a a=($raw_dir/flanders.*.times.tsv)
-  local -a b=($raw_dir/lisa.*.times.tsv)
-  tsv-concat ${a[-1]} ${b[-1]} > $times_tsv
+  x=$out/times.tsv
+  a=($raw_dir/flanders.*.times.tsv)
+  b=($raw_dir/lisa.*.times.tsv)
+  tsv-concat ${a[-1]} ${b[-1]} > $x
+
+  x=$out/bytecode-size.tsv
+  a=($raw_dir/flanders.*.bytecode-size.tsv)
+  b=($raw_dir/lisa.*.bytecode-size.tsv)
+  tsv-concat ${a[-1]} ${b[-1]} > $x
+
+  x=$out/bin-sizes.tsv
+  a=($raw_dir/flanders.*.bin-sizes.tsv)
+  b=($raw_dir/lisa.*.bin-sizes.tsv)
+  tsv-concat ${a[-1]} ${b[-1]} > $x

  # Construct a one-column TSV file
  local raw_data_tsv=$out/raw-data.tsv
@@ -335,7 +350,7 @@ print-report() {
    for comparison.
    </p>
 EOF
-  tsv2html $in_dir/times.tsv
+  tsv2html --css-class-pattern 'special ^oil' $in_dir/times.tsv

  cat <<EOF
    <h3>Binary Size</h3>
@@ -348,7 +363,8 @@ EOF
    </p>

 EOF
-  #csv2html $in_dir/virtual-memory.csv
+  tsv2html $in_dir/bytecode-size.tsv
+  tsv2html $in_dir/bin-sizes.tsv

  cat <<EOF

diff --git a/benchmarks/report.R b/benchmarks/report.R
@@ -378,18 +378,15 @@ WriteOvmBuildDetails = function(distinct_hosts, distinct_compilers, out_dir) {

 OvmBuildReport = function(in_dir, out_dir) {
  times = readTsv(file.path(in_dir, 'times.tsv'))
+  bytecode_size = readTsv(file.path(in_dir, 'bytecode-size.tsv'))
+  bin_sizes = readTsv(file.path(in_dir, 'bin-sizes.tsv'))
  raw_data = readTsv(file.path(in_dir, 'raw-data.tsv'))

  times %>% filter(status != 0) -> failed
  if (nrow(failed) != 0) {
    print(failed)
    stop('Some ovm-build tasks failed')
  }
-  #print(vm)
-
-  #times %>%
-  #  arrange(host, VmPeak_MB) ->
-  #  vm

  times %>% distinct(host_name, host_hash) -> distinct_hosts
  distinct_hosts$host_label = distinct_hosts$host_name
@@ -414,6 +411,10 @@ OvmBuildReport = function(in_dir, out_dir) {

  #print(times)

+  # NOTE: These don't have the host and compiler.
+  writeTsv(bytecode_size, file.path(out_dir, 'bytecode-size'))
+  writeTsv(bin_sizes, file.path(out_dir, 'bin-sizes'))
+
  writeTsv(times, file.path(out_dir, 'times'))

  # TODO: I want a size report too
diff --git a/benchmarks/report.sh b/benchmarks/report.sh
@@ -94,7 +94,7 @@ all() {
 # For view
 dev-index() {
  local out=_tmp/benchmarks.html
-  for name in osh-parser osh-runtime vm-baseline oheap; do
+  for name in osh-parser osh-runtime vm-baseline ovm-build oheap; do
    echo "<a href=\"$name/index.html\">$name</a> <br/>"
  done > $out
  log "Wrote $out"
diff --git a/web/benchmarks.css b/web/benchmarks.css
@@ -30,6 +30,6 @@ h3, h4 {
  background-color: oldlace;
 }
 /* rows */
-.osh-row {
+.special {
  background-color: oldlace;
 }
diff --git a/web/table/csv2html.py b/web/table/csv2html.py
@@ -39,6 +39,7 @@
 import csv
 import optparse
 import os
+import re
 import sys


@@ -224,7 +225,14 @@ def PrintColGroup(col_names, schema):
  print '  </colgroup>'


-def PrintTable(css_id, schema, col_names, rows):
+def PrintTable(css_id, schema, col_names, rows, css_class_pattern):
+  if css_class_pattern:
+    css_class, r = css_class_pattern.split(None, 2)
+    cell_regex = re.compile(r)
+  else:
+    css_class = None
+    cell_regex = None
+
  print '<table id="%s">' % css_id
  print '  <thead>'
  print '    <tr>'
@@ -244,12 +252,12 @@ def PrintTable(css_id, schema, col_names, rows):

    # TODO: There should be a special column called CSS_CLASS.  Output that
    # from R.
-    for cell in row:
-      if cell.startswith('osh'):
-        row_class = 'class="osh-row"'
-        break
-    else:
-      row_class = ''
+    row_class = ''
+    if cell_regex:
+      for cell in row:
+        if cell_regex.match(cell):
+          row_class = 'class="%s"' % css_class
+          break

    print '    <tr {}>'.format(row_class)

@@ -292,6 +300,11 @@ def CreateOptionsParser():
  p.add_option(
      '--tsv', dest='tsv', default=False, action='store_true',
      help='Read input in TSV format')
+  p.add_option(
+      '--css-class-pattern', dest='css_class_pattern', type='str',
+      help='A string of the form CSS_CLASS:PATTERN.  If the cell contents '
+           'matches the pattern, then apply the given CSS class. '
+           'Example: osh:^osh')
  return p


@@ -344,7 +357,7 @@ def main(argv):

  filename = os.path.basename(csv_path)
  css_id, _ = os.path.splitext(filename)
-  PrintTable(css_id, schema, col_names, rows)
+  PrintTable(css_id, schema, col_names, rows, opts.css_class_pattern)


 if __name__ == '__main__':