From https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
It says "-ftree-slp-vectorize: Perform basic block vectorization on trees. This flag is enabled by default at -O2 and by -ftree-vectorize, -fprofile-use, and -fauto-profile."
However it seems I have to pass a flag explicitly to turn on SIMD. Did I mis undertand something here? It is enabled at -O3 though.
Is
-ftree-slp-vectorizenot enabled by-O2in GCC?
Yes and no. It depends on the version of the compiler.
From https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
You have linked to the latest version of documentation. It applies to the version that is currently under development, which at the moment is version 12.
However it seems I have to pass a flag explicitly to turn on SIMD.
Your example uses GCC version 11.
Did I mis undertand something here?
You read the wrong version of documentation, or used wrong version of compiler and hence your assumption didn't hold.
latest or current in the url so I didn't know the doc is versioned. But thanks to you I looked into it and found https://gcc.gnu.org/onlinedocs/gcc-11.2.0/gcc/Optimize-Options.html is what I am looking for
-O3is "too aggressive" and/or often not faster. Clang enables auto-vectorization at-O2, so GCC before 12 looked bad by comparison if people benchmark at-O2. Simple loops on modern CPUs benefit significantly from SIMD.