Open smedilol opened 5 years ago
After investigation there are at least 3 areas that can be improved. I've updated the first post of the issue to detail this.
A intermediate status of the performance gain so far PerfStatus.xlsx
Except for a performance regression due to evolution of v1.3, we can see that they are major gain.
There is no more big performance impact when you use type
over plain Cobol 85
10/03/2019 | 30/03/2019 | |||
Test | e251ec0badd0536c80cc2f05072a9e1ca807c0b3 (time in ms) | TypeCobolLinker (time in ms) | Gain | |
Part1_FullParsing_Cobol85_NoRedefines | 2053 | 2123 | -3% | |
Part1_FullParsing_TC_BigTypesNoProcedure | 2990 | 1995 | 33% | |
Part1_FullParsing_TC_BigTypesWithProcedure | 2455 | 1996 | 19% | |
Part1_FullParsing_TC_GlobalStorage | 2255 | 1908 | 15% | |
Part1_Incremental_Cobol85_NoRedefines | 525 | 323 | 38% | |
Part1_Incremental_TC_BigTypesNoProcedure | 1288 | 331 | 74% | |
Part1_Incremental_TC_BigTypesWithProcedure | 742 | 402 | 46% | |
Part1_Incremental_TC_GlobaStorage | 743 | 401 | 46% | |
Part2_FullParsing_TC_UseALotOfTypes_001Time | 692 | 535 | 23% | |
Part2_FullParsing_TC_UseALotOfTypes_100Times | 2497 | 565 | 77% | |
Part2_FullParsing_TC_UseALotOfTypes_WithProc_100Times | 1435 | 586 | 59% | |
Part2_Incremental_TC_UseALotOfTypes_001Time | 161 | 64 | 60% | |
Part2_Incremental_TC_UseALotOfTypes_100Times | 1972 | 100 | 95% | |
Part2_Incremental_TC_UseALotOfTypes_WithProc_100Times | 926 | 95 | 90% | |
Part3_FullParsing_Cobol85_DeepVariables | 1967 | 1388 | 29% | |
Part3_FullParsing_TC_DeepTypes | 4078 | 1621 | 60% | |
Part3_Incremental_Cobol85_DeepVariables | 853 | 251 | 71% | |
Part3_Incremental_TC_DeepTypes | 3010 | 417 | 86% |
The parser has been optimized recently but there is still room for improvement.
This issue is a general issue about optimization.
After investigation these areas that can be improved:
More details on each issue.