https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92356
Bug ID: 92356
Summary: Missed optimization of std::find looking for item in
array of items [0..n]
Product: gcc
Version: 9.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: gsfreema at gmail dot com
Target Milestone: ---
Performing a std::find on an array with elements [0, 1, 2,..., n] does not
produce similar optimizations as implementing the find using a raw for loop.
This is true when using clang 9.0 or gcc 9.2 with flags `-O2 -std=c++17`.
Using libc++ with clang produces the same result with std::find as the raw loop
version.
Here is a link to compiler explore demonstrating the difference in generated
assembly.
https://godbolt.org/z/pM1FQQ
One item I just discovered that I did not expect is that a range-for loop in
gcc also has inefficiencies. That might be a separate issue, and I can file
one if needed.
I understand this looks like a contrived example, but I feel like this scenario
does come up sometimes with enums. People will have a large enum list where
they want to check if a value is in a small subset of the enum and that subset
just happens to be the first few items in the enum.
