Something that I think a lot of programming youtubers don't do is show an immediate example. a major reason for me to become interested in a programming concept would me being able to see in a very concise, quick and practical way how awesome this one thing actually is, and why I absolutely should want to actually be using it, almost to the point where it seems dumb NOT to be using it. Someone that has maybe heard about link time and might know what it is might come across this video and seek to learn more, but someone that has never ever heard of it might not know what it actually means. A good example is something like pointers. if you show people how they are practically used people suddenly think "oh fuck yeah, that's actually really cool" rather than sticking to what they already know. Not saying this isn't what you do. but it's something that could make me incredibly interested in concepts immediately off the bat.
While it's a good demo, your rationale for the vectorization not happening (In the multiple case) sounds incorrect. It looks like compiler is not able to inline the function call, which prevents vectorization.
Something that I think a lot of programming youtubers don't do is show an immediate example. a major reason for me to become interested in a programming concept would me being able to see in a very concise, quick and practical way how awesome this one thing actually is, and why I absolutely should want to actually be using it, almost to the point where it seems dumb NOT to be using it.
Someone that has maybe heard about link time and might know what it is might come across this video and seek to learn more, but someone that has never ever heard of it might not know what it actually means.
A good example is something like pointers. if you show people how they are practically used people suddenly think "oh fuck yeah, that's actually really cool" rather than sticking to what they already know.
Not saying this isn't what you do. but it's something that could make me incredibly interested in concepts immediately off the bat.
Would using the restrict keywork on the pointers of the base_mmul function lead to the same optimization?
Super helpful, thanks!
While it's a good demo, your rationale for the vectorization not happening (In the multiple case) sounds incorrect.
It looks like compiler is not able to inline the function call, which prevents vectorization.