diff options
Diffstat (limited to 'dev/perf-analysis')
-rw-r--r-- | dev/perf-analysis | 60 |
1 files changed, 28 insertions, 32 deletions
diff --git a/dev/perf-analysis b/dev/perf-analysis index 4295a573..23259156 100644 --- a/dev/perf-analysis +++ b/dev/perf-analysis @@ -1,51 +1,47 @@ -Performance analysis for V8-0-bugfix branch -------------------------------------------- +Performance analysis for V8-0 branch +------------------------------------ - Dec 27, 2005: contrib Karatsuba added (~ 24s) + Dec 29, 2005: new test and use of -vm in Stalmarck + + Dec 27, 2005: contrib Karatsuba added (~ 30s) + +Dec 28, 2005: size decrease + mainly due to Defined moved to Qed in FSets (reduction from 95M to 7Mo) Dec 1-14, 2005: benchmarking server down + between the two dates: Godel: -10%, CoRN: -10% + probably due to changes around vm (new informative Cast, + change of equality in named_context_val) -Nov 29 and Dec 16, 2005: size increase - due to new record flag in inductive for extraction + Oct 6, 2005: contribs IPC and Tait added (~ 22s and ~ 25s) - Oct 6, 2005: contribs IPC and Tait added (~ 22s and ~ 24s) - - Aug 1, 2005: contrib Kildall added (~ 64s) +Aug 19, 2005: time decrease after application of "Array.length x=0" Xavier's + suggestions for optimisation + (e.g. Nijmegen/QArith: -3%, Nijmegen/CoRN: -7%, Godel: -3%) + Aug 1, 2005: contrib Kildall added (~ 65s) + Jul 26-Aug 2, 2005: bench down Jul 14-15, 2005: 4 contribs failed including CoRN - Jul 7, 2005: adding contrib Fermat4: but not compabible and remove on Jul 8 +Jul 14, 2005: time increase after activation of "closure optimisation" + (e.g. Nijmegen/QArith: +8%, Nijmegen/CoRN: +3%, Godel: +13%) - Jun 17, 2005: contrib Goodstein extended and moved to CantorOrdinals (~ 28s) + Jul 7, 2005: adding contrib Fermat4 -Jun 4, 2005: significant time reduction - (e.g. Nijmegen/LinAlg: -15%, Nijmegen/QArith: stable; Nijmegen/CoRN: -1%) - only changes are the removal of an assert checking location and - the pre-definition of level 200 (could it be just a parsing improvement??) + Jun 17, 2005: contrib Goodstein extended and moved to CantorOrdinals (~ 30s) May 19, 2005: contrib Goodstein and prfx (~ 9s) added -Apr 30, 2005: evaluation order of atomic tactics changed - (e.g. Nijmegen/CoRN: stable, Nijmegen/QArith: -2%, Nijmegen/LinAlg: +20%) +Apr 21, 2005: strange time decrease + (could it be due to the change of Back and Reset mechanism) + (e.g. Nijmegen/CoRN: -2%, Nijmegen/QARITH: -4%, Godel: -11%) Mar 20, 2005: fixed Logic.with_check bug - improved whole V8-0-bugfix bench by 4 % - (e.g. Nijmegen/CoRN: - 7.5 %, Nijmegen/QARITH: - 1.5 %) - -Mar 7-10, 2005: unexplained time reduction - (on Mar 7, changed Ppconstrnew univ printer only) - (note also a server upgrade around Mar 10) - -Feb 17, 2005: fixed omega bug #922 (wrong STATE dependency): - improved whole V8-0-bugfix bench by 2 % - (e.g. Nijmegen/CoRN: - 6.5 %, Nijmegen/QARITH: - 3 %) - -Feb 2, 2005: fixed ltac var interpretation order - - Jan 13, 2005: contrib SumOfTwoSquare added (~ 37s) + global time decrease (e.g. Nijmegen/CoRN: -3%, Nijmegen/QARITH: -1.5%) -Dec 20-29, 2004: reduced whole V8-0-bugfix due to Berkeley/Godel failure +Jan 31-Feb 8, 2005: small instability + (e.g. CoRN: ~2015s -> ~1999s -> ~2032s, Godel: ~340s -> ~370s) -Nov 27 - Dec 10, 2004: strong instability + Jan 13, 2005: contrib SumOfTwoSquare added (~ 38s) |