cppcheck some XML parsing cleanups

Jun 12 '24 11:06 firewave

What speedup, if any, does using strcmp() buy us?

Jun 12 '24 14:06 chrchr-github

What speedup, if any, does using strcmp() buy us?

That was not done because of performance but so the code is aligned with the rest. The redundant calls should have more of an impact and were done for performance purposes.

callgrind in CI 64,844,065,823 -> 64,850,887,712

The increase seems to be caused by strcmp().

Jun 12 '24 17:06 firewave

Even with Clang 19 which includes a fix for https://github.com/llvm/llvm-project/issues/58003 using strcmp() uses more Ir.

Running just the start-up code of testrunner: 1,432,304,160 -> 1,459,503,454

Jun 12 '24 22:06 firewave

Same behavior with GCC.

Jun 12 '24 22:06 firewave

Even with Clang 19 which includes a fix for llvm/llvm-project#58003 using strcmp() uses more Ir.

Ah - turns out the default maximum length which is inlined is 3 characters. It can be controlled via the option strncmp-inline-threshold.

Jun 16 '24 13:06 firewave

I dropped the introduction of strcmp() as it makes things slower.

Sep 18 '24 10:09 firewave