http://gwt-code-reviews.appspot.com/1277801/diff/5001/6001 File plugins/npapi/NPVariantWrapper.h (right):
http://gwt-code-reviews.appspot.com/1277801/diff/5001/6001#newcode105 plugins/npapi/NPVariantWrapper.h:105: if (isInt(variant)) { On 2011/01/12 00:43:50, fabiomfv wrote:
my concern with this is something I chatted with Jat a couple of weeks
ago. this
will be called many times over and may affect perf. could we consider using modf() that is presumably optimized and it
would be a
one op to extract and check the integer part of a double.
i actually had a modf variant of this working locally, but I opted to change it to this to err on the side of caution since modf is still giving us a double that needs to be downcasted to an int and may be out of int range. just ran a quick test showing that adding modf into the mix slows things down: #include <stdlib.h> #include <stdio.h> #include <math.h> double d1 = 2141243; int main(int c, char **argv) { int count = 0; int x = 0; for (int i = 0; i < 100000000; i++) { #ifdef MODF double i; if (modf(d1, &i) == 0.0) { count++; } x = static_cast<int>(d1); #else x = static_cast<int>(d1); if (d1 == static_cast<double>(x)) { count++; } #endif } printf("done %d\n", count); } ... $ g++ -o modf -O0 modf.c && time ./modf done 100000000 real 0m0.363s user 0m0.354s sys 0m0.004s $ g++ -o modf -O0 -DMODF modf.c && time ./modf done 100000000 real 0m1.273s user 0m1.219s sys 0m0.012s http://gwt-code-reviews.appspot.com/1277801/diff/5001/6002 File plugins/npapi/main.cpp (right): http://gwt-code-reviews.appspot.com/1277801/diff/5001/6002#newcode241 plugins/npapi/main.cpp:241: //Debug::log(Debug::Spam) << "NPP_HandleEvent(instance=" << instance << ")" << Debug::flush; On 2011/01/12 00:43:50, fabiomfv wrote:
should this change be reverted?
as I said earlier, it gets called a ton on the Mac build and makes debugging a pain. (while working on this I needed Spam level verbosity to trace what was going on, but having this function fill up my terminal buffer was not helpful) http://gwt-code-reviews.appspot.com/1277801/show -- http://groups.google.com/group/Google-Web-Toolkit-Contributors