Community
Participate
Working Groups
I compare recompiling the source of the compiler using 1.6 mode and 1.5 mode and the 1.6 mode is around 600ms slower. Adding a static counter to record the time spent in the code generation, I could see that most of the difference is spent in the code generation. This requires investigation.
Maxime, I think I remember that Philippe used some windows bat files to compare batch compiler performances against different VMs. Isn't it something we can use for this?
I do not think this is about using various VMs, is it?
Created attachment 77162 [details] Proposed fix + updated regression tests
Created attachment 77163 [details] Updated local tests Update for ConformTests
This patch includes the fix for bug 182359
Released for 3.4M2. To verify, an improvement should be noticeable in the performance tests if one test is checking the 1.6 code generation.
Verified for 3.4M2