onsuccess: #157: 1: [TCWG CI] https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/157/

Results :
  | # reset_artifacts:
  | -10
  | # build_abe binutils:
  | -9
  | # build_abe stage1 -- --set gcc_override_configure=--with-mode=arm --set gcc_override_configure=--disable-libsanitizer:
  | -8
  | # build_abe linux:
  | -7
  | # build_abe glibc:
  | -6
  | # build_abe stage2 -- --set gcc_override_configure=--with-mode=arm --set gcc_override_configure=--disable-libsanitizer:
  | -5
  | # benchmark -- -O3_marm:
  | 1

check_regression status : 0

5 files changed, 431 insertions, 387 deletions

diff --git a/notify/any.skipped b/notify/any.skipped
index ce88d735..7977ccd9 100644
--- a/notify/any.skipped
+++ b/notify/any.skipped
@@ -1,4 +1,4 @@
-437.leslie3d,leslie3d_base.default,sped up by 17% - 437.leslie3d,sped up by 17% - 437.leslie3d - from 14368 to 11868 perf samples

-437.leslie3d,[.] fluxk_,sped up by 22% - 437.leslie3d:[.] fluxk_,sped up by 22% - 437.leslie3d:[.] fluxk_ - from 3127 to 2449 perf samples

-437.leslie3d,[.] fluxj_,sped up by 29% - 437.leslie3d:[.] fluxj_,sped up by 29% - 437.leslie3d:[.] fluxj_ - from 3049 to 2174 perf samples

-437.leslie3d,[.] fluxi_,sped up by 18% - 437.leslie3d:[.] fluxi_,sped up by 18% - 437.leslie3d:[.] fluxi_ - from 2491 to 2039 perf samples

+437.leslie3d,leslie3d_base.default,slowed down by 21% - 437.leslie3d,slowed down by 21% - 437.leslie3d - from 11868 to 14358 perf samples

+437.leslie3d,[.] fluxk_,slowed down by 27% - 437.leslie3d:[.] fluxk_,slowed down by 27% - 437.leslie3d:[.] fluxk_ - from 2449 to 3108 perf samples

+437.leslie3d,[.] fluxj_,slowed down by 40% - 437.leslie3d:[.] fluxj_,slowed down by 40% - 437.leslie3d:[.] fluxj_ - from 2174 to 3051 perf samples

+437.leslie3d,[.] fluxi_,slowed down by 23% - 437.leslie3d:[.] fluxi_,slowed down by 23% - 437.leslie3d:[.] fluxi_ - from 2039 to 2502 perf samples

diff --git a/notify/jira/comment-template.txt b/notify/jira/comment-template.txt
index f3147798..1f821dc6 100644
--- a/notify/jira/comment-template.txt
+++ b/notify/jira/comment-template.txt
@@ -1,3 +1,3 @@
 [GNU-689]
 No change
-Details: https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/156/artifact/artifacts/notify/mail-body.txt/*view*/
+Details: https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/157/artifact/artifacts/notify/mail-body.txt/*view*/
diff --git a/notify/mail-body.txt b/notify/mail-body.txt
index fb944dac..1162f3ee 100644
--- a/notify/mail-body.txt
+++ b/notify/mail-body.txt
@@ -2,25 +2,25 @@ Dear contributor, our automatic CI has detected problems related to your patch(e
 
 In CI config tcwg_bmk-code_speed-spec2k6/gnu-arm-master-O3 after:
 
-  | 105 commits in binutils,gcc,glibc
-  | f514e6e4806 Automatic date update in version.in
-  | a833790a626 [gdb/symtab] Work around gas PR28629
-  | 98ef1d81d6a [gdb/symtab] Add producer_is_gas
-  | 87e3cc466e8 Implement DAP setVariable request
-  | 8170af78e17 Support Intel USER_MSR
+  | 101 commits in binutils,gcc,glibc
+  | de2efa143e3 Automatic date update in version.in
+  | 8489362f002 Remove unused declaration
+  | 8cb2c7fc2ea gdbsupport: mark array_view::slice with [[nodiscard]]
+  | dbbfabb4415 gdbsupport: record and print failed selftest names
+  | 88bfe6ac8bc gas: correct ignoring of C-style number suffixes
   | ... and 17 more commits in binutils
-  | 7480dbb6e78 i386: Improve stack protector patterns and peephole2s
-  | 9693459e030 PR modula2/102989: reimplement overflow detection in ztype though WIDE_INT_MAX_PRECISION
-  | 084ea7ea5aa RISC-V: Support vundefine intrinsics for tuple types
-  | c9bb20f7c90 NFC: Fix whitespace
-  | eac0917bd3d Daily bump.
-  | ... and 62 more commits in gcc
-  | 9db31d7456 Use correct subdir when building tst-rfc3484* for mach and arm
-  | bc888a3976 stdlib: Add more qsort{_r} coverage
-  | 03bf8357e8 stdlib: Remove use of mergesort on qsort (BZ 21719)
-  | 274a46c9b2 stdlib: Implement introsort for qsort (BZ 19305)
-  | d097f3c79b stdlib: qsort: Move some macros to inline function
-  | ... and 11 more commits in glibc
+  | ada871cfadd libstdc++: Remove redundant partial specialization in _Nth_type
+  | 51f94778b45 libstdc++: Use strerror_r in std::generic_category()::message(int) [PR110133]
+  | 6933c05ade7 libstdc++: Replace "_N" in examples of naming conventions
+  | 8067caa85d0 openmp: Add omp::decl support for C2X
+  | 40b9af020fc openmp: Add support for omp::directive and omp::sequence attributes in C2X
+  | ... and 64 more commits in gcc
+  | f957f47df7 sysdeps: sem_open: Clear O_CREAT when semaphore file is expected to exist [BZ #30789]
+  | ac79930498 Add SEGV_CPERR from Linux 6.6 to bits/siginfo-consts.h
+  | d3476c20b2 linux: Sync Linux 6.6 elf.h
+  | 9b3cb0277e linux: Add HWCAP2_HBC from Linux 6.6 to AArch64 bits/hwcap.h
+  | 10b4c8b96f linux: Add FSCONFIG_CMD_CREATE_EXCL from Linux 6.6 to sys/mount.h
+  | ... and 5 more commits in glibc
 
 No change
 
@@ -40,6 +40,6 @@ This benchmarking CI is work-in-progress, and we welcome feedback and suggestion
 -----------------8<--------------------------8<--------------------------8<--------------------------
 The information below can be used to reproduce a debug environment:
 
-Current build   : https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/156/artifact/artifacts
-Reference build : https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/155/artifact/artifacts
+Current build   : https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/157/artifact/artifacts
+Reference build : https://ci.linaro.org/job/tcwg_bmk-code_speed-spec2k6--gnu-arm-master-O3-build/156/artifact/artifacts
 
diff --git a/notify/mail-subject.txt b/notify/mail-subject.txt
index 4073375d..8ccac36e 100644
--- a/notify/mail-subject.txt
+++ b/notify/mail-subject.txt
@@ -1 +1 @@
-[Linaro-TCWG-CI] 105 commits in binutils,gcc,glibc: No change on arm O3
+[Linaro-TCWG-CI] 101 commits in binutils,gcc,glibc: No change on arm O3
diff --git a/notify/output-bmk-results.log b/notify/output-bmk-results.log
index da25d7fe..a4236d6f 100644
--- a/notify/output-bmk-results.log
+++ b/notify/output-bmk-results.log
@@ -120,7 +120,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 400.perlbench,perlbench_base.default : sample=0% (threshold=3%)
+DEBUG: checking exe.regression : 400.perlbench,perlbench_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -174,7 +174,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 403.gcc,gcc_base.default : sample=0% (threshold=3%)
+DEBUG: checking exe.regression : 403.gcc,gcc_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -228,7 +228,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 416.gamess,gamess_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.regression : 416.gamess,gamess_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -390,12 +390,42 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 437.leslie3d,leslie3d_base.default : sample=17% (threshold=10.290000000000001%)
+DEBUG: checking exe.regression : 437.leslie3d,leslie3d_base.default : sample=-21% (threshold=10.32%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(176):         return (result - 100 > threshold)
-output-bmk-results.py(225):             continue
+output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
+ --- modulename: output-bmk-results, funcname: get_short_long_diag
+output-bmk-results.py(113):     bmk = row["benchmark"]
+output-bmk-results.py(114):     rel_value = row["rel_" + metric]
+output-bmk-results.py(115):     prev_value = row[metric + "_x"]
+output-bmk-results.py(116):     curr_value = row[metric + "_y"]
+output-bmk-results.py(118):     if metric == "sample":
+output-bmk-results.py(119):         if curr_value == 999999999:
+output-bmk-results.py(122):         elif curr_value == 888888888:
+output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
+output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
+output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
+output-bmk-results.py(144):     suffix = ""
+output-bmk-results.py(145):     if metric == "sample":
+output-bmk-results.py(146):         prefix_regression = "slowed down by"
+output-bmk-results.py(147):         prefix_improvement = "sped up by"
+output-bmk-results.py(148):         suffix = "perf samples"
+output-bmk-results.py(159):     if sym_type=="symbol":
+output-bmk-results.py(162):       item=bmk
+output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
+output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
+output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
+output-bmk-results.py(231):         if metric == "sample" \
+output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
+output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
+output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
+output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
+ --- modulename: output-bmk-results, funcname: write_csv
+output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
+output-bmk-results.py(43):     self.csvwriter.writerow(arr)
+output-bmk-results.py(236):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -444,7 +474,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 445.gobmk,gobmk_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.regression : 445.gobmk,gobmk_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -471,7 +501,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 447.dealII,dealII_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.regression : 447.dealII,dealII_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -525,7 +555,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 453.povray,povray_base.default : sample=-2% (threshold=3%)
+DEBUG: checking exe.regression : 453.povray,povray_base.default : sample=-1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -579,7 +609,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 456.hmmer,hmmer_base.default : sample=2% (threshold=3.96%)
+DEBUG: checking exe.regression : 456.hmmer,hmmer_base.default : sample=-1% (threshold=3.96%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -606,7 +636,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 458.sjeng,sjeng_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.regression : 458.sjeng,sjeng_base.default : sample=-1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -633,7 +663,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 459.GemsFDTD,GemsFDTD_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.regression : 459.GemsFDTD,GemsFDTD_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -687,7 +717,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 464.h264ref,h264ref_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.regression : 464.h264ref,h264ref_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -714,7 +744,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 465.tonto,tonto_base.default : sample=2% (threshold=3%)
+DEBUG: checking exe.regression : 465.tonto,tonto_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -768,7 +798,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 471.omnetpp,omnetpp_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.regression : 471.omnetpp,omnetpp_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -822,7 +852,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 481.wrf,wrf_base.default : sample=-2% (threshold=3%)
+DEBUG: checking exe.regression : 481.wrf,wrf_base.default : sample=2% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -849,7 +879,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.regression : 482.sphinx3,sphinx_livepretend_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.regression : 482.sphinx3,sphinx_livepretend_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -921,7 +951,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 400.perlbench,perlbench_base.default : sample=0% (threshold=3%)
+DEBUG: checking exe.improvement : 400.perlbench,perlbench_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -975,7 +1005,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 403.gcc,gcc_base.default : sample=0% (threshold=3%)
+DEBUG: checking exe.improvement : 403.gcc,gcc_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1029,7 +1059,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 416.gamess,gamess_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.improvement : 416.gamess,gamess_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1191,42 +1221,12 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 437.leslie3d,leslie3d_base.default : sample=17% (threshold=10.290000000000001%)
+DEBUG: checking exe.improvement : 437.leslie3d,leslie3d_base.default : sample=-21% (threshold=10.32%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(185):         return (100 - result > threshold)
-output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
- --- modulename: output-bmk-results, funcname: get_short_long_diag
-output-bmk-results.py(113):     bmk = row["benchmark"]
-output-bmk-results.py(114):     rel_value = row["rel_" + metric]
-output-bmk-results.py(115):     prev_value = row[metric + "_x"]
-output-bmk-results.py(116):     curr_value = row[metric + "_y"]
-output-bmk-results.py(118):     if metric == "sample":
-output-bmk-results.py(119):         if curr_value == 999999999:
-output-bmk-results.py(122):         elif curr_value == 888888888:
-output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
-output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
-output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
-output-bmk-results.py(144):     suffix = ""
-output-bmk-results.py(145):     if metric == "sample":
-output-bmk-results.py(146):         prefix_regression = "slowed down by"
-output-bmk-results.py(147):         prefix_improvement = "sped up by"
-output-bmk-results.py(148):         suffix = "perf samples"
-output-bmk-results.py(159):     if sym_type=="symbol":
-output-bmk-results.py(162):       item=bmk
-output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
-output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
-output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
-output-bmk-results.py(231):         if metric == "sample" \
-output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
-output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
-output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
-output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
- --- modulename: output-bmk-results, funcname: write_csv
-output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
-output-bmk-results.py(43):     self.csvwriter.writerow(arr)
-output-bmk-results.py(236):             continue
+output-bmk-results.py(225):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -1275,7 +1275,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 445.gobmk,gobmk_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.improvement : 445.gobmk,gobmk_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1302,7 +1302,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 447.dealII,dealII_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.improvement : 447.dealII,dealII_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1356,7 +1356,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 453.povray,povray_base.default : sample=-2% (threshold=3%)
+DEBUG: checking exe.improvement : 453.povray,povray_base.default : sample=-1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1410,7 +1410,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 456.hmmer,hmmer_base.default : sample=2% (threshold=3.96%)
+DEBUG: checking exe.improvement : 456.hmmer,hmmer_base.default : sample=-1% (threshold=3.96%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1437,7 +1437,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 458.sjeng,sjeng_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.improvement : 458.sjeng,sjeng_base.default : sample=-1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1464,7 +1464,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 459.GemsFDTD,GemsFDTD_base.default : sample=-1% (threshold=3%)
+DEBUG: checking exe.improvement : 459.GemsFDTD,GemsFDTD_base.default : sample=1% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1518,7 +1518,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 464.h264ref,h264ref_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.improvement : 464.h264ref,h264ref_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1545,7 +1545,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 465.tonto,tonto_base.default : sample=2% (threshold=3%)
+DEBUG: checking exe.improvement : 465.tonto,tonto_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1599,7 +1599,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 471.omnetpp,omnetpp_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.improvement : 471.omnetpp,omnetpp_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1653,7 +1653,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 481.wrf,wrf_base.default : sample=-2% (threshold=3%)
+DEBUG: checking exe.improvement : 481.wrf,wrf_base.default : sample=2% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1680,7 +1680,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking exe.improvement : 482.sphinx3,sphinx_livepretend_base.default : sample=1% (threshold=3%)
+DEBUG: checking exe.improvement : 482.sphinx3,sphinx_livepretend_base.default : sample=0% (threshold=3%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -1752,7 +1752,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 400.perlbench,[.] S_regmatch : sample=-2% (threshold=15%)
+DEBUG: checking symbol.regression : 400.perlbench,[.] S_regmatch : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1774,7 +1774,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 400.perlbench,libc.so.6 : sample=-2% (threshold=15%)
+DEBUG: checking symbol.regression : 400.perlbench,libc.so.6 : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1801,7 +1801,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 401.bzip2,[.] mainSort : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 401.bzip2,[.] mainSort : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1828,7 +1828,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 401.bzip2,[.] mainGtU.part.0 : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 401.bzip2,[.] mainGtU.part.0 : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1855,7 +1855,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 401.bzip2,[.] fallbackSort : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 401.bzip2,[.] fallbackSort : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1882,7 +1882,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 401.bzip2,[.] BZ2_compressBlock : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 401.bzip2,[.] BZ2_compressBlock : sample=4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1909,7 +1909,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 401.bzip2,[.] BZ2_decompress : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 401.bzip2,[.] BZ2_decompress : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1936,7 +1936,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 403.gcc,[.] reg_is_remote_constant_p : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 403.gcc,[.] reg_is_remote_constant_p : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -1990,7 +1990,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 403.gcc,[.] memset : sample=-5% (threshold=15%)
+DEBUG: checking symbol.regression : 403.gcc,[.] memset : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2044,7 +2044,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 410.bwaves,[.] bi_cgstab_block_ : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 410.bwaves,[.] bi_cgstab_block_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2071,7 +2071,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 416.gamess,[.] forms_ : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 416.gamess,[.] forms_ : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2098,7 +2098,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 416.gamess,[.] twotff_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 416.gamess,[.] twotff_ : sample=3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2125,7 +2125,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 416.gamess,[.] dirfck_ : sample=-4% (threshold=15%)
+DEBUG: checking symbol.regression : 416.gamess,[.] dirfck_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2179,7 +2179,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 429.mcf,[.] primal_bea_mpp : sample=-2% (threshold=15%)
+DEBUG: checking symbol.regression : 429.mcf,[.] primal_bea_mpp : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2206,7 +2206,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 429.mcf,[.] refresh_potential : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 429.mcf,[.] refresh_potential : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2233,7 +2233,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_na : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_na : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2260,7 +2260,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_nn : sample=-3% (threshold=15%)
+DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_nn : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2287,7 +2287,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_mat_vec : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 433.milc,[.] mult_su3_mat_vec : sample=4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2314,7 +2314,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 433.milc,[.] mult_adj_su3_mat_vec : sample=3% (threshold=15%)
+DEBUG: checking symbol.regression : 433.milc,[.] mult_adj_su3_mat_vec : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2368,7 +2368,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 434.zeusmp,[.] hsmoc_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 434.zeusmp,[.] hsmoc_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2395,7 +2395,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 434.zeusmp,[.] lorentz_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 434.zeusmp,[.] lorentz_ : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2422,7 +2422,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 435.gromacs,[.] inl1130_ : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 435.gromacs,[.] inl1130_ : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2449,7 +2449,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 435.gromacs,[.] search_neighbours : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 435.gromacs,[.] search_neighbours : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2503,39 +2503,42 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxk_ : sample=22% (threshold=15%)
-output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
- --- modulename: output-bmk-results, funcname: is_entry_regression
-output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
-output-bmk-results.py(176):         return (result - 100 > threshold)
-output-bmk-results.py(225):             continue
-output-bmk-results.py(216):     for index, row in out_df.iterrows():
-output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
- --- modulename: output-bmk-results, funcname: get_threshold
-output-bmk-results.py(98):     if metric == "sample":
-output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
- --- modulename: output-bmk-results, funcname: get_specific_thresholds
-output-bmk-results.py(57):     if specific_variability is None:
-output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
-output-bmk-results.py(61):     if var.empty:
-output-bmk-results.py(63):     elif len(var)>1:
-output-bmk-results.py(68):     if var.iloc[0]['sample_variation_average']>0 :
-output-bmk-results.py(69):        threshold = ( var.iloc[0]['sample_variation_average'] )
-output-bmk-results.py(70):        if mode == "build":
-output-bmk-results.py(74):           threshold *= 3
-output-bmk-results.py(81):        return threshold
-output-bmk-results.py(100):        if not np.isnan(spec_thr):
-output-bmk-results.py(104):           spec_thr=max(spec_thr, default_threshold[(change_kind,metric,mode)])
-output-bmk-results.py(105):           return spec_thr
-output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
-output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxj_ : sample=29% (threshold=15%)
+DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxk_ : sample=-27% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(176):         return (result - 100 > threshold)
-output-bmk-results.py(225):             continue
+output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
+ --- modulename: output-bmk-results, funcname: get_short_long_diag
+output-bmk-results.py(113):     bmk = row["benchmark"]
+output-bmk-results.py(114):     rel_value = row["rel_" + metric]
+output-bmk-results.py(115):     prev_value = row[metric + "_x"]
+output-bmk-results.py(116):     curr_value = row[metric + "_y"]
+output-bmk-results.py(118):     if metric == "sample":
+output-bmk-results.py(119):         if curr_value == 999999999:
+output-bmk-results.py(122):         elif curr_value == 888888888:
+output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
+output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
+output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
+output-bmk-results.py(144):     suffix = ""
+output-bmk-results.py(145):     if metric == "sample":
+output-bmk-results.py(146):         prefix_regression = "slowed down by"
+output-bmk-results.py(147):         prefix_improvement = "sped up by"
+output-bmk-results.py(148):         suffix = "perf samples"
+output-bmk-results.py(159):     if sym_type=="symbol":
+output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
+output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
+output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
+output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
+output-bmk-results.py(231):         if metric == "sample" \
+output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
+output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
+output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
+output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
+ --- modulename: output-bmk-results, funcname: write_csv
+output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
+output-bmk-results.py(43):     self.csvwriter.writerow(arr)
+output-bmk-results.py(236):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -2557,12 +2560,42 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxi_ : sample=18% (threshold=15%)
+DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxj_ : sample=-40% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(176):         return (result - 100 > threshold)
-output-bmk-results.py(225):             continue
+output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
+ --- modulename: output-bmk-results, funcname: get_short_long_diag
+output-bmk-results.py(113):     bmk = row["benchmark"]
+output-bmk-results.py(114):     rel_value = row["rel_" + metric]
+output-bmk-results.py(115):     prev_value = row[metric + "_x"]
+output-bmk-results.py(116):     curr_value = row[metric + "_y"]
+output-bmk-results.py(118):     if metric == "sample":
+output-bmk-results.py(119):         if curr_value == 999999999:
+output-bmk-results.py(122):         elif curr_value == 888888888:
+output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
+output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
+output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
+output-bmk-results.py(144):     suffix = ""
+output-bmk-results.py(145):     if metric == "sample":
+output-bmk-results.py(146):         prefix_regression = "slowed down by"
+output-bmk-results.py(147):         prefix_improvement = "sped up by"
+output-bmk-results.py(148):         suffix = "perf samples"
+output-bmk-results.py(159):     if sym_type=="symbol":
+output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
+output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
+output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
+output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
+output-bmk-results.py(231):         if metric == "sample" \
+output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
+output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
+output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
+output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
+ --- modulename: output-bmk-results, funcname: write_csv
+output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
+output-bmk-results.py(43):     self.csvwriter.writerow(arr)
+output-bmk-results.py(236):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -2584,12 +2617,42 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] extrapi_ : sample=8% (threshold=15%)
+DEBUG: checking symbol.regression : 437.leslie3d,[.] fluxi_ : sample=-23% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(176):         return (result - 100 > threshold)
-output-bmk-results.py(225):             continue
+output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
+ --- modulename: output-bmk-results, funcname: get_short_long_diag
+output-bmk-results.py(113):     bmk = row["benchmark"]
+output-bmk-results.py(114):     rel_value = row["rel_" + metric]
+output-bmk-results.py(115):     prev_value = row[metric + "_x"]
+output-bmk-results.py(116):     curr_value = row[metric + "_y"]
+output-bmk-results.py(118):     if metric == "sample":
+output-bmk-results.py(119):         if curr_value == 999999999:
+output-bmk-results.py(122):         elif curr_value == 888888888:
+output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
+output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
+output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
+output-bmk-results.py(144):     suffix = ""
+output-bmk-results.py(145):     if metric == "sample":
+output-bmk-results.py(146):         prefix_regression = "slowed down by"
+output-bmk-results.py(147):         prefix_improvement = "sped up by"
+output-bmk-results.py(148):         suffix = "perf samples"
+output-bmk-results.py(159):     if sym_type=="symbol":
+output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
+output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
+output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
+output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
+output-bmk-results.py(231):         if metric == "sample" \
+output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
+output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
+output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
+output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
+ --- modulename: output-bmk-results, funcname: write_csv
+output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
+output-bmk-results.py(43):     self.csvwriter.writerow(arr)
+output-bmk-results.py(236):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -2611,7 +2674,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] extrapj_ : sample=4% (threshold=15%)
+DEBUG: checking symbol.regression : 437.leslie3d,[.] extrapi_ : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2638,7 +2701,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 437.leslie3d,[.] extrapk_ : sample=4% (threshold=15%)
+DEBUG: checking symbol.regression : 437.leslie3d,[.] extrapj_ : sample=-5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2719,7 +2782,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil32calc_pair_energy_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil32calc_pair_energy_merge_fullelectEP9nonbonded : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2746,7 +2809,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil16calc_pair_energyEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil16calc_pair_energyEP9nonbonded : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2773,7 +2836,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil9calc_pairEP9nonbonded : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil25calc_pair_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2800,7 +2863,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil25calc_pair_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 444.namd,[.] _ZN20ComputeNonbondedUtil9calc_pairEP9nonbonded : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2827,7 +2890,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 445.gobmk,[.] do_play_move : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 445.gobmk,[.] do_play_move : sample=6% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2854,7 +2917,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 445.gobmk,[.] fastlib : sample=0% (threshold=15.09%)
+DEBUG: checking symbol.regression : 445.gobmk,[.] fastlib : sample=-6% (threshold=15.09%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2881,7 +2944,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 447.dealII,[.] _ZNK9MappingQ1ILi3EE12compute_fillERK12TriaIteratorILi3E15DoFCellAccessorILi3EEEjN10QProjectorILi3EE17DataSetDescriptorERNS0_12InternalDataERSt6vectorI5PointILi3EESaISE_EE : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 447.dealII,[.] _ZNK9MappingQ1ILi3EE12compute_fillERK12TriaIteratorILi3E15DoFCellAccessorILi3EEEjN10QProjectorILi3EE17DataSetDescriptorERNS0_12InternalDataERSt6vectorI5PointILi3EESaISE_EE : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2935,7 +2998,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 447.dealII,libstdc++.so.6.0.33 : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 447.dealII,libstdc++.so.6.0.33 : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2962,7 +3025,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 447.dealII,[.] _ZN13LaplaceSolver6SolverILi3EE22assemble_linear_systemERNS1_12LinearSystemE : sample=5% (threshold=15%)
+DEBUG: checking symbol.regression : 447.dealII,[.] _ZN13LaplaceSolver6SolverILi3EE22assemble_linear_systemERNS1_12LinearSystemE : sample=-5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -2989,7 +3052,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 447.dealII,[.] _ZSt18_Rb_tree_incrementPKSt18_Rb_tree_node_base : sample=3% (threshold=15%)
+DEBUG: checking symbol.regression : 447.dealII,[.] _ZSt18_Rb_tree_incrementPKSt18_Rb_tree_node_base : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3016,7 +3079,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex10SPxSteepPR8entered4ENS_5SPxIdEi : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex10SPxSteepPR8entered4ENS_5SPxIdEi : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3043,7 +3106,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex8SSVector18assign2productFullERKNS_5SVSetERKS0_ : sample=-5% (threshold=15%)
+DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex8SSVector18assign2productFullERKNS_5SVSetERKS0_ : sample=6% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3070,7 +3133,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex9CLUFactor16initFactorMatrixEPPNS_7SVectorEd : sample=-5% (threshold=15%)
+DEBUG: checking symbol.regression : 450.soplex,[.] _ZN6soplex9CLUFactor16initFactorMatrixEPPNS_7SVectorEd : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3097,7 +3160,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL23All_Plane_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL31All_CSG_Intersect_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3124,7 +3187,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL31All_CSG_Intersect_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=7% (threshold=15%)
+DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL23All_Plane_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3151,7 +3214,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL24All_Sphere_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-3% (threshold=15%)
+DEBUG: checking symbol.regression : 453.povray,[.] _ZN3povL24All_Sphere_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3178,7 +3241,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 453.povray,[.] _ZN3pov17Check_And_EnqueueEPNS_21Priority_Queue_StructEPNS_16BBox_Tree_StructEPNS_19Bounding_Box_StructEPNS_14Rayinfo_StructE : sample=-1% (threshold=15.36%)
+DEBUG: checking symbol.regression : 453.povray,[.] _ZN3pov17Check_And_EnqueueEPNS_21Priority_Queue_StructEPNS_16BBox_Tree_StructEPNS_19Bounding_Box_StructEPNS_14Rayinfo_StructE : sample=-3% (threshold=15.36%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3259,7 +3322,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 456.hmmer,[.] P7Viterbi : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 456.hmmer,[.] P7Viterbi : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3286,7 +3349,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 458.sjeng,[.] std_eval : sample=7% (threshold=15%)
+DEBUG: checking symbol.regression : 458.sjeng,[.] std_eval : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3313,7 +3376,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 458.sjeng,[.] gen : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 458.sjeng,[.] gen : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3340,7 +3403,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 458.sjeng,[.] setup_attackers : sample=5% (threshold=15%)
+DEBUG: checking symbol.regression : 458.sjeng,[.] setup_attackers : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3367,7 +3430,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdatee : sample=-3% (threshold=15%)
+DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdatee : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3394,7 +3457,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __nft_mod_MOD_nft_store : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __nft_mod_MOD_nft_store : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3421,7 +3484,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdateh : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdateh : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3448,7 +3511,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __update_mod_MOD_updatee_homo : sample=-2% (threshold=15%)
+DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __update_mod_MOD_updateh_homo : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3475,7 +3538,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __update_mod_MOD_updateh_homo : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 459.GemsFDTD,[.] __update_mod_MOD_updatee_homo : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3502,7 +3565,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 462.libquantum,[.] quantum_toffoli : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 462.libquantum,[.] quantum_toffoli : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3556,7 +3619,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 462.libquantum,[.] quantum_cnot : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 462.libquantum,[.] quantum_cnot : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3583,7 +3646,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 464.h264ref,[.] SetupFastFullPelSearch : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 464.h264ref,[.] SetupFastFullPelSearch : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3610,7 +3673,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 464.h264ref,libc.so.6 : sample=4% (threshold=15%)
+DEBUG: checking symbol.regression : 464.h264ref,libc.so.6 : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3637,7 +3700,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 464.h264ref,[.] __memcpy_neon : sample=4% (threshold=15%)
+DEBUG: checking symbol.regression : 464.h264ref,[.] __memcpy_neon : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3664,7 +3727,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 464.h264ref,[.] FastFullPelBlockMotionSearch : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 464.h264ref,[.] FastFullPelBlockMotionSearch : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3686,7 +3749,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 465.tonto,libm.so.6 : sample=-2% (threshold=15%)
+DEBUG: checking symbol.regression : 465.tonto,libm.so.6 : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3713,7 +3776,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 465.tonto,[.] __shell2_module_MOD_make_ft_1 : sample=4% (threshold=15%)
+DEBUG: checking symbol.regression : 465.tonto,[.] __shell2_module_MOD_make_ft_1 : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3735,7 +3798,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 465.tonto,[.] __sincosl : sample=-6% (threshold=15%)
+DEBUG: checking symbol.regression : 465.tonto,[.] __sincosl : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3762,7 +3825,29 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 465.tonto,[.] __shell1quartet_module_MOD_make_esfs.isra.0 : sample=2% (threshold=15%)
+DEBUG: checking symbol.regression : 465.tonto,[.] __shell1quartet_module_MOD_make_esfs.isra.0 : sample=7% (threshold=15%)
+output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
+ --- modulename: output-bmk-results, funcname: is_entry_regression
+output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
+output-bmk-results.py(176):         return (result - 100 > threshold)
+output-bmk-results.py(225):             continue
+output-bmk-results.py(216):     for index, row in out_df.iterrows():
+output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
+ --- modulename: output-bmk-results, funcname: get_threshold
+output-bmk-results.py(98):     if metric == "sample":
+output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
+ --- modulename: output-bmk-results, funcname: get_specific_thresholds
+output-bmk-results.py(57):     if specific_variability is None:
+output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
+output-bmk-results.py(61):     if var.empty:
+output-bmk-results.py(62):         return np.nan
+output-bmk-results.py(100):        if not np.isnan(spec_thr):
+output-bmk-results.py(107):     if metric == "num_vect_loops" or metric == "num_sve_loops":
+output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mode)]
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+DEBUG: checking symbol.regression : 465.tonto,[.] __cexpl : sample=-10% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3816,7 +3901,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 471.omnetpp,[.] _ZN12cMessageHeap8getFirstEv : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 471.omnetpp,[.] _ZN12cMessageHeap8getFirstEv : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3838,7 +3923,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 471.omnetpp,libc.so.6 : sample=-7% (threshold=15%)
+DEBUG: checking symbol.regression : 471.omnetpp,libc.so.6 : sample=3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3865,7 +3950,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 471.omnetpp,[.] _ZN5cGate7deliverEP8cMessaged : sample=-11% (threshold=15%)
+DEBUG: checking symbol.regression : 471.omnetpp,[.] _ZN5cGate7deliverEP8cMessaged : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3919,7 +4004,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 473.astar,[.] _ZN7way2obj12releasepointEii : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 473.astar,[.] _ZN7way2obj12releasepointEii : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3946,7 +4031,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 473.astar,[.] _ZN9regwayobj10makebound2ER9flexarrayIP6regobjES4_ : sample=1% (threshold=15%)
+DEBUG: checking symbol.regression : 473.astar,[.] _ZN9regwayobj10makebound2ER9flexarrayIP6regobjES4_ : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -3973,7 +4058,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 481.wrf,[.] __module_advect_em_MOD_advect_scalar : sample=0% (threshold=15%)
+DEBUG: checking symbol.regression : 481.wrf,[.] __module_advect_em_MOD_advect_scalar : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4000,7 +4085,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 481.wrf,[.] __module_small_step_em_MOD_advance_w : sample=-3% (threshold=15%)
+DEBUG: checking symbol.regression : 481.wrf,[.] __module_small_step_em_MOD_advance_w : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4027,7 +4112,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 481.wrf,libm.so.6 : sample=6% (threshold=15%)
+DEBUG: checking symbol.regression : 481.wrf,libm.so.6 : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4049,7 +4134,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 481.wrf,libc.so.6 : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 481.wrf,libc.so.6 : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4076,7 +4161,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 482.sphinx3,[.] vector_gautbl_eval_logs3 : sample=3% (threshold=15%)
+DEBUG: checking symbol.regression : 481.wrf,[.] powf@@GLIBC_2.27 : sample=-11% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4103,7 +4188,34 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.regression : 482.sphinx3,[.] mgau_eval : sample=-1% (threshold=15%)
+DEBUG: checking symbol.regression : 482.sphinx3,[.] vector_gautbl_eval_logs3 : sample=0% (threshold=15%)
+output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
+ --- modulename: output-bmk-results, funcname: is_entry_regression
+output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
+output-bmk-results.py(176):         return (result - 100 > threshold)
+output-bmk-results.py(225):             continue
+output-bmk-results.py(216):     for index, row in out_df.iterrows():
+output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
+ --- modulename: output-bmk-results, funcname: get_threshold
+output-bmk-results.py(98):     if metric == "sample":
+output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
+ --- modulename: output-bmk-results, funcname: get_specific_thresholds
+output-bmk-results.py(57):     if specific_variability is None:
+output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
+output-bmk-results.py(61):     if var.empty:
+output-bmk-results.py(63):     elif len(var)>1:
+output-bmk-results.py(68):     if var.iloc[0]['sample_variation_average']>0 :
+output-bmk-results.py(69):        threshold = ( var.iloc[0]['sample_variation_average'] )
+output-bmk-results.py(70):        if mode == "build":
+output-bmk-results.py(74):           threshold *= 3
+output-bmk-results.py(81):        return threshold
+output-bmk-results.py(100):        if not np.isnan(spec_thr):
+output-bmk-results.py(104):           spec_thr=max(spec_thr, default_threshold[(change_kind,metric,mode)])
+output-bmk-results.py(105):           return spec_thr
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+DEBUG: checking symbol.regression : 482.sphinx3,[.] mgau_eval : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_regression
 output-bmk-results.py(175):     if metric in metric_utils.higher_regress_metrics:
@@ -4148,7 +4260,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 400.perlbench,[.] S_regmatch : sample=-2% (threshold=15%)
+DEBUG: checking symbol.improvement : 400.perlbench,[.] S_regmatch : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4170,7 +4282,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 400.perlbench,libc.so.6 : sample=-2% (threshold=15%)
+DEBUG: checking symbol.improvement : 400.perlbench,libc.so.6 : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4197,7 +4309,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 401.bzip2,[.] mainSort : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 401.bzip2,[.] mainSort : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4224,7 +4336,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 401.bzip2,[.] mainGtU.part.0 : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 401.bzip2,[.] mainGtU.part.0 : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4251,7 +4363,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 401.bzip2,[.] fallbackSort : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 401.bzip2,[.] fallbackSort : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4278,7 +4390,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 401.bzip2,[.] BZ2_compressBlock : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 401.bzip2,[.] BZ2_compressBlock : sample=4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4305,7 +4417,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 401.bzip2,[.] BZ2_decompress : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 401.bzip2,[.] BZ2_decompress : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4332,7 +4444,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 403.gcc,[.] reg_is_remote_constant_p : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 403.gcc,[.] reg_is_remote_constant_p : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4386,7 +4498,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 403.gcc,[.] memset : sample=-5% (threshold=15%)
+DEBUG: checking symbol.improvement : 403.gcc,[.] memset : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4440,7 +4552,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 410.bwaves,[.] bi_cgstab_block_ : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 410.bwaves,[.] bi_cgstab_block_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4467,7 +4579,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 416.gamess,[.] forms_ : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 416.gamess,[.] forms_ : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4494,7 +4606,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 416.gamess,[.] twotff_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 416.gamess,[.] twotff_ : sample=3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4521,7 +4633,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 416.gamess,[.] dirfck_ : sample=-4% (threshold=15%)
+DEBUG: checking symbol.improvement : 416.gamess,[.] dirfck_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4575,7 +4687,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 429.mcf,[.] primal_bea_mpp : sample=-2% (threshold=15%)
+DEBUG: checking symbol.improvement : 429.mcf,[.] primal_bea_mpp : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4602,7 +4714,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 429.mcf,[.] refresh_potential : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 429.mcf,[.] refresh_potential : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4629,7 +4741,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_na : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_na : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4656,7 +4768,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_nn : sample=-3% (threshold=15%)
+DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_nn : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4683,7 +4795,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_mat_vec : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 433.milc,[.] mult_su3_mat_vec : sample=4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4710,7 +4822,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 433.milc,[.] mult_adj_su3_mat_vec : sample=3% (threshold=15%)
+DEBUG: checking symbol.improvement : 433.milc,[.] mult_adj_su3_mat_vec : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4764,7 +4876,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 434.zeusmp,[.] hsmoc_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 434.zeusmp,[.] hsmoc_ : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4791,7 +4903,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 434.zeusmp,[.] lorentz_ : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 434.zeusmp,[.] lorentz_ : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4818,7 +4930,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 435.gromacs,[.] inl1130_ : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 435.gromacs,[.] inl1130_ : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4845,7 +4957,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 435.gromacs,[.] search_neighbours : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 435.gromacs,[.] search_neighbours : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -4899,99 +5011,12 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxk_ : sample=22% (threshold=15%)
+DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxk_ : sample=-27% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(185):         return (100 - result > threshold)
-output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
- --- modulename: output-bmk-results, funcname: get_short_long_diag
-output-bmk-results.py(113):     bmk = row["benchmark"]
-output-bmk-results.py(114):     rel_value = row["rel_" + metric]
-output-bmk-results.py(115):     prev_value = row[metric + "_x"]
-output-bmk-results.py(116):     curr_value = row[metric + "_y"]
-output-bmk-results.py(118):     if metric == "sample":
-output-bmk-results.py(119):         if curr_value == 999999999:
-output-bmk-results.py(122):         elif curr_value == 888888888:
-output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
-output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
-output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
-output-bmk-results.py(144):     suffix = ""
-output-bmk-results.py(145):     if metric == "sample":
-output-bmk-results.py(146):         prefix_regression = "slowed down by"
-output-bmk-results.py(147):         prefix_improvement = "sped up by"
-output-bmk-results.py(148):         suffix = "perf samples"
-output-bmk-results.py(159):     if sym_type=="symbol":
-output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
-output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
-output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
-output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
-output-bmk-results.py(231):         if metric == "sample" \
-output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
-output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
-output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
-output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
- --- modulename: output-bmk-results, funcname: write_csv
-output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
-output-bmk-results.py(43):     self.csvwriter.writerow(arr)
-output-bmk-results.py(236):             continue
-output-bmk-results.py(216):     for index, row in out_df.iterrows():
-output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
- --- modulename: output-bmk-results, funcname: get_threshold
-output-bmk-results.py(98):     if metric == "sample":
-output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
- --- modulename: output-bmk-results, funcname: get_specific_thresholds
-output-bmk-results.py(57):     if specific_variability is None:
-output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
-output-bmk-results.py(61):     if var.empty:
-output-bmk-results.py(63):     elif len(var)>1:
-output-bmk-results.py(68):     if var.iloc[0]['sample_variation_average']>0 :
-output-bmk-results.py(69):        threshold = ( var.iloc[0]['sample_variation_average'] )
-output-bmk-results.py(70):        if mode == "build":
-output-bmk-results.py(74):           threshold *= 3
-output-bmk-results.py(81):        return threshold
-output-bmk-results.py(100):        if not np.isnan(spec_thr):
-output-bmk-results.py(104):           spec_thr=max(spec_thr, default_threshold[(change_kind,metric,mode)])
-output-bmk-results.py(105):           return spec_thr
-output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
-output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxj_ : sample=29% (threshold=15%)
-output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
- --- modulename: output-bmk-results, funcname: is_entry_improvement
-output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
-output-bmk-results.py(185):         return (100 - result > threshold)
-output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
- --- modulename: output-bmk-results, funcname: get_short_long_diag
-output-bmk-results.py(113):     bmk = row["benchmark"]
-output-bmk-results.py(114):     rel_value = row["rel_" + metric]
-output-bmk-results.py(115):     prev_value = row[metric + "_x"]
-output-bmk-results.py(116):     curr_value = row[metric + "_y"]
-output-bmk-results.py(118):     if metric == "sample":
-output-bmk-results.py(119):         if curr_value == 999999999:
-output-bmk-results.py(122):         elif curr_value == 888888888:
-output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
-output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
-output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
-output-bmk-results.py(144):     suffix = ""
-output-bmk-results.py(145):     if metric == "sample":
-output-bmk-results.py(146):         prefix_regression = "slowed down by"
-output-bmk-results.py(147):         prefix_improvement = "sped up by"
-output-bmk-results.py(148):         suffix = "perf samples"
-output-bmk-results.py(159):     if sym_type=="symbol":
-output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
-output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
-output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
-output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
-output-bmk-results.py(231):         if metric == "sample" \
-output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
-output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
-output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
-output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
- --- modulename: output-bmk-results, funcname: write_csv
-output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
-output-bmk-results.py(43):     self.csvwriter.writerow(arr)
-output-bmk-results.py(236):             continue
+output-bmk-results.py(225):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -5013,42 +5038,12 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxi_ : sample=18% (threshold=15%)
+DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxj_ : sample=-40% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
 output-bmk-results.py(185):         return (100 - result > threshold)
-output-bmk-results.py(227):         percent_change, short_diag, long_diag = get_short_long_diag(row, metric, sym_type, change_kind)
- --- modulename: output-bmk-results, funcname: get_short_long_diag
-output-bmk-results.py(113):     bmk = row["benchmark"]
-output-bmk-results.py(114):     rel_value = row["rel_" + metric]
-output-bmk-results.py(115):     prev_value = row[metric + "_x"]
-output-bmk-results.py(116):     curr_value = row[metric + "_y"]
-output-bmk-results.py(118):     if metric == "sample":
-output-bmk-results.py(119):         if curr_value == 999999999:
-output-bmk-results.py(122):         elif curr_value == 888888888:
-output-bmk-results.py(125):         elif prev_value == 999999999 and curr_value == 888888888:
-output-bmk-results.py(128):         elif prev_value == 888888888 and curr_value < 888888888:
-output-bmk-results.py(131):         elif prev_value == 999999999 and curr_value < 888888888:
-output-bmk-results.py(144):     suffix = ""
-output-bmk-results.py(145):     if metric == "sample":
-output-bmk-results.py(146):         prefix_regression = "slowed down by"
-output-bmk-results.py(147):         prefix_improvement = "sped up by"
-output-bmk-results.py(148):         suffix = "perf samples"
-output-bmk-results.py(159):     if sym_type=="symbol":
-output-bmk-results.py(160):       item=bmk+":"+row["symbol"]
-output-bmk-results.py(164):     short_diag = "{1} {2}% - {0}".format(item, locals()["prefix_" + change_kind], abs(rel_value - 100))
-output-bmk-results.py(165):     long_diag = "{0} - from {1} to {2} {3}".format(short_diag, prev_value, curr_value, suffix)
-output-bmk-results.py(166):     return abs(rel_value - 100), short_diag, long_diag
-output-bmk-results.py(231):         if metric == "sample" \
-output-bmk-results.py(232):           and row['symbol_md5sum_x'] == row['symbol_md5sum_y'] \
-output-bmk-results.py(233):           and row['symbol_md5sum_x'] != "-1" \
-output-bmk-results.py(234):           and row['symbol_md5sum_x'] != "d41d8cd98f00b204e9800998ecf8427e":
-output-bmk-results.py(235):             f_skip.write_csv((row["benchmark"], row["symbol"], short_diag, long_diag))
- --- modulename: output-bmk-results, funcname: write_csv
-output-bmk-results.py(41):     if not self.predicate or not self.csvwriter:
-output-bmk-results.py(43):     self.csvwriter.writerow(arr)
-output-bmk-results.py(236):             continue
+output-bmk-results.py(225):             continue
 output-bmk-results.py(216):     for index, row in out_df.iterrows():
 output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
  --- modulename: output-bmk-results, funcname: get_threshold
@@ -5070,7 +5065,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] extrapi_ : sample=8% (threshold=15%)
+DEBUG: checking symbol.improvement : 437.leslie3d,[.] fluxi_ : sample=-23% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5097,7 +5092,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] extrapj_ : sample=4% (threshold=15%)
+DEBUG: checking symbol.improvement : 437.leslie3d,[.] extrapi_ : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5124,7 +5119,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 437.leslie3d,[.] extrapk_ : sample=4% (threshold=15%)
+DEBUG: checking symbol.improvement : 437.leslie3d,[.] extrapj_ : sample=-5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5205,7 +5200,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil32calc_pair_energy_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil32calc_pair_energy_merge_fullelectEP9nonbonded : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5232,7 +5227,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil16calc_pair_energyEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil16calc_pair_energyEP9nonbonded : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5259,7 +5254,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil9calc_pairEP9nonbonded : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil25calc_pair_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5286,7 +5281,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil25calc_pair_merge_fullelectEP9nonbonded : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 444.namd,[.] _ZN20ComputeNonbondedUtil9calc_pairEP9nonbonded : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5313,7 +5308,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 445.gobmk,[.] do_play_move : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 445.gobmk,[.] do_play_move : sample=6% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5340,7 +5335,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 445.gobmk,[.] fastlib : sample=0% (threshold=15.09%)
+DEBUG: checking symbol.improvement : 445.gobmk,[.] fastlib : sample=-6% (threshold=15.09%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5367,7 +5362,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 447.dealII,[.] _ZNK9MappingQ1ILi3EE12compute_fillERK12TriaIteratorILi3E15DoFCellAccessorILi3EEEjN10QProjectorILi3EE17DataSetDescriptorERNS0_12InternalDataERSt6vectorI5PointILi3EESaISE_EE : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 447.dealII,[.] _ZNK9MappingQ1ILi3EE12compute_fillERK12TriaIteratorILi3E15DoFCellAccessorILi3EEEjN10QProjectorILi3EE17DataSetDescriptorERNS0_12InternalDataERSt6vectorI5PointILi3EESaISE_EE : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5421,7 +5416,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 447.dealII,libstdc++.so.6.0.33 : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 447.dealII,libstdc++.so.6.0.33 : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5448,7 +5443,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 447.dealII,[.] _ZN13LaplaceSolver6SolverILi3EE22assemble_linear_systemERNS1_12LinearSystemE : sample=5% (threshold=15%)
+DEBUG: checking symbol.improvement : 447.dealII,[.] _ZN13LaplaceSolver6SolverILi3EE22assemble_linear_systemERNS1_12LinearSystemE : sample=-5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5475,7 +5470,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 447.dealII,[.] _ZSt18_Rb_tree_incrementPKSt18_Rb_tree_node_base : sample=3% (threshold=15%)
+DEBUG: checking symbol.improvement : 447.dealII,[.] _ZSt18_Rb_tree_incrementPKSt18_Rb_tree_node_base : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5502,7 +5497,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex10SPxSteepPR8entered4ENS_5SPxIdEi : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex10SPxSteepPR8entered4ENS_5SPxIdEi : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5529,7 +5524,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex8SSVector18assign2productFullERKNS_5SVSetERKS0_ : sample=-5% (threshold=15%)
+DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex8SSVector18assign2productFullERKNS_5SVSetERKS0_ : sample=6% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5556,7 +5551,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex9CLUFactor16initFactorMatrixEPPNS_7SVectorEd : sample=-5% (threshold=15%)
+DEBUG: checking symbol.improvement : 450.soplex,[.] _ZN6soplex9CLUFactor16initFactorMatrixEPPNS_7SVectorEd : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5583,7 +5578,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL23All_Plane_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL31All_CSG_Intersect_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5610,7 +5605,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL31All_CSG_Intersect_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=7% (threshold=15%)
+DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL23All_Plane_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5637,7 +5632,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL24All_Sphere_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-3% (threshold=15%)
+DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3povL24All_Sphere_IntersectionsEPNS_13Object_StructEPNS_10Ray_StructEPNS_13istack_structE : sample=-4% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5664,7 +5659,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3pov17Check_And_EnqueueEPNS_21Priority_Queue_StructEPNS_16BBox_Tree_StructEPNS_19Bounding_Box_StructEPNS_14Rayinfo_StructE : sample=-1% (threshold=15.36%)
+DEBUG: checking symbol.improvement : 453.povray,[.] _ZN3pov17Check_And_EnqueueEPNS_21Priority_Queue_StructEPNS_16BBox_Tree_StructEPNS_19Bounding_Box_StructEPNS_14Rayinfo_StructE : sample=-3% (threshold=15.36%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5745,7 +5740,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 456.hmmer,[.] P7Viterbi : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 456.hmmer,[.] P7Viterbi : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5772,7 +5767,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 458.sjeng,[.] std_eval : sample=7% (threshold=15%)
+DEBUG: checking symbol.improvement : 458.sjeng,[.] std_eval : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5799,7 +5794,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 458.sjeng,[.] gen : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 458.sjeng,[.] gen : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5826,7 +5821,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 458.sjeng,[.] setup_attackers : sample=5% (threshold=15%)
+DEBUG: checking symbol.improvement : 458.sjeng,[.] setup_attackers : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5853,7 +5848,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdatee : sample=-3% (threshold=15%)
+DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdatee : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5880,7 +5875,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __nft_mod_MOD_nft_store : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __nft_mod_MOD_nft_store : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5907,7 +5902,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdateh : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __upml_mod_MOD_upmlupdateh : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5934,7 +5929,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __update_mod_MOD_updatee_homo : sample=-2% (threshold=15%)
+DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __update_mod_MOD_updateh_homo : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5961,7 +5956,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __update_mod_MOD_updateh_homo : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 459.GemsFDTD,[.] __update_mod_MOD_updatee_homo : sample=1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -5988,7 +5983,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 462.libquantum,[.] quantum_toffoli : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 462.libquantum,[.] quantum_toffoli : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6042,7 +6037,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 462.libquantum,[.] quantum_cnot : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 462.libquantum,[.] quantum_cnot : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6069,7 +6064,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 464.h264ref,[.] SetupFastFullPelSearch : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 464.h264ref,[.] SetupFastFullPelSearch : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6096,7 +6091,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 464.h264ref,libc.so.6 : sample=4% (threshold=15%)
+DEBUG: checking symbol.improvement : 464.h264ref,libc.so.6 : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6123,7 +6118,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 464.h264ref,[.] __memcpy_neon : sample=4% (threshold=15%)
+DEBUG: checking symbol.improvement : 464.h264ref,[.] __memcpy_neon : sample=2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6150,7 +6145,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 464.h264ref,[.] FastFullPelBlockMotionSearch : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 464.h264ref,[.] FastFullPelBlockMotionSearch : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6172,7 +6167,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 465.tonto,libm.so.6 : sample=-2% (threshold=15%)
+DEBUG: checking symbol.improvement : 465.tonto,libm.so.6 : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6199,7 +6194,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 465.tonto,[.] __shell2_module_MOD_make_ft_1 : sample=4% (threshold=15%)
+DEBUG: checking symbol.improvement : 465.tonto,[.] __shell2_module_MOD_make_ft_1 : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6221,7 +6216,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 465.tonto,[.] __sincosl : sample=-6% (threshold=15%)
+DEBUG: checking symbol.improvement : 465.tonto,[.] __sincosl : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6248,7 +6243,29 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 465.tonto,[.] __shell1quartet_module_MOD_make_esfs.isra.0 : sample=2% (threshold=15%)
+DEBUG: checking symbol.improvement : 465.tonto,[.] __shell1quartet_module_MOD_make_esfs.isra.0 : sample=7% (threshold=15%)
+output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
+ --- modulename: output-bmk-results, funcname: is_entry_improvement
+output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
+output-bmk-results.py(185):         return (100 - result > threshold)
+output-bmk-results.py(225):             continue
+output-bmk-results.py(216):     for index, row in out_df.iterrows():
+output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
+ --- modulename: output-bmk-results, funcname: get_threshold
+output-bmk-results.py(98):     if metric == "sample":
+output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
+ --- modulename: output-bmk-results, funcname: get_specific_thresholds
+output-bmk-results.py(57):     if specific_variability is None:
+output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
+output-bmk-results.py(61):     if var.empty:
+output-bmk-results.py(62):         return np.nan
+output-bmk-results.py(100):        if not np.isnan(spec_thr):
+output-bmk-results.py(107):     if metric == "num_vect_loops" or metric == "num_sve_loops":
+output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mode)]
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+DEBUG: checking symbol.improvement : 465.tonto,[.] __cexpl : sample=-10% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6302,7 +6319,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 471.omnetpp,[.] _ZN12cMessageHeap8getFirstEv : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 471.omnetpp,[.] _ZN12cMessageHeap8getFirstEv : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6324,7 +6341,7 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 471.omnetpp,libc.so.6 : sample=-7% (threshold=15%)
+DEBUG: checking symbol.improvement : 471.omnetpp,libc.so.6 : sample=3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6351,7 +6368,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 471.omnetpp,[.] _ZN5cGate7deliverEP8cMessaged : sample=-11% (threshold=15%)
+DEBUG: checking symbol.improvement : 471.omnetpp,[.] _ZN5cGate7deliverEP8cMessaged : sample=-3% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6405,7 +6422,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 473.astar,[.] _ZN7way2obj12releasepointEii : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 473.astar,[.] _ZN7way2obj12releasepointEii : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6432,7 +6449,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 473.astar,[.] _ZN9regwayobj10makebound2ER9flexarrayIP6regobjES4_ : sample=1% (threshold=15%)
+DEBUG: checking symbol.improvement : 473.astar,[.] _ZN9regwayobj10makebound2ER9flexarrayIP6regobjES4_ : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6459,7 +6476,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 481.wrf,[.] __module_advect_em_MOD_advect_scalar : sample=0% (threshold=15%)
+DEBUG: checking symbol.improvement : 481.wrf,[.] __module_advect_em_MOD_advect_scalar : sample=-1% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6486,7 +6503,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 481.wrf,[.] __module_small_step_em_MOD_advance_w : sample=-3% (threshold=15%)
+DEBUG: checking symbol.improvement : 481.wrf,[.] __module_small_step_em_MOD_advance_w : sample=5% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6513,7 +6530,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 481.wrf,libm.so.6 : sample=6% (threshold=15%)
+DEBUG: checking symbol.improvement : 481.wrf,libm.so.6 : sample=-8% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6535,7 +6552,34 @@ output-bmk-results.py(110):     return default_threshold[(change_kind,metric,mod
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 481.wrf,libc.so.6 : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 481.wrf,libc.so.6 : sample=2% (threshold=15%)
+output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
+ --- modulename: output-bmk-results, funcname: is_entry_improvement
+output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
+output-bmk-results.py(185):         return (100 - result > threshold)
+output-bmk-results.py(225):             continue
+output-bmk-results.py(216):     for index, row in out_df.iterrows():
+output-bmk-results.py(218):         threshold = get_threshold(sym_type, metric, mode, row["benchmark"], row["symbol"])
+ --- modulename: output-bmk-results, funcname: get_threshold
+output-bmk-results.py(98):     if metric == "sample":
+output-bmk-results.py(99):        spec_thr = get_specific_thresholds(metric, mode, bmk, symb)
+ --- modulename: output-bmk-results, funcname: get_specific_thresholds
+output-bmk-results.py(57):     if specific_variability is None:
+output-bmk-results.py(60):     var = specific_variability[ (specific_variability['benchmark'] == bmk) & (specific_variability['symbol'].str.strip() == symb)]
+output-bmk-results.py(61):     if var.empty:
+output-bmk-results.py(63):     elif len(var)>1:
+output-bmk-results.py(68):     if var.iloc[0]['sample_variation_average']>0 :
+output-bmk-results.py(69):        threshold = ( var.iloc[0]['sample_variation_average'] )
+output-bmk-results.py(70):        if mode == "build":
+output-bmk-results.py(74):           threshold *= 3
+output-bmk-results.py(81):        return threshold
+output-bmk-results.py(100):        if not np.isnan(spec_thr):
+output-bmk-results.py(104):           spec_thr=max(spec_thr, default_threshold[(change_kind,metric,mode)])
+output-bmk-results.py(105):           return spec_thr
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
+output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
+DEBUG: checking symbol.improvement : 481.wrf,[.] powf@@GLIBC_2.27 : sample=-11% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6562,7 +6606,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 482.sphinx3,[.] vector_gautbl_eval_logs3 : sample=3% (threshold=15%)
+DEBUG: checking symbol.improvement : 482.sphinx3,[.] vector_gautbl_eval_logs3 : sample=0% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics:
@@ -6589,7 +6633,7 @@ output-bmk-results.py(105):           return spec_thr
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
 output-bmk-results.py(221):              .format(sym_type, change_kind, row["benchmark"], row["symbol"], metric, 100-row["rel_" + metric], threshold))
 output-bmk-results.py(220):         print("DEBUG: checking {0}.{1} : {2},{3} : {4}={5}% (threshold={6}%)"\
-DEBUG: checking symbol.improvement : 482.sphinx3,[.] mgau_eval : sample=-1% (threshold=15%)
+DEBUG: checking symbol.improvement : 482.sphinx3,[.] mgau_eval : sample=-2% (threshold=15%)
 output-bmk-results.py(224):         if not is_entry_xxx[change_kind](metric, row["rel_" + metric], threshold):
  --- modulename: output-bmk-results, funcname: is_entry_improvement
 output-bmk-results.py(184):     if metric in metric_utils.higher_regress_metrics: