C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  function pointer Goto page Previous  1, 2, 3 ... , 11, 12, 13  Next          C++Talk.NET Forum Index -> C Language (Moderated) View previous topic :: View next topic   Author Message Thomas Richter Guest Posted: Mon Apr 23, 2012 1:31 pm    Post subject: Re: which one is faster? On 05.04.2012 01:05, echo ma wrote: Quote: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? Sorry, I don't understand. To have something "the same", you need to compare two different things. So which two "are the same" or "not the same"? I see only one piece of code, and this is not even portable. (#pragma pack() is not ANSI-C). Anyhow, before performing micro-optimizations like the one you made above, profile your code and identify whether this is really the cause of a performance (or memory) problem. You should probably post a more complete example to allow any judgement - and measure first before trying non-portable constructs. Greetings, Thomas -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Dag-Erling Smørgrav Guest Posted: Mon Apr 23, 2012 1:31 pm    Post subject: Re: portable code rangsynth (AT) gmail (DOT) com writes: Quote: You can say write((i >> 24) & 0xFF) write((i >> 16) & 0xFF) write((i >> & 0xFF) write(i & 0xFF) There is no write() in C, but there is one in POSIX that doesn't match this usage: ssize_t write(int fd, void *buf, size_t len); Let's assume for now that this in fact a private function based on a global FILE *f: void write(unsigned char i) { if (fwrite(&i, 1, 1, f) != 1) { fprintf(stderr, "fwrite() failed\n"); exit(EX_FAILURE); } } (actually, we should assume that this function is called funwrite() and that write() is the following macro: #define write(i) funwrite(i); but I'm being charitable) Quote: int out = 0; out += (read() >> 24); out += (read() >> 16); out += (read() >> ; out += (read()); There is no read() in C, but there is one in POSIX that doesn't match this usage: ssize_t read(int fd, void *buf, size_t len); Let's assume for now that this is in fact a private function based on a global FILE *f: unsigned char read(void) { unsigned char i; if (fread(&i, 1, 1, f) != 1) { fprintf(stderr, "fread() failed\n"); exit(EX_FAILURE); } return i; } then the following: unsigned int i = 0x01020304; write((i >> 24) & 0xFF); write((i >> 16) & 0xFF); write((i >> & 0xFF); write(i & 0xFF); rewind(f); i = 0; i += (read() >> 24); i += (read() >> 16); i += (read() >> ; i += (read()); printf("i = 0x%x\n", i); will print i = 0x4 The correct answer is "store your data in a textual representation". DES -- Dag-Erling Smørgrav - des (AT) des (DOT) no -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Barry Schwarz Guest Posted: Mon Apr 23, 2012 1:32 pm    Post subject: Re: which one is faster? On Wed, 4 Apr 2012 18:05:42 -0500 (CDT), echo ma <fatmck (AT) gmail (DOT) com> wrote: Quote: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? Sorry for my poor english. It will depend on your system (hardware, compiler, and options). There is no guarantee that t.a = 0; and t.a = 1; will have the same performance. (Some systems have methods of zeroing memory that are more efficient than storing an arbitrary value.) For that matter, there is no guarantee that t.a = 0; and t.b = 0; have the same performance. (On some systems, access performance is different for different alignments.) And then step three accesses "more" memory than the first two so why should its performance be the same. The only way to answer these types of questions is to benchmark your code and even that is questionable unless you disable cache and your code is the ONLY process executing on the system, something pretty rare on most user systems these days. -- Remove del for email -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Keith Thompson Guest Posted: Mon Apr 23, 2012 1:32 pm    Post subject: Re: which one is faster? echo ma <fatmck (AT) gmail (DOT) com> writes: Quote: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? The C language standard says nothing about the relative performance of various operations, so there is no general answer to your question. The language specifies the behavior of programs, not their performance. You can answer it for your own system by measuring the performance. Suppose you find out that one operation is faster than another on your system; what would you do with that information? If you need to assign a value to t.c, assigning a value to t.a might be faster, but that doesn't do yhou much good if it's not the right thing to do. (Note that "#pragma pack" is non-standard and potentially unsafe; see <http://stackoverflow.com/q/8568432/827263>. For your particular struct definition, it's unlikely to have any effect.) -- Keith Thompson (The_Other_Keith) kst-u (AT) mib (DOT) org <http://www.ghoti.net/~kst> Will write code for food. "We must do something. This is something. Therefore, we must do this." -- Antony Jay and Jonathan Lynn, "Yes Minister" -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Bharat Guest Posted: Mon Apr 23, 2012 1:32 pm    Post subject: Re: portable code See http://code.google.com/p/protobuf-c/ http://en.wikipedia.org/wiki/Type-length-value you will have portable at the cost of slight cpu overhead. On Friday, 17 December 2010 00:15:38 UTC+5:30, raffamaiden wrote: Quote: Hi all. I'm writing a program wich will write some variables to an output file. I do something like int a =5; fwrite(&a, sizeof(int), 1, my_file_ptr); This will write an int to the file pointed by my_file_ptr. But i know that the c standard does not specify the exact size in bytes for its primitive type, as far as i know it only specifies that and int is an integer type that rapresents a number with a sign, but different implementations\operating systems can have different size for an int. So this mean that my program will write a 32 bit integer with one implementation and a 16 bit with another implementation. This would also mean that the file generated by the program that is running in one implementation will not be readable in another implementation, unless the program knows also in which implementation the instance that generated the file was running. That is right? I do not want such a behavior. How can i solve this? -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top echo ma Guest Posted: Mon Apr 23, 2012 1:32 pm    Post subject: Re: portable code You can use int32_t in <stdint.h>, this is part of C99. You should use int32_t in both source code and the file format document, so other developers will know the size of your int. Another thing you should know, the byte order things. Intel X86 and IA64 are using little endian, while IBM PowerPC and Sparc are using big endian. Memory layout are totally defferent when you assign 5 for an interger variable on there CPUs. You should use the a uniform byte order independent of CPUs, so you would like to use htons() and htonl() to made it. Or, you can use text representations, this is more portable and easy for human reading. Sorry for my poor English. Wish this help you. -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top echo ma Guest Posted: Mon Apr 23, 2012 1:33 pm    Post subject: Re: C Data Structures Libraries 在 2011å¹´2月5日星期六UTC+8上å�ˆ2æ—¶46分35秒，Nimo写é�“： Quote: Hello, where can I find free C Data Structures Libraries.. Thank You cheers -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. You must mean the common List / Map / Queue data structures. I want them too, currently always using STL in C++... I know there are double-linked-list and fifo implemented by linux kernel in C, but I have never used them. Check on this page: http://www.kernel.org/doc/htmldocs/kernel-api/ -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Joel C. Salomon Guest Posted: Mon Apr 23, 2012 1:33 pm    Post subject: Re: which one is faster? More CPU cycles have been spent passing this message across the Internet than would be saved by any such optimization. -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top David Brown Guest Posted: Mon Apr 23, 2012 1:33 pm    Post subject: Re: which one is faster? On 05/04/2012 01:05, echo ma wrote: Quote: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? Taken out of context, it is impossible to say. The speed of code depends on a huge number of factors, including the compiler, compiler options, processor (there are several dozens of commonly used 32-bit processor architectures), cache, memory, etc., as well as the surrounding code. As a wild generalisation, the timing will be much the same if the processor uses absolute addressing modes, but for processors that make more use of pointers there will be some initial code to set up the pointer before the first access, then each access will be quickly done. Quote: Sorry for my poor english. There is no problem with your English - the problem lies in not knowing what you are trying to ask. It is seldom a good plan to take statements out of context and ask if that one statement is fast or slow - you have to look at the bigger picture. -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Keith Thompson Guest Posted: Tue May 01, 2012 2:57 am    Post subject: Re: which one is faster? Thomas Richter <thor (AT) math (DOT) tu-berlin.de> writes: Quote: On 05.04.2012 01:05, echo ma wrote: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? Sorry, I don't understand. To have something "the same", you need to compare two different things. So which two "are the same" or "not the same"? I see only one piece of code, and this is not even portable. (#pragma pack() is not ANSI-C). I think echo ma is asking about the relative performance of the three statements "t.a = 0;", "t.b = 1;", and "t.c = 2;". I suppose if, for example, the assignment to t.a turned out to be significantly faster than the assignment to t.b, then it might make sense to rearrange the struct definition so the most frequently accessed members are first. In practice, such rearrangement is unlikely to be helpful. [...] -- Keith Thompson (The_Other_Keith) kst-u (AT) mib (DOT) org <http://www.ghoti.net/~kst> Will write code for food. "We must do something. This is something. Therefore, we must do this." -- Antony Jay and Jonathan Lynn, "Yes Minister" -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top George Neuner Guest Posted: Tue May 01, 2012 2:58 am    Post subject: Re: which one is faster? On Mon, 23 Apr 2012 08:32:16 -0500 (CDT), Keith Thompson <kst-u (AT) mib (DOT) org> wrote: Quote: echo ma <fatmck (AT) gmail (DOT) com> writes: on a 32bit system. #pragma pack(4) struct TestStruct { unsigned short a; unsigned short b; unsigned int c; }; #pragma pack() struct TestStruct t; t.a = 0; //step 1 t.b = 1; //step 2 t.c = 2; //step 3 Questing is : Are these 3 steps having the same perfomance time? Suppose you find out that one operation is faster than another on your system; what would you do with that information? If you need to assign a value to t.c, assigning a value to t.a might be faster, but that doesn't do yhou much good if it's not the right thing to do. (Note that "#pragma pack" is non-standard and potentially unsafe; see <http://stackoverflow.com/q/8568432/827263>. For your particular struct definition, it's unlikely to have any effect.) The only real reasons for such tightly packed structures are to save memory in *very* small systems, or to perform memory mapped device access ... which brings up both coding issues such as register value reuse and CPU/platform issues such as bus width, access alignment, write-back caching, out-of-order execution, write combining, etc. x86 being an exception, most 32-bit CPUs use memory mapped I/O exclusively, and it often is desirable to create structures for accessing groups of contiguous device "registers". However, in such cases, making a single 32-bit access to A+B in order to write a 16-bit value to B would be incorrect (and potentially destructive). As someone else said already, it would be nice if the OP had given a more complete description of the intended use. George -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Jorgen Grahn Guest Posted: Tue May 01, 2012 2:59 am    Post subject: Re: portable code On Mon, 2012-04-23, Dag-Erling Smørgrav wrote: Quote: rangsynth (AT) gmail (DOT) com writes: You can say write((i >> 24) & 0xFF) write((i >> 16) & 0xFF) write((i >> & 0xFF) write(i & 0xFF) There is no write() in C, but there is one in POSIX that doesn't match this usage: I am pretty sure rangsynth just used "write" as a generic "write an octet to somewhere" function. The details are irrelevant here. [snip irrelevant] Quote: but I'm being charitable) int out = 0; out += (read() >> 24); out += (read() >> 16); out += (read() >> ; out += (read()); There is no read() in C, but there is one in POSIX that doesn't match this usage: [snip irrelevant] Quote: then the following: unsigned int i = 0x01020304; write((i >> 24) & 0xFF); write((i >> 16) & 0xFF); write((i >> & 0xFF); write(i & 0xFF); rewind(f); i = 0; i += (read() >> 24); i += (read() >> 16); i += (read() >> ; i += (read()); printf("i = 0x%x\n", i); will print i = 0x4 If it does, it's because you introduced a bug somewhere. Why? I feel I'm missing some point here; can you please explain it more clearly? Quote: The correct answer is "store your data in a textual representation". I prefer that too, but you don't always get to choose your requirements. Especially not file formats and protocols! /Jorgen -- // Jorgen Grahn <grahn@ Oo o. . . \X/ snipabacken.se> O o . -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Bharat Guest Posted: Tue May 01, 2012 2:59 am    Post subject: Re: C Data Structures Libraries did you try google? here is the first listing i got http://home.gna.org/gdsl/ On Saturday, 5 February 2011 00:16:35 UTC+5:30, Nimo wrote: Quote: Hello, where can I find free C Data Structures Libraries.. Thank You cheers -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Bharat Guest Posted: Tue May 01, 2012 3:00 am    Post subject: Re: portable code On Friday, 17 December 2010 00:15:38 UTC+5:30, raffamaiden wrote: Quote: Hi all. I'm writing a program wich will write some variables to an output file. I do something like int a =5; fwrite(&a, sizeof(int), 1, my_file_ptr); This will write an int to the file pointed by my_file_ptr. But i know that the c standard does not specify the exact size in bytes for its primitive type, as far as i know it only specifies that and int is an integer type that rapresents a number with a sign, but different implementations\operating systems can have different size for an int. So this mean that my program will write a 32 bit integer with one implementation and a 16 bit with another implementation. This would also mean that the file generated by the program that is running in one implementation will not be readable in another implementation, unless the program knows also in which implementation the instance that generated the file was running. That is right? I do not want such a behavior. How can i solve this? -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. See usage of TLV http://en.wikipedia.org/wiki/Type-length-value See http://code.google.com/p/protobuf/ -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top Dag-Erling Smørgrav Guest Posted: Fri May 04, 2012 10:29 pm    Post subject: Re: portable code Jorgen Grahn <grahn+nntp (AT) snipabacken (DOT) se> writes: Quote: Dag-Erling Smørgrav <des (AT) des (DOT) no> writes: There is no write() in C, but there is one in POSIX that doesn't match this usage: I am pretty sure rangsynth just used "write" as a generic "write an octet to somewhere" function. The details are irrelevant here. The details are not irrelevant. When you post code that operates on the value returned by a function, the semantics of that function and the range and representation of its return value are highly relevant. Quote: then the following: unsigned int i = 0x01020304; write((i >> 24) & 0xFF); write((i >> 16) & 0xFF); write((i >> & 0xFF); write(i & 0xFF); rewind(f); i = 0; i += (read() >> 24); i += (read() >> 16); i += (read() >> ; i += (read()); printf("i = 0x%x\n", i); will print i = 0x4 If it does, it's because you introduced a bug somewhere. Why? I feel I'm missing some point here; can you please explain it more clearly? I didn't introduce a bug anywhere. On the contrary - I was pointing out a bug in rangsynth's code. I'm sure you'll see it too if you look closely. DES -- Dag-Erling Smørgrav - des (AT) des (DOT) no -- comp.lang.c.moderated - moderation address: clcm (AT) plethora (DOT) net -- you must have an appropriate newsgroups line in your header for your mail to be seen, or the newsgroup name in square brackets in the subject line. Sorry. Back to top        C++Talk.NET Forum Index -> C Language (Moderated) All times are GMT Goto page Previous  1, 2, 3 ... , 11, 12, 13  Next Page 12 of 13 Powered by phpBB © 2001, 2006 phpBB Group