Fails to vectorize when compiled with g++ -ftree-vectorizer-verbose=7 -S -O3 -march=core2 -msse3 -mfpmath=sse -Wall -o t test.c
$ gcc -v Using built-in specs. Target: x86_64-redhat-linux Configured with: ../configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla --enable-bootstrap --enable-shared --enable-threads=posix --enable-checking=release --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-languages=c,c++,objc,obj-c++,java,fortran,ada --enable-java-awt=gtk --disable-dssi --enable-plugin --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-1.5.0.0/jre --enable-libgcj-multifile --enable-java-maintainer-mode --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --disable-libjava-multilib --with-ppl --with-cloog --with-tune=generic --with-arch_32=i586 --build=x86_64-redhat-linux Thread model: posix gcc version 4.4.0 20090506 (Red Hat 4.4.0-4) (GCC) unsigned char cr[4096] __attribute((aligned(16))); unsigned char cg[4096] __attribute((aligned(16))); unsigned char cb[4096] __attribute((aligned(16))); unsigned char pr[4096] __attribute((aligned(16))); unsigned char pg[4096] __attribute((aligned(16))); unsigned char pb[4096] __attribute((aligned(16))); unsigned short noise[4096] __attribute((aligned(16))); unsigned char out[4096] __attribute((aligned(16))); int main(int argc, char * argv[]) { int i; unsigned char * __restrict pcr = cr; unsigned char * __restrict pcg = cg; unsigned char * __restrict pcb = cb; unsigned char * __restrict ppr = pr; unsigned char * __restrict ppg = pg; unsigned char * __restrict ppb = pb; unsigned short * __restrict n = noise; unsigned char * __restrict o = out; //#define VECTORIZE 1 #define FAILVECTORIZE for (i = 1024;i > 0; --i) { unsigned short r0, r1, r2; r0 = pcr[0] - ppr[0]; r1 = pcg[0] - ppg[0]; r2 = pcb[0] - ppb[0]; #ifdef VECTORIZE // This successfully vectorizes. unsigned short a0, a1, a2, b0, b1, b2; a0 = (r0 >> 8); b0 = a0 | 0xff00; r0 &= b0; a1 = (r1 >> 8); b1 = a1 | 0xff00; r1 &= b1; a2 = (r2 >> 8); b2 = a2 | 0xff00; r2 &= a2; #endif #ifdef FAILVECTORIZE // This fails to vectorize, even though it apparently // does exactly the same thing. r0 &= (r0>>8) | 0xff00; r1 &= (r1>>8) | 0xff00; r2 &= (r2>>8) | 0xff00; #endif unsigned short r = r0 + r1 + r2; *n = *n + r - (*n>>5); *o = r; ++n; ++o; ++pcr; ++pcg; ++pcb; ++ppr; ++ppg; ++ppb; } } -- Summary: Code fails to vectorize in g++ but does vectorize in gcc. Product: gcc Version: unknown Status: UNCONFIRMED Severity: normal Priority: P3 Component: c++ AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: public at dgmo dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40927
