Update gemm performance monitoring tool:

 - permit to recompute a subset of changesets
 - update changeset list
 - add a few more cases

3 files changed, 78 insertions, 40 deletions

diff --git a/bench/perf_monitoring/gemm/changesets.txt b/bench/perf_monitoring/gemm/changesets.txt
index f19b4287d..b379d7bd2 100644
--- a/bench/perf_monitoring/gemm/changesets.txt
+++ b/bench/perf_monitoring/gemm/changesets.txt
@@ -1,39 +1,42 @@
-3.0.1
-3.1.1
-3.2.0
+#3.0.1
+#3.1.1
+#3.2.0
 3.2.4
-5745:37f59e65eb6c
-5891:d8652709345d
-5893:24b4dc92c6d3
-5895:997c2ef9fc8b
-5904:e1eafd14eaa1
-5908:f8ee3c721251
-5921:ca808bb456b0
-5927:8b1001f9e3ac
-5937:5a4ca1ad8c53
-5949:f3488f4e45b2
-5969:e09031dccfd9
-5992:4a429f5e0483
+#5745:37f59e65eb6c
+5891:d8652709345d  # introduce AVX
+#5893:24b4dc92c6d3  # merge
+5895:997c2ef9fc8b  # introduce FMA
+#5904:e1eafd14eaa1  # complex and AVX
+5908:f8ee3c721251  # improve packing with ptranspose
+#5921:ca808bb456b0  # merge
+#5927:8b1001f9e3ac
+5937:5a4ca1ad8c53  # New gebp kernel handling up to 3 packets x 4 register-level blocks
+#5949:f3488f4e45b2  # merge
+#5969:e09031dccfd9  # Disable 3pX4 kernel on Altivec
+#5992:4a429f5e0483  # merge
 before-evaluators
-6334:f6a45e5b8b7c
-6639:c9121c60b5c7
-6655:06f163b5221f
-6677:700e023044e7   # FMA has been wrongly disabled
-6681:11d31dafb0e3
-6699:5e6e8e10aad1   # merge default to tensors
-6726:ff2d2388e7b9   # merge default to tensors
-6742:0cbd6195e829   # merge default to tensors
-6747:853d2bafeb8f   # Generalized the gebp apis
+#6334:f6a45e5b8b7c  # Implement evaluator for sparse outer products
+#6639:c9121c60b5c7
+#6655:06f163b5221f  # Properly detect FMA support on ARM
+#6677:700e023044e7   # FMA has been wrongly disabled
+#6681:11d31dafb0e3
+#6699:5e6e8e10aad1   # merge default to tensors
+#6726:ff2d2388e7b9   # merge default to tensors
+#6742:0cbd6195e829   # merge default to tensors
+#6747:853d2bafeb8f   # Generalized the gebp apis
 6765:71584fd55762   # Made the blocking computation aware of the l3 cache; Also optimized the blocking parameters to take into account the number of threads used for a computation
-6781:9cc5a931b2c6   # generalized gemv
-6792:f6e1daab600a   # ensured that contractions that can be reduced to a matrix vector product
-6844:039efd86b75c   # merge tensor
+#6781:9cc5a931b2c6   # generalized gemv
+#6792:f6e1daab600a   # ensured that contractions that can be reduced to a matrix vector product
+#6844:039efd86b75c   # merge tensor
 6845:7333ed40c6ef   # change prefetching in gebp
-6856:b5be5e10eb7f   # merge index conversion
-6893:c3a64aba7c70   # clean blocking size computation
-6898:6fb31ebe6492   # rotating kernel for ARM
+#6856:b5be5e10eb7f   # merge index conversion
+#6893:c3a64aba7c70   # clean blocking size computation
+#6898:6fb31ebe6492   # rotating kernel for ARM
 6899:877facace746   # rotating kernel for ARM only
-6904:c250623ae9fa   # result_of
+#6904:c250623ae9fa   # result_of
 6921:915f1b1fc158   # fix prefetching change for ARM
 6923:9ff25f6dacc6   # prefetching
-6933:52572e60b5d3   # blocking size strategy
\ No newline at end of file
+6933:52572e60b5d3   # blocking size strategy
+6937:c8c042f286b2   # avoid redundant pack_rhs
+6981:7e5d6f78da59   # dynamic loop swapping
+6984:45f26866c091   # rm dynamic loop swapping, adjust lhs's micro panel height to fully exploit L1 cache
diff --git a/bench/perf_monitoring/gemm/run_gemm.sh b/bench/perf_monitoring/gemm/run_gemm.sh
index d3a9fadc9..3fa6a3661 100755
--- a/bench/perf_monitoring/gemm/run_gemm.sh
+++ b/bench/perf_monitoring/gemm/run_gemm.sh
@@ -6,6 +6,7 @@
 
 # Options:
 #   -up : enforce the recomputation of existing data, and keep best results as a merging strategy
+#   -s  : recompute selected changesets only and keep bests
 
 
 if echo "$*" | grep '\-up' > /dev/null; then
@@ -14,14 +15,30 @@ else
   update=false
 fi
 
-if [ $update == true ]; then
+if echo "$*" | grep '\-s' > /dev/null; then
+  selected=true
+else
+  selected=false
+fi
+
+global_args="$*"
+
+if [ $selected == true ]; then
+ echo "Recompute selected changesets only and keep bests"
+elif [ $update == true ]; then
  echo "(Re-)Compute all changesets and keep bests"
 else
  echo "Skip previously computed changesets"
 fi
 
+
+
 if [ ! -d "eigen_src" ]; then
   hg clone https://bitbucket.org/eigen/eigen eigen_src
+else
+  cd eigen_src
+  hg pull -u
+  cd ..
 fi
 
 if [ ! -z '$CXX' ]; then
@@ -61,17 +78,31 @@ function test_current
   scalar=$2
   name=$3
   
-  prev=`grep $rev "$name.backup" | cut -c 14-`
+  prev=""
+  if [ -e "$name.backup" ]; then
+    prev=`grep $rev "$name.backup" | cut -c 14-`
+  fi
   res=$prev
   count_rev=`echo $prev |  wc -w`
   count_ref=`cat "settings.txt" |  wc -l`
-  if [ $update == true ] || [ $count_rev != $count_ref ]; then
+  if echo "$global_args" | grep "$rev" > /dev/null; then
+    rev_found=true
+  else
+    rev_found=false
+  fi
+#  echo $update et $selected et $rev_found because $rev et "$global_args"
+#  echo $count_rev et $count_ref
+  if [ $update == true ] || [ $count_rev != $count_ref ] || ([ $selected == true ] &&  [ $rev_found == true ]); then
     if $CXX -O2 -DNDEBUG -march=native $CXX_FLAGS -I eigen_src gemm.cpp -DSCALAR=$scalar -o $name; then
       curr=`./$name`
-      echo merge $prev
-      echo with  $curr
+      if [ $count_rev == $count_ref ]; then
+        echo "merge previous $prev"
+        echo "with new       $curr"
+      else
+        echo "got            $curr"
+      fi
       res=`merge "$curr" "$prev"`
-      echo $res
+#       echo $res
       echo "$rev $res" >> $name.out
     else
       echo "Compilation failed, skip rev $rev"
@@ -86,12 +117,12 @@ make_backup $PREFIX"sgemm"
 make_backup $PREFIX"dgemm"
 make_backup $PREFIX"cgemm"
 
-cut -f1 -d"#" < changesets.txt | while read rev
+cut -f1 -d"#" < changesets.txt | grep -E '[[:alnum:]]' | while read rev
 do
   if [ ! -z '$rev' ]; then
     echo "Testing rev $rev"
     cd eigen_src
-    hg up -C $rev
+    hg up -C $rev > /dev/null
     actual_rev=`hg identify | cut -f1 -d' '`
     cd ..
     
diff --git a/bench/perf_monitoring/gemm/settings.txt b/bench/perf_monitoring/gemm/settings.txt
index 6ef690708..5c43e1c7d 100644
--- a/bench/perf_monitoring/gemm/settings.txt
+++ b/bench/perf_monitoring/gemm/settings.txt
@@ -1,5 +1,6 @@
 8 8 8
 9 9 9
+24 24 24
 239 239 239
 240 240 240
 2400 24 24
@@ -8,4 +9,7 @@
 24 2400 2400
 2400 24 2400
 2400 2400 24
+2400 2400 64
+4800 23 160
+23 4800 160
 2400 2400 2400