macro help

[inftoconc@gmail.com]

Hi all,

I wrote a macro to read 8 bytes from a byte stream. Am not
sure if it is working ok. Can anyone please point out if there looks
to be a problem!

#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
8) + ( *(b+7))

Here b is a pointer to an unsigned char.

Thanks,

Vishal

[Ian Collins]

Re: macro help

inftoconc@gmail .com wrote:

Hi all,
>
I wrote a macro to read 8 bytes from a byte stream. Am not
sure if it is working ok. Can anyone please point out if there looks
to be a problem!
>

What did your test cases show you?

#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
8) + ( *(b+7))
>

Why on Earth did you write a totally horrible macro when a function can
do the job and is a kinder on the eye and debugger?

--
Ian Collins.

[Army1987]

Re: macro help

On Sun, 30 Sep 2007 09:26:36 +0000, inftoconc wrote:

Hi all,
>
I wrote a macro to read 8 bytes from a byte stream. Am not
sure if it is working ok. Can anyone please point out if there looks
to be a problem!
>
#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
8) + ( *(b+7))
>
Here b is a pointer to an unsigned char.

1. The syntax of the cast is (type-name)cast-expression, e.g.
(uint64_t)*b, not uint64_t(*b). Maybe you were thinking about C++?
2. What's wrong with good ol' b[0], b[1] etc.? They're somewhat
clearer than (*(b+1)).

Try:
#define READ64(b) ( ((uint64_t)(b)[0] << 56) + ((uint64_t)(b)[1] << 48) +\
((uint64_t)(b)[2] << 40) + etc... )
(Also, if you're ever going to port this to somewhere CHAR_BIT 8
you should mask each addend with & 0xFF, and if you aren't you
could add #if CHAR_BIT != 8
#error something
#endif
somewhere.)
--
Army1987 (Replace "NOSPAM" with "email")
A hamburger is better than nothing.
Nothing is better than eternal happiness.
Therefore, a hamburger is better than eternal happiness.

[Army1987]

Re: macro help

On Sun, 30 Sep 2007 04:14:19 -0700, William Pursell wrote:

On 30 Sep, 10:26, inftoc...@gmail .com wrote:

> I wrote a macro to read 8 bytes from a byte stream. Am not
>sure if it is working ok. Can anyone please point out if there looks
>to be a problem!
>>
>#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
>+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
>+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
>8) + ( *(b+7))
>>
>Here b is a pointer to an unsigned char.

>
Lots of problems:
>
The endianness of the machine will have
an impact, certainly.

I think he did that to have the same format on all machines.
Otherwise he could use memcpy().
--
Army1987 (Replace "NOSPAM" with "email")
A hamburger is better than nothing.
Nothing is better than eternal happiness.
Therefore, a hamburger is better than eternal happiness.

[Chris Torek]

Re: macro help

In article <46ff902c$0$159 55$426a74cc@new s.free.fr>
Charlie Gordon <news@chqrlie.o rgwrote:

>You SHOULD use a function for this, possibly an inline function.

Indeed, in some cases it turns out to be faster to call a regular
(non-inline) function than to do the code in line (either with a
macro, or with an inline function). (As always, the way to find
out is to measure it. If you do a lot of these conversions, it
may be worth looking at your compiler's output. Many of today's
ILP32-LL64 compilers produce poor 64-bit shift-and-mask code, so
this sort of micro-optimization is occasionally worthwhile.)

>Function:
>
>static inline uint64_t read64(unsigned char const *b) {
return ((uint64_t)b[0] << 56) | ((uint64_t)b[1] << 48) |
((uint64_t)b[2] << 40) | ((uint64_t)b[3] << 32) |
((uint64_t)b[4] << 24) | ((uint64_t)b[5] << 16) |
((uint64_t)b[6] << 8) | ((uint64_t)b[7] << 0);
>}
>
>If your compiler is not too smart, this might be better:
>
>static inline uint64_t read64(unsigned char const *b) {
return ((uint64_t)(((u int32_t)b[0] << 24) | ((uint32_t)b[1] << 16) |
((uint32_t)b[2] << 8) | b[3]) << 32) |
(((uint32_t)b[4] << 24) | ((uint32_t)b[5] << 16) |
((uint32_t)b[6] << 8) | b[7]);
>}

I find that applications that do this tend also to do 16 and/or
32-bit numbers, so for the second case, consider, e.g.:

static inline uint32_t read32(const unsigned char *b) {
return ((uint32_t)b[0] << 24) | ((uint32_t)b[1] << 16) |
((unsigned)b[2] << 8) | b[3];
}
static inline uint64_t read64(const unsigned char *b) {
return ((uint64_t)read 32(b) << 32) | read32(b + 4);
}

One might also consider also using "unsigned long" and "unsigned
long long" instead of the (possibly overspecified) uint32_t and
uint64_t aliases; and read32 can call read16() twice, and read16()
can use plain "unsigned int" (as the code in read32 above shows).
--
In-Real-Life: Chris Torek, Wind River Systems
Salt Lake City, UT, USA (40°39.22'N, 111°50.29'W) +1 801 277 2603
email: forget about it http://web.torek.net/torek/index.html
Reading email is like searching for food in the garbage, thanks to spammers.

[Charlie Gordon]

Re: macro help

"Thad Smith" <ThadSmith@acm. orga écrit dans le message de news:
47004884$0$9042 1$892e0abb@auth .newsreader.oct anews.com...

William Pursell wrote:

>On 30 Sep, 10:26, inftoc...@gmail .com wrote:

>> I wrote a macro to read 8 bytes from a byte stream. Am not
>>sure if it is working ok. Can anyone please point out if there looks
>>to be a problem!
>>>
>>#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
>>+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
>>+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
>>8) + ( *(b+7))
>>>
>>Here b is a pointer to an unsigned char.

>>

>

>What if a caller wants chars 2 through 9 from a buffer
>and calls the macro as:
>uint64_t x = READ64( b+2 );

>
It works properly. It would, however, fail with arguments such as b+1<<2.

No, you are mistaken: it does not work for b+2 either. It compiles, but
does not behave as intended. The MSB is read from b[0] and offset by 2.

--
Chqrlie.

[Charlie Gordon]

Re: macro help

"Charlie Gordon" <news@chqrlie.o rga écrit dans le message de news:
4700cef6$0$7288 $426a34cc@news. free.fr...

"Thad Smith" <ThadSmith@acm. orga écrit dans le message de news:
47004884$0$9042 1$892e0abb@auth .newsreader.oct anews.com...

>William Pursell wrote:

>>On 30 Sep, 10:26, inftoc...@gmail .com wrote:
>>> I wrote a macro to read 8 bytes from a byte stream. Am not
>>>sure if it is working ok. Can anyone please point out if there looks
>>>to be a problem!
>>>>
>>>#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
>>>+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
>>>+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
>>>8) + ( *(b+7))
>>>>
>>>Here b is a pointer to an unsigned char.
>>>

>>

>>What if a caller wants chars 2 through 9 from a buffer
>>and calls the macro as:
>>uint64_t x = READ64( b+2 );

>>
>It works properly. It would, however, fail with arguments such as
>b+1<<2.

>
No, you are mistaken: it does not work for b+2 either. It compiles, but
does not behave as intended. The MSB is read from b[0] and offset by 2.

Actually, it might compile as C++, but one has to fix the cast syntax for it
to compile as C ;-)

Complete answers have been posted on a different thread.

--
Chqrlie.

[Thad Smith]

Re: macro help

Charlie Gordon wrote:

"Charlie Gordon" <news@chqrlie.o rga écrit dans le message de news:
4700cef6$0$7288 $426a34cc@news. free.fr...

>"Thad Smith" <ThadSmith@acm. orga écrit dans le message de news:
>47004884$0$904 21$892e0abb@aut h.newsreader.oc tanews.com...

>>William Pursell wrote:
>>>On 30 Sep, 10:26, inftoc...@gmail .com wrote:
>>>> I wrote a macro to read 8 bytes from a byte stream. Am not
>>>>sure if it is working ok. Can anyone please point out if there looks
>>>>to be a problem!
>>>>>
>>>>#define READ64(b) ( (uint64_t(*b)) << 56) + ((uint64_t(*(b+ 1))) << 48)
>>>>+ ((uint64_t(*(b+ 2))) << 40) +((uint64_t(*(b +3))) << 32)+((uint64_t( *(b
>>>>+4))) << 24) + (( uint64_t(*(b+5) )) << 16) + (( uint64_t(*(b+6) )) <<
>>>>8) + ( *(b+7))
>>>>>
>>>>Here b is a pointer to an unsigned char.
>>>What if a caller wants chars 2 through 9 from a buffer
>>>and calls the macro as:
>>>uint64_t x = READ64( b+2 );
>>It works properly. It would, however, fail with arguments such as
>>b+1<<2.

>No, you are mistaken: it does not work for b+2 either. It compiles, but
>does not behave as intended. The MSB is read from b[0] and offset by 2.

>
Actually, it might compile as C++, but one has to fix the cast syntax for it
to compile as C ;-)

Thanks for correcting my blunders!


--
Thad