Considering that "complex" turns basically any basic type into a vector type,
complex number addition and subtraction could utilize SSE instructions to
perform the operation on real and imaginary parts simultaneously. (Only applies
to addition and subtraction.)
Code:
#include <complex.h>
typedef float complex ss1;
typedef float ss2 __attribute__((vector_size(sizeof(ss1))));
ss1 add1(ss1 a, ss1 b) { return a + b; }
ss2 add2(ss2 a, ss2 b) { return a + b; }
Produces:
add1:
movq %xmm0, -8(%rsp)
movq %xmm1, -16(%rsp)
movss -4(%rsp), %xmm0
movss -8(%rsp), %xmm1
addss -12(%rsp), %xmm0
addss -16(%rsp), %xmm1
movss %xmm0, -20(%rsp)
movss %xmm1, -24(%rsp)
movq -24(%rsp), %xmm0
ret
add2:
movlps %xmm0, -16(%rsp)
movlps %xmm1, -24(%rsp)
movaps -24(%rsp), %xmm0
addps -16(%rsp), %xmm0
movaps %xmm0, -56(%rsp)
movlps -56(%rsp), %xmm0
ret
Command line:
gcc -msse -O3 -S test2.c
(Results are same with -ffast-math)
Architecture:
CPU=AMD Athlon(tm) 64 X2 Dual Core Processor 4600+
CPU features=fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt lm 3dnowext 3dnow
pni lahf_lm cmp_legacy
GCC is:
Target: x86_64-linux-gnu
Configured with: ../src/configure -v
--enable-languages=c,c++,fortran,objc,obj-c++,treelang --prefix=/usr
--enable-shared --with-system-zlib --libexecdir=/usr/lib
--without-included-gettext --enable-threads=posix --enable-nls
--program-suffix=-4.1 --enable-__cxa_atexit --enable-clocale=gnu
--enable-libstdcxx-debug --enable-mpfr --enable-checking=release
x86_64-linux-gnu
Thread model: posix
gcc version 4.1.2 20061115 (prerelease) (Debian 4.1.1-21)
--
Summary: C complex numbers, amd64 SSE, missed optimization
opportunity
Product: gcc
Version: 4.1.2
Status: UNCONFIRMED
Severity: enhancement
Priority: P3
Component: rtl-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: bisqwit at iki dot fi
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31485
