As suggested by @laanwj on IRC:
<cfields> whoa
<cfields> Leaving test case "knapsack_solver_test"; testing time: 358694ms
<cfields> i386 + old wine ^^
<cfields> Leaving test case "knapsack_solver_test"; testing time: 6781ms
<cfields> ^^ same, but with the LogPrint commented out
...
<wumpus> if both log-to-file and log-to-console is disabled, it should probably bypass all logging
Edit: The painful line commented out being the LogPrintf in CWallet::AddToWallet.
utACK 8ca1f6ecba5efc278927fd34eb04f6a4254dfe03
Wine i686 build down to 11 min :)
utACK 8ca1f6e (Run time seems to be about the same or slightly faster than with the wine-hq)
While this is an easy win for tests, we should still track down why LogTimestampStr() is so painful.
While this is an easy win for tests, we should still track down why LogTimestampStr() is so painful.
It looks inefficient too:
std::string DateTimeStrFormat(const char* pszFormat, int64_t nTime)
{
static std::locale classic(std::locale::classic());
// std::locale takes ownership of the pointer
std::locale loc(classic, new boost::posix_time::time_facet(pszFormat));
std::stringstream ss;
ss.imbue(loc);
ss << boost::posix_time::from_time_t(nTime);
return ss.str();
}
338 | @@ -339,6 +339,10 @@ int LogPrintStr(const std::string &str) 339 | int ret = 0; // Returns total number of characters written 340 | static std::atomic_bool fStartedNewLine(true); 341 | 342 | + if (!fPrintToConsole && !fPrintToDebugLog) {
nit: might be even more efficient to do this check in util.h already, before calling tfm::format.
Good point, done.
This leads to massive speedups under Wine.
utACK 339730a
utACK 339730a6d84a3a90ae23d732fbfc1c81af88d435
utACK 339730a
utACK 339730a6d84a3a90ae23d732fbfc1c81af88d435
Impressive! Very nice find!