Author: Gerd Isenberg
Date: 08:41:39 09/26/03
Go up one level in this thread
On September 26, 2003 at 07:59:42, Bo Persson wrote: >On September 26, 2003 at 06:50:59, Gerd Isenberg wrote: > >>Hi All, >> >>in C++ we have the boolean type "bool" with the value range true/false. >>I'm not sure about ANSI-C. > >There is a _Bool type in C99. If you include bool.h you get the usual typedefs >etc for bool, true, and false. > >> >>Due to some C-related "portability" problems and possible performance drawbacks >>due sizeof(bool) == 1 implies partial register handling or zero extending to >>native word lenght, you find in most MS-sources an "own" boolean type BOOL: >> >>typedef int BOOL; // e.g. WINDEF.H > >The reason for this is not performance, but that Windows is older than standard >C and C++. There was no bool type back then. I see, but performance is an (negligible?) issue here with partial registers. Pushing actual bool parameters on the stack requires zero extensions - or you have other missaligned 32-bit values. IIRC it was a bit more than the "random" noise improvement in IsiChess, when i switched from bool to BOOL. I use a lot of boolean inlines like isKing()... > >You don't really have to extend the value to test for zero/non-zero. If you have >several bools in a struct, the smaller size could be good too. Very seldom, instead of more (than eight) bools in a structure or an array of bools i prefere array of bits, but hide the implementation with boolean inliners. > >> >>This own BOOL type is of course not "typesafe" as bool, as you may assign any >>"int" expressions to it. With BOOL one should interprete zero as FALSE and any >>other value as TRUE. Due to this ambiguity, comparing BOOL-expressions with TRUE >>may be erroneous, so better compare with != FALSE. > >There are also potential overload problems. Yes, int and BOOL have the same signature, like #define BOOL int > >> >>See also this bugreport related to this issue: >> >>http://www.codeproject.com/buglist/virtualboolbug.asp >> >>Actually i have some "disputes" with colleagues obout it. >>I found bool better for didactical reasons, >>but stay with BOOL for pragmatical reasons. >> >>A few questions: >> >>Is sizeof(bool) == 1 per definition, >>or is it compiler implementation depending? > >That is up to the compiler. > > >> >>Is there any conditional preprocessor directive to ask whether a user defined >>type is already typedefined, similar to #ifdef? > >No. > >> >>What is your opinion / experience with bool versus BOOL? > >I use bool as much as possible. BOOL is just for interfacing with Windows. > The most correct way - but unfortunately not the fastest. Finally i will use a conditional compiled types: #ifdef QUICK_AND_DIRTY #define MyBool bool #else #define MyBool BOOL #endif Gerd > >> >>Thanks in advance for your suggestions, >>Gerd > > >Bo Persson
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.