on Linux 2.6.18-164.11.1.el5 #1 SMP Wed Jan 20 12:36:24 CET 2010 x86_64 x86_64 x86_64 GNU/Linux gcc version 4.6.0 20100408 (experimental) (GCC) and GNU C++ (GCC) version 4.5.0 (x86_64-unknown-linux-gnu) #include<cmath> float sqr(float x) { return std::pow(x,2);} produces 0000000000000020 <sqr(float)>: 20: 0f 14 c0 unpcklps %xmm0,%xmm0 23: 0f 5a c0 cvtps2pd %xmm0,%xmm0 26: f2 0f 59 c0 mulsd %xmm0,%xmm0 2a: 66 0f 14 c0 unpcklpd %xmm0,%xmm0 2e: 66 0f 5a c0 cvtpd2ps %xmm0,%xmm0 32: c3 retq 33: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1) 39: 0f 1f 80 00 00 00 00 nopl 0x0(%rax) when compiled with c++ -O2 -std=gnu++0x while c++ -O2 emits (as expected) 0000000000000020 <sqr(float)>: 20: f3 0f 59 c0 mulss %xmm0,%xmm0 24: c3 retq 25: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1) 2a: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1) compilation step details c++ -O2 -std=gnu++0x -c -v assocMath.cpp Using built-in specs. COLLECT_GCC=c++ COLLECT_LTO_WRAPPER=/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/lto-wrapper Target: x86_64-unknown-linux-gnu Configured with: /build/LCG/work/gcc-4.5.0_20100409/configure --prefix=/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt --with-mpfr=/afs/cern.ch/sw/lcg/external/mpfr/2.4.2/x86_64-slc5-gcc43-opt --with-gmp=/afs/cern.ch/sw/lcg/external/gmp/4.3.2/x86_64-slc5-gcc43-opt --with-mpc=/afs/cern.ch/sw/lcg/external/mpc/0.8.1/x86_64-slc5-gcc43-opt --enable-lto --with-libelf=/afs/cern.ch/sw/lcg/external/libelf/0.8.13/x86_64-slc5-gcc43-opt --with-ppl=/afs/cern.ch/sw/lcg/external/ppl/0.10.2/x86_64-slc5-gcc43-opt --with-cloog=/afs/cern.ch/sw/lcg/external/cloog-ppl/0.15.7/x86_64-slc5-gcc43-opt --enable-languages=c,c++,fortran --disable-werror Thread model: posix gcc version 4.6.0 20100408 (experimental) (GCC) COLLECT_GCC_OPTIONS='-O2' '-std=gnu++0x' '-c' '-v' '-shared-libgcc' '-mtune=generic' '-march=x86-64' /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus -quiet -v -D_GNU_SOURCE assocMath.cpp -quiet -dumpbase assocMath.cpp -mtune=generic -march=x86-64 -auxbase assocMath -O2 -std=gnu++0x -version -o /tmp/innocent/ccQs7YW8.s GNU C++ (GCC) version 4.6.0 20100408 (experimental) (x86_64-unknown-linux-gnu) compiled by GNU C version 4.6.0 20100408 (experimental), GMP version 4.3.2, MPFR version 2.4.2, MPC version 0.8.1 GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096 ignoring nonexistent directory "/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../../x86_64-unknown-linux-gnu/include" #include "..." search starts here: #include <...> search starts here: /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../../include/c++/4.6.0 /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../../include/c++/4.6.0/x86_64-unknown-linux-gnu /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../../include/c++/4.6.0/backward /usr/local/include /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/include /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/include /afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/include-fixed /usr/include End of search list. GNU C++ (GCC) version 4.6.0 20100408 (experimental) (x86_64-unknown-linux-gnu) compiled by GNU C version 4.6.0 20100408 (experimental), GMP version 4.3.2, MPFR version 2.4.2, MPC version 0.8.1 GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096 Compiler executable checksum: c599e68b8adba71320a274af3bca4ed6 COLLECT_GCC_OPTIONS='-O2' '-std=gnu++0x' '-c' '-v' '-shared-libgcc' '-mtune=generic' '-march=x86-64' as -V -Qy --64 -o assocMath.o /tmp/innocent/ccQs7YW8.s GNU assembler version 2.17.50.0.6-12.el5 (x86_64-redhat-linux) using BFD version 2.17.50.0.6-12.el5 20061020 COMPILER_PATH=/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/libexec/gcc/x86_64-unknown-linux-gnu/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/ LIBRARY_PATH=/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../../lib64/:/lib/../lib64/:/usr/lib/../lib64/:/afs/cern.ch/sw/lcg/contrib/gcc/4.5.0_20100409/x86_64-slc5-gcc43-opt/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/../../../:/lib/:/usr/lib/ COLLECT_GCC_OPTIONS='-O2' '-std=gnu++0x' '-c' '-v' '-shared-libgcc' '-mtune=generic' '-march=x86-64' same with GNU C++ (GCC) version 4.5.0 (x86_64-unknown-linux-gnu) compiled by GNU C version 4.5.0, GMP version 4.3.2, MPFR version 2.4.2, MPC version 0.8.1 if fixed please backport to 4.5.x
There is nothing to fix here, see: http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#550
Subject: Re: std::pow(float) converts to double when compiled with -std=gnu++0x Ciao Paolo, On 5 Sep, 2010, at 2:42 PM, paolo dot carlini at oracle dot com wrote: > > > ------- Comment #1 from paolo dot carlini at oracle dot com 2010-09-05 12:42 ------- > There is nothing to fix here, see: > > http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#550 > > 1) this is VERY unfortunate 2) and then why std::exp(float) invokes expf? (a well as all other transcendental and trigonometric functions that invoke xyzf) also std::sqrt converts to sqrtss ? is this not a bit inconsistent? vincenzo > -- > > paolo dot carlini at oracle dot com changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > Status|UNCONFIRMED |RESOLVED > Resolution| |INVALID > > > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45542 > > ------- You are receiving this mail because: ------- > You reported the bug, or are watching the reporter.
The issue affects only mixed mode arithmetic (thus, functions taking at least two arguments), and in that case, as Howard explained, C++0x does what Fortran and C do. In any case, we are implementing correctly the FCD, for sure.
By the way, if a function taking a single argument is passed and integer, the return type is double, not float or long double and one can see that the underlying mechanism is the same. All in all, I agreed with the resolution suggested by Howard, at the time, and frankly I don't thick the DR should be re-opened. CERN should send somebody to the ISO Meetings, like Fermilab does...
Subject: Re: std::pow(float) converts to double when compiled with -std=gnu++0x On 5 Sep, 2010, at 6:15 PM, paolo dot carlini at oracle dot com wrote: > > > ------- Comment #4 from paolo dot carlini at oracle dot com 2010-09-05 16:15 ------- > By the way, if a function taking a single argument is passed and integer, the > return type is double, not float or long double and one can see that the > underlying mechanism is the same. All in all, I agreed with the resolution > suggested by Howard, at the time, and frankly I don't thick the DR should be > re-opened. CERN should send somebody to the ISO Meetings, like Fermilab does… After reading the standard a couple of times, I fully agree that gcc implements the standard as written including the last resolution to remove std::(T,int) (this last bit escaped my first reading) I've discovered that std::pow(float,float) (called for instance as std::pow(x,2.f)) behaves as "I" expect. The fact that CERN is not represented in ISO is an old story (we are NOT affiliated to any national committee) and we rely on Fermilab representatives. In any case this seems to be a C99 issue and my understanding is that C++ never overrules C. The conclusion is that we will have now to recommend to use pow(float,float) not pow(float, int) if float behavior is desired… vincenzo >
Ok... We can discuss these issues in better detail when we met. Well, remember that this is Free Software, thus, if you are unsure about a behavior, just open the header in an editor and look inside it: isn't only allowed, is encouraged!