Hello,
I noted that the optimizations turned on by O3 was outdated in the
documentation. Here is the discussion that brought it to my attention:
https://gcc.gnu.org/pipermail/gcc/2020-April/232164.html
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index e2bc2559218..a28001a5adf 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -9241,14 +9241,14 @@ by @option{-O2} and also turns on the following
optimization flags:
-floop-unroll-and-jam @gol
-fpeel-loops @gol
-fpredictive-commoning @gol
+-fsplit-loops @gol
-fsplit-paths @gol
--ftree-loop-distribute-patterns @gol
-ftree-loop-distribution @gol
-ftree-loop-vectorize @gol
-ftree-partial-pre @gol
-ftree-slp-vectorize @gol
-funswitch-loops @gol
--fvect-cost-model @gol
+-fvect-cost-model=dynamic @gol
-fversion-loops-for-strides}
@item -O0
