doing some benchmarks on attached file (preprocessed source), i found, that
gcc-4.4 is somehow (in my case about 4%) slower than gcc-4.3 on x86_64, tuned
for core2:
code compiled with -O3 -march=core2
versions:
g++-4.3 -v
Using built-in specs.
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu 4.3.2-0ubuntu3'
--with-bugurl=file:///usr/share/doc/gcc-4.3/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --enable-shared
--with-system-zlib --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --enable-nls --with-gxx-include-dir=/usr/include/c++/4.3
--program-suffix=-4.3 --enable-clocale=gnu --enable-libstdcxx-debug
--enable-objc-gc --enable-mpfr --enable-checking=release
--build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 4.3.2 (Ubuntu 4.3.2-0ubuntu3)
g++-4.4 -v
Using built-in specs.
Target: x86_64-linux-gnu
Configured with: ../gcc-4.4-20080815/configure
--enable-languages=c,c++,fortran,objc,obj-c++ --enable-shared
--with-system-zlib --enable-mpfr --enable-checking=release
--build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
--without-included-gettext --enable-threads=posix --enable-nls
--with-gxx-include-dir=/usr/local/include/c++/4.4 --program-suffix=-4.4
--enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc
Thread model: posix
gcc version 4.4.0 20080815 (experimental) (GCC)
gcc-4.3 produces:
0000000000401040 <loop(nova::biquad<float, float, false, true>&, float*,
float*, int)>:
401040: ff c9 dec %ecx
401042: f3 0f 10 6f 18 movss 0x18(%rdi),%xmm5
401047: f3 0f 10 67 14 movss 0x14(%rdi),%xmm4
40104c: 31 c0 xor %eax,%eax
40104e: f3 0f 10 35 6a 0d 00 movss 0xd6a(%rip),%xmm6 #
401dc0 <boost::array<float, 3ul>::operator[](unsigned
long)::__PRETTY_FUNCTION__+0x60>
401055: 00
401056: 48 8d 0c 8d 04 00 00 lea 0x4(,%rcx,4),%rcx
40105d: 00
40105e: 66 90 xchg %ax,%ax
while gcc-4.4 produces:
0000000000400fe0 <loop(nova::biquad<float, float, false, true>&, float*,
float*, int)>:
400fe0: 49 89 d0 mov %rdx,%r8
400fe3: f3 0f 10 67 14 movss 0x14(%rdi),%xmm4
400fe8: 8b 47 18 mov 0x18(%rdi),%eax
400feb: 66 0f 7e e2 movd %xmm4,%edx
400fef: 48 c1 e0 20 shl $0x20,%rax
400ff3: 89 d2 mov %edx,%edx
400ff5: ff c9 dec %ecx
400ff7: 48 09 d0 or %rdx,%rax
400ffa: f3 0f 10 35 3e 0d 00 movss 0xd3e(%rip),%xmm6 #
401d40 <boost::array<float, 3ul>::operator[](unsigned
long)::__PRETTY_FUNCTION__+0x60>
401001: 00
401002: 48 c1 e8 20 shr $0x20,%rax
401006: 48 8d 14 8d 04 00 00 lea 0x4(,%rcx,4),%rdx
40100d: 00
40100e: 66 0f 6e e8 movd %eax,%xmm5
401012: 31 c0 xor %eax,%eax
the rest of the code is equivalent ...
i am not really familiar with x86_64 assembly, but
mov 0x18(%rdi),%eax
movd %eax,%xmm5
has been realizied by gcc-4.3 as
movss 0x18(%rdi),%xmm5
and for the other code, registers seem to be allocated and reused in a more
efficient way ...
--
Summary: [4.4 regression] speed regression
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: tim at klingt dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37437
