| Description | Reading a large 1GB STEP file shows enormous difference in time on different platforms (Windows 10 (msvc14), Ubuntu 18.04 (gcc7.5), macOS 10.14 (CLang11)):
|===========================================|
| | AMD Ryzen 3900X | Intel Core i5 |
|---------|-----------------|---------------|
| msvc14 | 19:13 | 16:30 |
| gcc7.5 | 06:13 | 14:54 |
| CLang11 | N/A | 07:31 |
|===========================================|
gcc7.5 is 3x times faster than msvc14 on Ryzen 3900X (but not that much on Intel Core i5)
and CLang11 is 2x times faster than gcc7.5 and msvc14 on Intel Core i5.
Such a large difference (and extremely long STEP import time in case of msvc14 outsider) looks suspicious.
It is proposed to:
- Collect more data samples to verify if deviation is general or specific to one / few STEP files.
- Collect more statistics, including Visual Studio 2019, GCC 10 and CLang (Linux/Windows) to determine if difference is caused only by compiler or also by platform (dubious).
- Localize hot-spots in the algorithm showing most noticeable performance deviations and analyze options which would reduce the difference. |
Relationships 
Relationships 