C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  'Portable' Measurement of Pointer Alignment in C?          C++Talk.NET Forum Index -> C Language (Moderated) View previous topic :: View next topic   Author Message Brian Gladman Guest Posted: Thu Mar 29, 2007 6:24 am    Post subject: 'Portable' Measurement of Pointer Alignment in C? A lot of low level cryptographic code and some hardware cryptographic accelerators either fail completely or perform very poorly if their input, output and key storage areas in memory are not aligned on specific memory boundaries. Moreover, in many situations the cryptographic code does not itself have any control over the memory alignment of its parameters so the best it can do is to detect if these aligmments are correct and act accordingly. This hence rasises the question of the most appropriate way of determining the aligment of a pointer. Here my interests is less in the 'political' correctness of C code but rather in what method is most likely to work in practice on the highest proportion of widely deployed processors and C compilers. For example, when 'x' is a pointer of some kind, 'n' is a power of two and 'pint' is a pointer sized integer type, on what proportion of systems will: #define PTR_OFFSET(x,n) (((pint)x) & ((n) - 1)) return the alignment of 'x ' from an 'n' byte boundary in memory? Is: #define PTR_OFFSET(x,n) (((char*)(x) - (char*)0) & ((n) - 1)) any better (or worse)? Or, on systems that allow for the declaration of aligned variables, is: declare_aligned(n) type var; #define PTR_OFFSET(x,n) (((char*)(x) - (char*)&var) & ((n) - 1)) any better? I would much appreciate any observations that people here might have on the most portable (in practice) ways in which pointer alignment might be determined when the pointer itself is not under the control of the software that needs to determine this. Brian Gladman -- 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 Hans-Bernhard Bröker Guest Posted: Mon Apr 16, 2007 8:17 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? Brian Gladman wrote: Quote: A lot of low level cryptographic code and some hardware cryptographic accelerators either fail completely or perform very poorly if their input, output and key storage areas in memory are not aligned on specific memory boundaries. Well, that's far into "Don't do that, then!" country. At least one instance of either undefined behaviour (bad pointer cast) or decidedly suboptimal toolchain behaviour (insufficiently aligned heap allocations) has to occur before this can become a real problem. Quote: Here my interests is less in the 'political' correctness of C code but rather in what method is most likely to work in practice on the highest proportion of widely deployed processors and C compilers. Cast the pointer to unsigned long, if you're in C90, or to C99's intptr_t, and assume that integer is the actual address. If that doesn't work, nothing else is likely to. -- 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 WillerZ Guest Posted: Mon Apr 16, 2007 8:18 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? Brian Gladman wrote: Quote: A lot of low level cryptographic code and some hardware cryptographic accelerators either fail completely or perform very poorly if their input, output and key storage areas in memory are not aligned on specific memory boundaries. Moreover, in many situations the cryptographic code does not itself have any control over the memory alignment of its parameters so the best it can do is to detect if these aligmments are correct and act accordingly. This hence rasises the question of the most appropriate way of determining the aligment of a pointer. Here my interests is less in the 'political' correctness of C code but rather in what method is most likely to work in practice on the highest proportion of widely deployed processors and C compilers. For example, when 'x' is a pointer of some kind, 'n' is a power of two and 'pint' is a pointer sized integer type, on what proportion of systems will: #define PTR_OFFSET(x,n) (((pint)x) & ((n) - 1)) return the alignment of 'x ' from an 'n' byte boundary in memory? Is: #define PTR_OFFSET(x,n) (((char*)(x) - (char*)0) & ((n) - 1)) any better (or worse)? Or, on systems that allow for the declaration of aligned variables, is: declare_aligned(n) type var; #define PTR_OFFSET(x,n) (((char*)(x) - (char*)&var) & ((n) - 1)) any better? I would much appreciate any observations that people here might have on the most portable (in practice) ways in which pointer alignment might be determined when the pointer itself is not under the control of the software that needs to determine this. Brian Gladman In C99 I would prefer: #include <stdint.h> inline uintptr_t PRT_OFFSET(void *x, uintptr_t n) { return ((uintptr_t)x) & (n-1); } Of the three #define statements above, the central one is (IMO) most likely to work. I would expect difficulties with the top definition on IBM System i. The third definition is obviously useless without the option of declaring aligned variables. -- 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 Thomas Richter Guest Posted: Mon Apr 16, 2007 8:18 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? Brian Gladman wrote: Quote: A lot of low level cryptographic code and some hardware cryptographic accelerators either fail completely or perform very poorly if their input, output and key storage areas in memory are not aligned on specific memory boundaries. Moreover, in many situations the cryptographic code does not itself have any control over the memory alignment of its parameters so the best it can do is to detect if these aligmments are correct and act accordingly. This hence rasises the question of the most appropriate way of determining the aligment of a pointer. Here my interests is less in the 'political' correctness of C code but rather in what method is most likely to work in practice on the highest proportion of widely deployed processors and C compilers. For example, when 'x' is a pointer of some kind, 'n' is a power of two and 'pint' is a pointer sized integer type, on what proportion of systems will: #define PTR_OFFSET(x,n) (((pint)x) & ((n) - 1)) The above would work with: #include <stdint.h> typedef intptr_t pint; The "intptr_t" type of stdint.h is an integer large enough to hold a pointer without loss. Another candidate (I'm using here) is to cast the pointer to ptrdiff_t, though (on wierd systems) this might be smaller than an intptr_t, thus the first option should probably be your best candidate. So long, 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 Douglas A. Gwyn Guest Posted: Thu Apr 19, 2007 7:13 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? David Given wrote: Quote: However, I don't think there are any platforms where pointers aren't unit aligned; ... It's fairly common for some of the wider types (long long and double, for example) to have a hardware-mandated alignment that is not as strict as the full width. Some architectures support arbitrary byte alignment of the built-in types, although it is usually more efficient to nevertheless use a wider alignment (to match memory subsystem width). -- 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: Thu Apr 19, 2007 7:13 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? "Douglas A. Gwyn" <DAGwyn (AT) null (DOT) net> writes: Quote: Brian Gladman wrote: I would much appreciate any observations that people here might have on the most portable (in practice) ways in which pointer alignment might be determined when the pointer itself is not under the control of the software that needs to determine this. I think a much better approach would be for the library to allocate its own buffers, ensuring that they are properly aligned. On the rare occasions when I am trying to ensure such address alignment, I use the following macros: #define AlignDown(p) ((uintptr_t)(p)/ALIGN_SIZE*ALIGN_SIZE) #define AlignUp(p) (((uintptr_t)(p)+(ALIGN_SIZE-1))/ALIGN_SIZE*ALIGN_SIZE) and let the compiler take care of optimization. That's likely to work on most systems, but it's not 100% portable. It depends on assumptions about pointer representations that aren't guaranteed by the standard. I've worked on systems where this would fail (because the low-order bits of a pointer don't correspond to byte addresses). -- Keith Thompson (The_Other_Keith) kst-u (AT) mib (DOT) org <http://www.ghoti.net/~kst> San Diego Supercomputer Center <*> <http://users.sdsc.edu/~kst> "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 Brian Gladman Guest Posted: Thu Apr 19, 2007 7:15 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? "Douglas A. Gwyn" <DAGwyn (AT) null (DOT) net> wrote in message news:clcm-20070416-0022 (AT) plethora (DOT) net... Quote: Brian Gladman wrote: I would much appreciate any observations that people here might have on the most portable (in practice) ways in which pointer alignment might be determined when the pointer itself is not under the control of the software that needs to determine this. My thanks to everyone who responded to my post. I should mention that, as this moderated group didn't appear to work very rapidly, I also aksed this question comp.lang.c and got quite a few helpful answers. Quote: I think a much better approach would be for the library to allocate its own buffers, ensuring that they are properly aligned. I agree and where I have control over buffer allocation, this is what I do. The situations in which the issue I raised comes up is when I am delivered a buffer and I don't know its alignment but what I have to do depends on this. Quote: On the rare occasions when I am trying to ensure such address alignment, I use the following macros: #define AlignDown(p) ((uintptr_t)(p)/ALIGN_SIZE*ALIGN_SIZE) #define AlignUp(p) (((uintptr_t)(p)+(ALIGN_SIZE-1))/ALIGN_SIZE*ALIGN_SIZE) and let the compiler take care of optimizations Thanks for this suggestion. Right now I am casting to an unsigned char and using an AND mask (I know that my alignment requirement is a power of 2 not more than 256) because most people who responded on comp.lang.c seemed to think that this was the approach most likely to work on most systems. I didn't use uintptr_t since I am suspicious that this is a C99 invention - was it available in the earlier C standards? Thanks again to all who have contributed. Brian Gladman -- 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 Brian Gladman Guest Posted: Thu Apr 19, 2007 7:15 pm    Post subject: Re: 'Portable' Measurement of Pointer Alignment in C? "David Given" <dg (AT) cowlark (DOT) com> wrote in message news:clcm-20070416-0012 (AT) plethora (DOT) net... Quote: Brian Gladman wrote: [...] I would much appreciate any observations that people here might have on the most portable (in practice) ways in which pointer alignment might be determined when the pointer itself is not under the control of the software that needs to determine this. I'm not entirely sure what you're asking here; do you want to know the alignment requirements of a pointer type on your system, or the alignment of a specific pointer argument you've been given? The latter. Quote: For (b), you should be able to do this: intptr_t arg_alignment = (intptr_t)arg & (page_size-1); ...where page_size varies according to the size of the thing you want to measure alignment against. But that's so trivial that I don't think it was really what you wanted to know. Yes, this was what I was after. I was wondering about the proportion of widely deployed systems this would work on. I know of some systems where the low end of a pointer won't indicate a pointer's alignment so I was seeking evidence on whether such systems are around in sufficient numbers to make this simple aligmnment test dangerous in practice. Thanks for your help on this (thanks also to other who have responded). Brian Gladman -- 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 Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.