Antonio Stano

I don't think it makes the code slower since -O3 and -Ofast are enabled and only -ffast-math is disabled, I don't know what the contribution of -ffast-math is in terms...

yes, -Ofast with -ffast-math enabled seems to be the best option, but looks like only works for macos, -Ofast -fno-fast-math is a compromise but should work for everyone.

@DongbinNie i can confirm that perfs are 2x with -fno-finite-math-only instead of -fno-fast-math, thanks for the help! I was doing kinda the same thing testing the opt flags enabled by...

@karpathy have you tried compiling with `-fno-finite-math-only` instead of `-fno-fast-math`?

yes @karpathy, -fno-finite-math-only instead of -fno-fast-math for almost 2x improvements

I went from ~17 seconds per step with -fno-fast-math, to this with -fno-finite-math-only & -march=native: ```bash step 0: train loss 5.356086 (took 9611.016869 ms) step 1: train loss 4.300644 (took...

i close this to open a new one synced

@chadbrewbaker does CFLAGS = -O3 -Ofast works togheter on your machine? i got this problem 08e618e73f13544f24122196fa1890a39868f94f, and needed to use -O3 alone, or disable the -ffast-math flag which was called...

Can I have more informations about this position? Il giorno mar 20 feb 2024 alle ore 01:45 Ralf Rentschler < ***@***.***> ha scritto: > Hello, > > We have an...

what is this? Il giorno mar 20 feb 2024 alle ore 12:23 Antonio Stano < ***@***.***> ha scritto: > Can I have more informations about this position? > > Il...