Remove an unnecessary null pointer check about filament HOT 6 CLOSED

Indeed, thanks for the report!

from filament.

Would you like to check if any similar update candidates were left over in the source code for other software components?

from filament.

I'm sure there are other instances. I'd love to know about them if it's not too much work for you to look for them. Otherwise I'll take a look myself.

from filament.

How do you think about to recheck the need for null pointer checks before calls of resource release functions (like “free”) with the help of the semantic patch language (for example)?

from filament.

@elfring Note that there are a few places in filament where we intentionally check for nullptr before free() for performance reason. This is because a call to free() is expensive -- .so calls are actually pretty heavy, they cost the equivalent of a jump + global variable dereference + function pointer branch. So in case where the call to free() is NOT the common case, it's preferable to check for nullptr.

from filament.

… for performance reason.

Would you like to point any corresponding source code places out?

This is because a call to free() is expensive …

• Did you measure such software costs?
• Do extra pointer checks influence the run time characteristics really in desired ways?

So in case where the call to free() is NOT the common case, it's preferable to check for nullptr.

Do you care for the concrete function call frequency where null pointers will be passed actually?

🔮 How do you think about to update (bundled) software libraries any further?

from filament.