 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Mac Forum / Programming / Mac Programming / December 2005float.h
I have printed the values LDBL_MAX_10_EXP LDBL_MIN_10_EXP and i have obtained 308 and -291 while DBL_MAX_10_EXP DBL_MIN_10_EXP are 308 and -307. Shouldn't the values for long double be much bigger or at least equal to those of double? Furthermore how is it possible that LDBL_EPSILON equals 4.940656e-324 if LDBL_MIN_10_EXP is -291? What's wrong? I'm using OS X 10.4.3 on a powerbook g4. Thanks. > I have printed the values LDBL_MAX_10_EXP LDBL_MIN_10_EXP and i have > obtained 308 and -291 while DBL_MAX_10_EXP DBL_MIN_10_EXP are 308 and [quoted text clipped - 3 lines] > What's wrong? I'm using OS X 10.4.3 on a powerbook g4. > Thanks. I don't know where the -291 is coming from - if you look at float.h you should see LDBL_MIN_10_EXP defined as -307. BTW, it probably helps to mention which compiler you're using (gcc 3.3, gcc 4.0, CodeWarrior, whatever). Paul >> I have printed the values LDBL_MAX_10_EXP LDBL_MIN_10_EXP and i have >> obtained 308 and -291 while DBL_MAX_10_EXP DBL_MIN_10_EXP are 308 and [quoted text clipped - 11 lines] > > Paul I'm using gcc 4.0. By looking inside float.h i only see the following: #define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ However even if it would be -307 what would be the meaning of LDBL_EPSILON equals 4.940656e-324? > I'm using gcc 4.0. By looking inside float.h i only see the following: > #define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ Right, and you should see in the same header that __LDBL_MIN_10_EXP__ is #defined as (-307). > However even if it would be -307 what would be the meaning of > LDBL_EPSILON equals 4.940656e-324? I'm not 100% certain but I would guess that 1e-307 is the smallest normalised float, while 4.940656e-324 is either the smallest denormal or the size of one LSB in the smallest float ? Paul > > I'm using gcc 4.0. By looking inside float.h i only see the following: > > #define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ > > Right, and you should see in the same header that __LDBL_MIN_10_EXP__ is > #defined as (-307). I think it's actually built in to the compiler: PowerBookG412:~> cpp -E -dM < /dev/null | grep __LDBL_MIN_10_EXP__ #define __LDBL_MIN_10_EXP__ (-291) -Eric  SignatureEric Albert ejalbert@cs.stanford.edu http://outofcheese.org/ >>>I'm using gcc 4.0. By looking inside float.h i only see the following: >>>#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ [quoted text clipped - 8 lines] > > -Eric Interesting - float.h has #ifndef __LDBL_MIN_10_EXP__ #define __LDBL_MIN_10_EXP__ (-307) #endif so I guess the compiler is overriding float.h. Paul > >>>I'm using gcc 4.0. By looking inside float.h i only see the following: > >>>#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ [quoted text clipped - 6 lines] > > PowerBookG412:~> cpp -E -dM < /dev/null | grep __LDBL_MIN_10_EXP__ > > #define __LDBL_MIN_10_EXP__ (-291) > Interesting - float.h has > [quoted text clipped - 3 lines] > > so I guess the compiler is overriding float.h. Look a bit closer and you'll see that's in an #if __MWERKS__ block. It's only there for CodeWarrior support. -Eric SignatureEric Albert ejalbert@cs.stanford.edu http://outofcheese.org/ >>>>>I'm using gcc 4.0. By looking inside float.h i only see the following: >>>>>#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ [quoted text clipped - 17 lines] > Look a bit closer and you'll see that's in an #if __MWERKS__ block. > It's only there for CodeWarrior support. D'oh ! Interestingly though gcc 3.3 seems to give the correct value: > prussell$ gcc-3.3 -E -dM - < /dev/null | grep __LDBL_MIN_10_EXP__ > #define __LDBL_MIN_10_EXP__ (-307) Paul > >>>>>I'm using gcc 4.0. By looking inside float.h i only see the following: > >>>>>#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ [quoted text clipped - 24 lines] > > prussell$ gcc-3.3 -E -dM - < /dev/null | grep __LDBL_MIN_10_EXP__ > > #define __LDBL_MIN_10_EXP__ (-307) I wonder if the different value is due to the different default size of long double between 3.3 and 4.0. Long double technically wasn't supported on Mac OS X until 128-bit long doubles were introduced with gcc 4 and Tiger. Each compiler may be correct for its long double size. This example shows, well, something: PowerBookG412:~> cat test.c #include <stdio.h> int main(void) { printf("Size of long double: %lu\n", sizeof(long double)); return 0; } PowerBookG412:~> gcc-3.3 -Wall -g -o test test.c test.c: In function `main': test.c:4: warning: use of `long double' type; its size may change in a future release test.c:4: warning: (Long double usage is reported only once for each file. test.c:4: warning: To disable this warning, use -Wno-long-double.) PowerBookG412:~> ./test Size of long double: 8 PowerBookG412:~> gcc-4.0 -Wall -g -o test test.c PowerBookG412:~> ./test Size of long double: 16 PowerBookG412:~> -Eric SignatureEric Albert ejalbert@cs.stanford.edu http://outofcheese.org/ >>>>>>>I'm using gcc 4.0. By looking inside float.h i only see the following: >>>>>>>#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__ [quoted text clipped - 52 lines] > Size of long double: 16 > PowerBookG412:~> I can only guess, than, that some of the negative exponent range in long double has been stolen for NANs and other special cases, and the change in gcc 4.0 reflects this ? Paul |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2008 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.