Wednesday, January 09, 2013

Hidden Variables: Arduino Due and the Cortex-M3

Once again, I'm curious about what preprocessor symbols are predefined by the GNU C compiler. This time it's for the ARM Cortex-M3 core in the Atmel SAM3X8E chip on my Arduino Due. Here is what the command

arm-none-eabi-g++ -dM -E -mcpu=cortex-m3 -mthumb - < /dev/null

for GCC 4.7.2 from the CodeSourcery CodeBench Lite 2012.09-63 toolchain yields.

#define __DBL_MIN_EXP__ (-1021)
#define __HQ_FBIT__ 15
#define __UINT_LEAST16_MAX__ 65535
#define __ATOMIC_ACQUIRE 2
#define __SFRACT_IBIT__ 0
#define __FLT_MIN__ 1.1754943508222875e-38F
#define __UFRACT_MAX__ 0XFFFFP-16UR
#define __UINT_LEAST8_TYPE__ unsigned char
#define __DQ_FBIT__ 63
#define __INTMAX_C(c) c ## LL
#define __CS_SOURCERYGXX_REV__ 63
#define __ULFRACT_FBIT__ 32
#define __SACCUM_EPSILON__ 0x1P-7HK
#define __CHAR_BIT__ 8
#define __USQ_IBIT__ 0
#define __UINT8_MAX__ 255
#define __ACCUM_FBIT__ 15
#define __WINT_MAX__ 4294967295U
#define __USFRACT_FBIT__ 8
#define __ORDER_LITTLE_ENDIAN__ 1234
#define __SIZE_MAX__ 4294967295U
#define __WCHAR_MAX__ 4294967295U
#define __LACCUM_IBIT__ 32
#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_1 1
#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_2 1
#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4 1
#define __DBL_DENORM_MIN__ ((double)4.9406564584124654e-324L)
#define __GCC_ATOMIC_CHAR_LOCK_FREE 2
#define __FLT_EVAL_METHOD__ 0
#define __LLACCUM_MAX__ 0X7FFFFFFFFFFFFFFFP-31LLK
#define __GCC_ATOMIC_CHAR32_T_LOCK_FREE 2
#define __FRACT_FBIT__ 15
#define __UINT_FAST64_MAX__ 18446744073709551615ULL
#define __SIG_ATOMIC_TYPE__ int
#define __UACCUM_FBIT__ 16
#define __DBL_MIN_10_EXP__ (-307)
#define __FINITE_MATH_ONLY__ 0
#define __ARMEL__ 1
#define __ARM_FEATURE_UNALIGNED 1
#define __ARM_ARCH_7M__ 1
#define __LFRACT_IBIT__ 0
#define __GNUC_PATCHLEVEL__ 2
#define __LFRACT_MAX__ 0X7FFFFFFFP-31LR
#define __UINT_FAST8_MAX__ 4294967295U
#define __DEC64_MAX_EXP__ 385
#define __INT8_C(c) c
#define __UINT_LEAST64_MAX__ 18446744073709551615ULL
#define __SA_FBIT__ 15
#define __SHRT_MAX__ 32767
#define __LDBL_MAX__ 1.7976931348623157e+308L
#define __FRACT_MAX__ 0X7FFFP-15R
#define __thumb2__ 1
#define __UFRACT_FBIT__ 16
#define __UFRACT_MIN__ 0.0UR
#define __UINT_LEAST8_MAX__ 255
#define __GCC_ATOMIC_BOOL_LOCK_FREE 2
#define __UINTMAX_TYPE__ long long unsigned int
#define __LLFRACT_EPSILON__ 0x1P-63LLR
#define __DEC32_EPSILON__ 1E-6DF
#define __CHAR_UNSIGNED__ 1
#define __UINT32_MAX__ 4294967295UL
#define __ULFRACT_MAX__ 0XFFFFFFFFP-32ULR
#define __TA_IBIT__ 64
#define __LDBL_MAX_EXP__ 1024
#define __WINT_MIN__ 0U
#define __ULLFRACT_MIN__ 0.0ULLR
#define __SCHAR_MAX__ 127
#define __WCHAR_MIN__ 0U
#define __INT64_C(c) c ## LL
#define __DBL_DIG__ 15
#define __GCC_ATOMIC_POINTER_LOCK_FREE 2
#define __LLACCUM_MIN__ (-0X1P31LLK-0X1P31LLK)
#define __SIZEOF_INT__ 4
#define __SIZEOF_POINTER__ 4
#define __USACCUM_IBIT__ 8
#define __USER_LABEL_PREFIX__ 
#define __STDC_HOSTED__ 1
#define __LDBL_HAS_INFINITY__ 1
#define __LFRACT_MIN__ (-0.5LR-0.5LR)
#define __HA_IBIT__ 8
#define __TQ_IBIT__ 0
#define __FLT_EPSILON__ 1.1920928955078125e-7F
#define __APCS_32__ 1
#define __USFRACT_IBIT__ 0
#define __LDBL_MIN__ 2.2250738585072014e-308L
#define __FRACT_MIN__ (-0.5R-0.5R)
#define __DEC32_MAX__ 9.999999E96DF
#define __DA_IBIT__ 32
#define __INT32_MAX__ 2147483647L
#define __UQQ_FBIT__ 8
#define __SIZEOF_LONG__ 4
#define __UACCUM_MAX__ 0XFFFFFFFFP-16UK
#define __UINT16_C(c) c
#define __DECIMAL_DIG__ 17
#define __LFRACT_EPSILON__ 0x1P-31LR
#define __ULFRACT_MIN__ 0.0ULR
#define __LDBL_HAS_QUIET_NAN__ 1
#define __ULACCUM_IBIT__ 32
#define __UACCUM_EPSILON__ 0x1P-16UK
#define __GNUC__ 4
#define __ULLACCUM_MAX__ 0XFFFFFFFFFFFFFFFFP-32ULLK
#define __HQ_IBIT__ 0
#define __FLT_HAS_DENORM__ 1
#define __SIZEOF_LONG_DOUBLE__ 8
#define __BIGGEST_ALIGNMENT__ 8
#define __DQ_IBIT__ 0
#define __DBL_MAX__ ((double)1.7976931348623157e+308L)
#define __ULFRACT_IBIT__ 0
#define __INT_FAST32_MAX__ 2147483647
#define __DBL_HAS_INFINITY__ 1
#define __ACCUM_IBIT__ 16
#define __DEC32_MIN_EXP__ (-94)
#define __THUMB_INTERWORK__ 1
#define __LACCUM_MAX__ 0X7FFFFFFFFFFFFFFFP-31LK
#define __INT_FAST16_TYPE__ int
#define __LDBL_HAS_DENORM__ 1
#define __DEC128_MAX__ 9.999999999999999999999999999999999E6144DL
#define __INT_LEAST32_MAX__ 2147483647L
#define __ARM_PCS 1
#define __DEC32_MIN__ 1E-95DF
#define __ACCUM_MAX__ 0X7FFFFFFFP-15K
#define __DBL_MAX_EXP__ 1024
#define __USACCUM_EPSILON__ 0x1P-8UHK
#define __DEC128_EPSILON__ 1E-33DL
#define __SFRACT_MAX__ 0X7FP-7HR
#define __FRACT_IBIT__ 0
#define __PTRDIFF_MAX__ 2147483647
#define __UACCUM_MIN__ 0.0UK
#define __UACCUM_IBIT__ 16
#define __LONG_LONG_MAX__ 9223372036854775807LL
#define __SIZEOF_SIZE_T__ 4
#define __ULACCUM_MAX__ 0XFFFFFFFFFFFFFFFFP-32ULK
#define __SIZEOF_WINT_T__ 4
#define __SA_IBIT__ 16
#define __ULLACCUM_MIN__ 0.0ULLK
#define __GXX_ABI_VERSION 1002
#define __UTA_FBIT__ 64
#define __SOFTFP__ 1
#define __FLT_MIN_EXP__ (-125)
#define __USFRACT_MAX__ 0XFFP-8UHR
#define __UFRACT_IBIT__ 0
#define __INT_FAST64_TYPE__ long long int
#define __DBL_MIN__ ((double)2.2250738585072014e-308L)
#define __LACCUM_MIN__ (-0X1P31LK-0X1P31LK)
#define __ULLACCUM_FBIT__ 32
#define __GXX_TYPEINFO_EQUALITY_INLINE 0
#define __ULLFRACT_EPSILON__ 0x1P-64ULLR
#define __USES_INITFINI__ 1
#define __DEC128_MIN__ 1E-6143DL
#define __REGISTER_PREFIX__ 
#define __UINT16_MAX__ 65535
#define __DBL_HAS_DENORM__ 1
#define __ACCUM_MIN__ (-0X1P15K-0X1P15K)
#define __SQ_IBIT__ 0
#define __UINT8_TYPE__ unsigned char
#define __UHA_FBIT__ 8
#define __NO_INLINE__ 1
#define __SFRACT_MIN__ (-0.5HR-0.5HR)
#define __UTQ_FBIT__ 128
#define __FLT_MANT_DIG__ 24
#define __VERSION__ "4.7.2"
#define __UINT64_C(c) c ## ULL
#define __ULLFRACT_FBIT__ 64
#define __FRACT_EPSILON__ 0x1P-15R
#define __ULACCUM_MIN__ 0.0ULK
#define __UDA_FBIT__ 32
#define __LLACCUM_EPSILON__ 0x1P-31LLK
#define __GCC_ATOMIC_INT_LOCK_FREE 2
#define __FLOAT_WORD_ORDER__ __ORDER_LITTLE_ENDIAN__
#define __USFRACT_MIN__ 0.0UHR
#define __UQQ_IBIT__ 0
#define __INT32_C(c) c ## L
#define __DEC64_EPSILON__ 1E-15DD
#define __ORDER_PDP_ENDIAN__ 3412
#define __DEC128_MIN_EXP__ (-6142)
#define __UHQ_FBIT__ 16
#define __LLACCUM_FBIT__ 31
#define __INT_FAST32_TYPE__ int
#define __UINT_LEAST16_TYPE__ short unsigned int
#define __INT16_MAX__ 32767
#define __SIZE_TYPE__ unsigned int
#define __UINT64_MAX__ 18446744073709551615ULL
#define __UDQ_FBIT__ 64
#define __INT8_TYPE__ signed char
#define __thumb__ 1
#define __ELF__ 1
#define __ULFRACT_EPSILON__ 0x1P-32ULR
#define __LLFRACT_FBIT__ 63
#define __FLT_RADIX__ 2
#define __INT_LEAST16_TYPE__ short int
#define __LDBL_EPSILON__ 2.2204460492503131e-16L
#define __UINTMAX_C(c) c ## ULL
#define __SACCUM_MAX__ 0X7FFFP-7HK
#define __SIG_ATOMIC_MAX__ 2147483647
#define __GCC_ATOMIC_WCHAR_T_LOCK_FREE 2
#define __VFP_FP__ 1
#define __SIZEOF_PTRDIFF_T__ 4
#define __CS_SOURCERYGXX_MIN__ 9
#define __LACCUM_EPSILON__ 0x1P-31LK
#define __DEC32_SUBNORMAL_MIN__ 0.000001E-95DF
#define __INT_FAST16_MAX__ 2147483647
#define __UINT_FAST32_MAX__ 4294967295U
#define __UINT_LEAST64_TYPE__ long long unsigned int
#define __USACCUM_MAX__ 0XFFFFP-8UHK
#define __SFRACT_EPSILON__ 0x1P-7HR
#define __FLT_HAS_QUIET_NAN__ 1
#define __FLT_MAX_10_EXP__ 38
#define __LONG_MAX__ 2147483647L
#define __DEC128_SUBNORMAL_MIN__ 0.000000000000000000000000000000001E-6143DL
#define __FLT_HAS_INFINITY__ 1
#define __USA_FBIT__ 16
#define __UINT_FAST16_TYPE__ unsigned int
#define __DEC64_MAX__ 9.999999999999999E384DD
#define __CHAR16_TYPE__ short unsigned int
#define __PRAGMA_REDEFINE_EXTNAME 1
#define __CS_SOURCERYGXX_MAJ__ 2012
#define __INT_LEAST16_MAX__ 32767
#define __DEC64_MANT_DIG__ 16
#define __INT64_MAX__ 9223372036854775807LL
#define __UINT_LEAST32_MAX__ 4294967295UL
#define __SACCUM_FBIT__ 7
#define __GCC_ATOMIC_LONG_LOCK_FREE 2
#define __INT_LEAST64_TYPE__ long long int
#define __INT16_TYPE__ short int
#define __INT_LEAST8_TYPE__ signed char
#define __SQ_FBIT__ 31
#define __DEC32_MAX_EXP__ 97
#define __INT_FAST8_MAX__ 2147483647
#define __INTPTR_MAX__ 2147483647
#define __QQ_FBIT__ 7
#define __UTA_IBIT__ 64
#define __LDBL_MANT_DIG__ 53
#define __SFRACT_FBIT__ 7
#define __SACCUM_MIN__ (-0X1P7HK-0X1P7HK)
#define __DBL_HAS_QUIET_NAN__ 1
#define __SIG_ATOMIC_MIN__ (-__SIG_ATOMIC_MAX__ - 1)
#define __INTPTR_TYPE__ int
#define __UINT16_TYPE__ short unsigned int
#define __WCHAR_TYPE__ unsigned int
#define __SIZEOF_FLOAT__ 4
#define __THUMBEL__ 1
#define __USQ_FBIT__ 32
#define __UINTPTR_MAX__ 4294967295U
#define __DEC64_MIN_EXP__ (-382)
#define __ULLACCUM_IBIT__ 32
#define __INT_FAST64_MAX__ 9223372036854775807LL
#define __GCC_ATOMIC_TEST_AND_SET_TRUEVAL 1
#define __FLT_DIG__ 6
#define __UINT_FAST64_TYPE__ long long unsigned int
#define __INT_MAX__ 2147483647
#define __LACCUM_FBIT__ 31
#define __USACCUM_MIN__ 0.0UHK
#define __UHA_IBIT__ 8
#define __INT64_TYPE__ long long int
#define __FLT_MAX_EXP__ 128
#define __UTQ_IBIT__ 0
#define __DBL_MANT_DIG__ 53
#define __INT_LEAST64_MAX__ 9223372036854775807LL
#define __GCC_ATOMIC_CHAR16_T_LOCK_FREE 2
#define __DEC64_MIN__ 1E-383DD
#define __WINT_TYPE__ unsigned int
#define __UINT_LEAST32_TYPE__ long unsigned int
#define __SIZEOF_SHORT__ 2
#define __ULLFRACT_IBIT__ 0
#define __LDBL_MIN_EXP__ (-1021)
#define __arm__ 1
#define __UDA_IBIT__ 32
#define __INT_LEAST8_MAX__ 127
#define __LFRACT_FBIT__ 31
#define __LDBL_MAX_10_EXP__ 308
#define __ATOMIC_RELAXED 0
#define __DBL_EPSILON__ ((double)2.2204460492503131e-16L)
#define __UINT8_C(c) c
#define __INT_LEAST32_TYPE__ long int
#define __SIZEOF_WCHAR_T__ 4
#define __UINT64_TYPE__ long long unsigned int
#define __LLFRACT_MAX__ 0X7FFFFFFFFFFFFFFFP-63LLR
#define __TQ_FBIT__ 127
#define __INT_FAST8_TYPE__ int
#define __ULLACCUM_EPSILON__ 0x1P-32ULLK
#define __UHQ_IBIT__ 0
#define __LLACCUM_IBIT__ 32
#define __DBL_DECIMAL_DIG__ 17
#define __DEC_EVAL_METHOD__ 2
#define __TA_FBIT__ 63
#define __UDQ_IBIT__ 0
#define __ORDER_BIG_ENDIAN__ 4321
#define __ACCUM_EPSILON__ 0x1P-15K
#define __UINT32_C(c) c ## UL
#define __INTMAX_MAX__ 9223372036854775807LL
#define __BYTE_ORDER__ __ORDER_LITTLE_ENDIAN__
#define __FLT_DENORM_MIN__ 1.4012984643248171e-45F
#define __LLFRACT_IBIT__ 0
#define __INT8_MAX__ 127
#define __UINT_FAST32_TYPE__ unsigned int
#define __CHAR32_TYPE__ long unsigned int
#define __FLT_MAX__ 3.4028234663852886e+38F
#define __USACCUM_FBIT__ 8
#define __INT32_TYPE__ long int
#define __SIZEOF_DOUBLE__ 8
#define __FLT_MIN_10_EXP__ (-37)
#define __UFRACT_EPSILON__ 0x1P-16UR
#define __INTMAX_TYPE__ long long int
#define __DEC128_MAX_EXP__ 6145
#define __ATOMIC_CONSUME 1
#define __GNUC_MINOR__ 7
#define __UINTMAX_MAX__ 18446744073709551615ULL
#define __DEC32_MANT_DIG__ 7
#define __HA_FBIT__ 7
#define __DBL_MAX_10_EXP__ 308
#define __LDBL_DENORM_MIN__ 4.9406564584124654e-324L
#define __INT16_C(c) c
#define __STDC__ 1
#define __PTRDIFF_TYPE__ int
#define __LLFRACT_MIN__ (-0.5LLR-0.5LLR)
#define __ATOMIC_SEQ_CST 5
#define __DA_FBIT__ 31
#define __UINT32_TYPE__ long unsigned int
#define __ARM_ARCH_EXT_IDIV__ 1
#define __UINTPTR_TYPE__ unsigned int
#define __USA_IBIT__ 16
#define __DEC64_SUBNORMAL_MIN__ 0.000000000000001E-383DD
#define __ARM_EABI__ 1
#define __DEC128_MANT_DIG__ 34
#define __LDBL_MIN_10_EXP__ (-307)
#define __SIZEOF_LONG_LONG__ 8
#define __ULACCUM_EPSILON__ 0x1P-32ULK
#define __SACCUM_IBIT__ 8
#define __GCC_ATOMIC_LLONG_LOCK_FREE 1
#define __LDBL_DIG__ 15
#define __FLT_DECIMAL_DIG__ 9
#define __UINT_FAST16_MAX__ 4294967295U
#define __GNUC_GNU_INLINE__ 1
#define __GCC_ATOMIC_SHORT_LOCK_FREE 2
#define __ULLFRACT_MAX__ 0XFFFFFFFFFFFFFFFFP-64ULLR
#define __UINT_FAST8_TYPE__ unsigned int
#define __USFRACT_EPSILON__ 0x1P-8UHR
#define __ULACCUM_FBIT__ 32
#define __QQ_IBIT__ 0
#define __ATOMIC_ACQ_REL 4
#define __ATOMIC_RELEASE 3

I find this kind of stuff fascinating. And necessary, for the kind of work that I do down close to bare metal where the underlying details are often important. But I'm careful about writing code that depends on these symbols unless it's really necessary. For example, I much prefer to use sizeof(size_t) instead of __SIZEOF_SIZE_T__ if possible (sometimes it isn't).

It's been interesting having spent the past several years developing low level code for embedded systems whose processors range from powerful ARM microprocessors like the Cortex-A8 -- a 32-bit von Neumann architecture with a memory management unit and on which I run Linux -- to Atmel AVR and Microchip PIC16 microcontrollers -- 8-bit Harvard architectures sometimes with as little as ninety-six bytes of RAM on which I may run just a task loop -- to see the ARM Cortex-M3 which straddles both worlds -- a 32-bit ARM core yet with separate program and data memories.

My friends who are Java developers may safely go back to downloading an additional terabyte of framework to add yet another layer of abstraction to whatever they're trying to accomplish.

3 comments:

Anonymous said...

Hello John, I'm a loyal reader from Argentina. I have to say that your blog is a never-ending source of technical knowledge and carreer advices.

While I was reading your last post, I found something that made me think (again) how much I don't know about programming, even though I've been doing it for the past 13 years:

"I much prefer to use sizeof(size_t) instead of __SIZEOF_SIZE_T__ if possible (sometimes it isn't)."

Would you mind to explain the difference, please?

Best regards,
Fernando

Chip Overclock said...

When you submit your source code to the C compiler, whether you're using GNU C or some other compiler, there are two distinct phases: the preprocessor phase and the compile phase.

The preprocessor is the component that understands stuff that starts with a hash or pound symbol, like "#if" or "#define" or even "#include". The symbols I show in this article are the result of telling the GNU C compiler front-end to only run the preprocessor ("-E") and to emit the symbols that it defines itself ("-dM") as a result of my telling it the type of processor for which I am compiling ("-mcpu=cortex-mx" and "-mthumb"). These symbols are preprocessor symbols, and they and the "#" operators that manipulate them really amount to a completely different programming language with its own rules quite separate from the C language itself. It's a programming language that is interpreted and executed _before_ the real C compiler actually sees your code.

This is not a new thing, BTW. For example, back when I was writing assembler code for an IBM 360, when glaciers still covered the land, the process of running the IBM assembler also had two phases, the macro expansion phase that had its own quite complex language, and the the actual phase that turns assembler code into binary machine code. This kind of thing turns up a lot.

The C preprocessor, and the IBM assembler macro expansion, and other tools like the "m4" utility, are immensely powerful tools that perform "code generation". As you might expect, with great power comes great responsibility. What comes out of the C preprocessor is pure C code, without any "#" preprocessor directives _and_ without any preprocessor symbols; all of those have been replaced with whatever they were defined to be during the preprocessor step. The preprocessor has also merged all of the header files you referenced, directly or indirectly, into your code, into one immense "translation unit".

The preprocessor doesn't understand the C "sizeof" operator because that's done in the second phase, the actual code compilation that turns the code that you wrote AND the code generated as a result of the preprocessor into executable machine code. The preprocessor can't really grok "sizeof" because it doesn't know squat about things like variables, types, structures, and what not. These are things only known to the compiler, and their size is the result of the compiler doing its thing as it turns code into machine instructions.

If you want to _generate_ different code depending on the size of the size_t data type, that's a preprocessor operation using something like this.

#if (__SIZEOF_SIZE_T__ == 4)

This only works because you assume that the preprocessor predefined __SIZEOF_SIZE_T__ to be the value value as the actual value of sizeof(size_t). Once the preprocessor step is done, this "#if" directive is gone, and all instances of "__SIZEOF_SIZE_T__" have also been replaced and all the header files merged in with your code. It's this new code that the C compiler actually sees. (BTW, you can actually use the C preprocessor independent of the C compiler; I've used it with assembler code, for example.)

If you want to _execute_ different code depending on the size of the "size_t" type then you write C code something like this.

if (sizeof(size_t) == 4)

Once this is compiled, it will be executed at run-time.

(Well, not really: the GNU compiler is really clever about optimization, and it will know that this conditional statement is always true -- or never true, depending on the underlying machine architecture -- and optimize the statement away. But you get the idea.)

Anonymous said...

Thanks John for your fast and very complete answer.

I've only used the preprocessor output once in the past, to find a bug in an x-macro declaration I wasn't able to find by watching the code.

I guess the learning process is a never ending endeavour, and fortunately there are people like you (the gurus) that help us (the mere mortals) to fill-up our gaps.

Best regards,
Fernando