https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100393
Bug ID: 100393
Summary: Very slow compilation of switch statement with
thousands of cases
Product: gcc
Version: 10.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: curiousdannii at gmail dot com
Target Milestone: ---
I have a switch statement with almost 11,000 cases (produced by a decompiler),
which is taking almost 5 minutes to compile with GCC 10.2. Helpful people on
Stack Overflow suggested that it was a regression, that it only took 3 seconds
to compile with GCC 8.4, and for comparison, only 1.3 seconds in Clang. I
haven't been able to confirm those results yet, but I'll look into doing so.
SO discussion: https://stackoverflow.com/q/67363813/2854284
Preprocessed source:
https://gist.githubusercontent.com/curiousdannii/9476375ff3ae22c403ce2a8132e6a5dc/raw/568f519f2f1b599e98c514f3609a4968d4153eed/functions_unsafe.i
The `-ftime-report -ftime-report-details` results (full results in the SO page)
show that the slow part is in "tree switch lowering".
I also tried compiling it with `-fno-jump-tables`, which, rather than helping,
made it much worse, taking over 33 minutes to compile.
