[Bug tree-optimization/71775] New: Redundant move instruction for sign extension

[wmhkebe at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71775

            Bug ID: 71775
           Summary: Redundant move instruction for sign extension
           Product: gcc
           Version: 6.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: wmhkebe at gmail dot com
  Target Milestone: ---

Created attachment 38838
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38838&action=edit
preprocessed file for the test program

Compiling the following program with g++ -O3 -S:

#include <cstdint>

unsigned long long f(std::uint64_t x) {
  std::uint64_t ans = 0;
  for (; x != 0; x &= (x - 1)) {
    ans ^= __builtin_ctzll(x);
  }
  return ans;
}

yields the following assembly code:

        .text
        .align 4,0x90
        .globl __Z1fy
__Z1fy:
LFB0:
        xorl    %eax, %eax
        testq   %rdi, %rdi
        je      L4
        .align 4,0x90
L3:
        bsfq    %rdi, %rdx
        movslq  %edx, %rdx
        xorq    %rdx, %rax
        leaq    -1(%rdi), %rdx
        andq    %rdx, %rdi
        jne     L3
        ret
L4:
        ret

Note that the movslq instruction inside the loop is redundant. Clang is able to
generate code without this move. Benchmark shows, this represents a 5%-10%
difference in the running time of this function.


Version info:
Using built-in specs.
COLLECT_GCC=gcc-6
COLLECT_LTO_WRAPPER=/usr/local/Cellar/gcc6/6.1.0/libexec/gcc/x86_64-apple-darwin15.5.0/6.1.0/lto-wrapper
Target: x86_64-apple-darwin15.5.0
Configured with: ../configure --build=x86_64-apple-darwin15.5.0
--prefix=/usr/local/Cellar/gcc6/6.1.0
--libdir=/usr/local/Cellar/gcc6/6.1.0/lib/gcc/6
--enable-languages=c,c++,objc,obj-c++,fortran --program-suffix=-6
--with-gmp=/usr/local/opt/gmp --with-mpfr=/usr/local/opt/mpfr
--with-mpc=/usr/local/opt/libmpc --with-isl=/usr/local/opt/isl014
--with-system-zlib --enable-libstdcxx-time=yes --enable-stage1-checking
--enable-checking=release --enable-lto --with-build-config=bootstrap-debug
--disable-werror --with-pkgversion='Homebrew gcc6 6.1.0'
--with-bugurl=https://github.com/Homebrew/homebrew-versions/issues
--enable-plugin --disable-nls --enable-multilib
--with-native-system-header-dir=/usr/include
--with-sysroot=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.11.sdk
Thread model: posix
gcc version 6.1.0 (Homebrew gcc6 6.1.0)