Replies: 1 comment
-
We partially have it (a lot of constants are in types.h). |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
My question is: what do we have now: two assemblies from the same source in a row will have a different hash, i.e. they are not binary equal. This means that there is a very high probability that one of them will be 0.1 - 1% faster than the other. Again, it is quite possible to measure it, usually the difference is small, but sometimes it can be around that same 1%.
What follows from this:
If we run the tournament master vs patch, then if the "patch" is a little faster, simply due to the randomness of the compiler, then the patch can gain a certain plus, having in fact a minus. And this code will pass!
But if you put the variables / constants into a file, then the patches that change them (but not the code itself) can be run on the same binar, getting greater accuracy.
Beta Was this translation helpful? Give feedback.
All reactions