[llvm-bugs] [Bug 47794] New: missed optimization to use MOVS to copy arrays

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=47794

            Bug ID: 47794
           Summary: missed optimization to use MOVS to copy arrays
           Product: clang
           Version: trunk
          Hardware: All
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: enhex0 at gmail.com
                CC: llvm-bugs at lists.llvm.org, neeilans at live.com,
                    richard-llvm at metafoo.co.uk

For example the following code generates `MOVSQ` with GCC, and `call memcpy`
with Clang:
https://godbolt.org/z/15bPh5
```
#include <iostream>
#include <array>
#include <random>
using namespace std;

random_device rd;

int main()
{
    std::array<char, 512> a;
    std::array<char, 512> b;

    for(auto& e : a) { // prevent optimizing away
        e = rd();
    }

    for(int i=0; i < a.size(); ++i) { // copy
        b[i] = a[i];
    }

    for(auto const& e : b) { // prevent optimizing away
        cout << e;
    }
}
```

It seems that using MOVSQ is about 4-5 times faster:
Clang: https://quick-bench.com/q/LK5SmbOYZc8T-Kvy48jsf1tAL9w
GCC: https://quick-bench.com/q/qG8ApWJV7Ad-ghub2of6C_WWMYM

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20201011/bae3eea4/attachment.html>