question about Zend MM

13 years ago by Adi Mutu — view source

unread

Hello,

I want to understand how Zend MM works, so i'm looking trought the sources and i see this:

#define ZEND_MM_ALIGNMENT_MASK ~(ZEND_MM_ALIGNMENT-1) #define ZEND_MM_ALIGNED_SIZE(size) (((size) + ZEND_MM_ALIGNMENT - 1) & ZEND_MM_ALIGNMENT_MASK)

I understand that the first define will create something like 11111000 ( it will clear last 3 bits)
but what does the 2nd define? Before clearing the last 3 bytes why does it add ZEND_MM_ALIGNMENT- 1
to size?

Thanks,

13 years ago by Gustavo Lopes — view source

unread

I want to understand how Zend MM works, so i'm looking trought the
sources and i see this:

#define ZEND_MM_ALIGNMENT_MASK ~(ZEND_MM_ALIGNMENT-1) #define
ZEND_MM_ALIGNED_SIZE(size) (((size) + ZEND_MM_ALIGNMENT - 1) &
ZEND_MM_ALIGNMENT_MASK)

I understand that the first define will create something like 11111000 (
it will clear last 3 bits)
but what does the 2nd define? Before clearing the last 3 bytes why does
it add ZEND_MM_ALIGNMENT- 1
to size?

It basically just rounds to the next multiple of ZEND_MM_ALIGNMENT,
assuming ZEND_MM_ALIGNMENT is a power of 2.

ZEND_MM_ALIGNMENT is a power of 2, so it has 1 bit set. Subtracting 1 will
zero that bit and and flip on all the other less significant bits. Then
negating flips the bits so that now the bits less significant than the
log2(ZEND_MM_ALIGNMENT)-th will be zero and the others will be one.

ZEND_MM_ALIGNED_SIZE adds ZEND_MM_ALIGNMENT - 1 and applies the mask. The
effect is that the result will be >= size and it will be a multiple of
ZEND_MM_ALIGNMENT (in particular the smallest multiple of
ZEND_MM_ALIGNMENT that's >= the argument) because the bits less
significant than the log2(ZEND_MM_ALIGNMENT)-th will be zero. "a & n-1" is
the same as "a mod n" with n being a power of 2. So if a &
ZEND_MM_ALIGNMENT - 1 == 0 then a mod ZEND_MM_ALIGNMENT == 0 and a is a
multiple of ZEND_MM_ALIGNMENT.

--
Gustavo Lopes

13 years ago by Adi Mutu — view source

unread

Ah, now i got it......so it can also return size, if size is directly a multiple of ZEND_MM_ALIGNMENT.
I was convinced that it has to be stricly > than size, i guess i was wrong.

From: Gustavo Lopes glopes@nebm.ist.utl.pt
To: Adi Mutu adi_mutu06@yahoo.com; internals@lists.php.net
Sent: Thursday, March 1, 2012 6:44 PM
Subject: Re: [PHP-DEV] question about Zend MM

I want to understand how Zend MM works, so i'm looking trought the sources and i see this:

#define ZEND_MM_ALIGNMENT_MASK ~(ZEND_MM_ALIGNMENT-1) #define ZEND_MM_ALIGNED_SIZE(size) (((size) + ZEND_MM_ALIGNMENT - 1) & ZEND_MM_ALIGNMENT_MASK)

I understand that the first define will create something like 11111000 ( it will clear last 3 bits)
but what does the 2nd define? Before clearing the last 3 bytes why does it add ZEND_MM_ALIGNMENT- 1
to size?

It basically just rounds to the next multiple of ZEND_MM_ALIGNMENT, assuming ZEND_MM_ALIGNMENT is a power of 2.

ZEND_MM_ALIGNMENT is a power of 2, so it has 1 bit set. Subtracting 1 will zero that bit and and flip on all the other less significant bits. Then negating flips the bits so that now the bits less significant than the log2(ZEND_MM_ALIGNMENT)-th will be zero and the others will be one.

ZEND_MM_ALIGNED_SIZE adds ZEND_MM_ALIGNMENT - 1 and applies the mask. The effect is that the result will be >= size and it will be a multiple of ZEND_MM_ALIGNMENT (in particular the smallest multiple of ZEND_MM_ALIGNMENT that's >= the argument) because the bits less significant than the log2(ZEND_MM_ALIGNMENT)-th will be zero. "a & n-1" is the same as "a mod n" with n being a power of 2. So if a & ZEND_MM_ALIGNMENT - 1 == 0 then a mod ZEND_MM_ALIGNMENT == 0 and a is a multiple of ZEND_MM_ALIGNMENT.

--Gustavo Lopes

13 years ago by Richard Lynch — view source

unread

WILD GUESS ALERT!

I'm guessing that this is for byte-alignment on big-endian versus
little-endian...

So it's more like 11110000 and 00001111 as masks to flip-flop bytes by
some binary logic / magic.

The -1 is to "wrap" the byte to binary inversion.

You might want to grep the code and see how they are actually used.

That helps me a lot when I'm totally lost what the Magic Numbers are
used for.

Not knowing what the initial value of ZEND_MM_ALIGNMENT is in the
first place, this is just wild guess...

Hello,

I want to understand how Zend MM works, so i'm looking trought the
sources and i see this:

#define ZEND_MM_ALIGNMENT_MASK ~(ZEND_MM_ALIGNMENT-1) #define
ZEND_MM_ALIGNED_SIZE(size) (((size) + ZEND_MM_ALIGNMENT - 1) &
ZEND_MM_ALIGNMENT_MASK)

I understand that the first define will create something like 11111000
( it will clear last 3 bits)
but what does the 2nd define? Before clearing the last 3 bytes why
does it add ZEND_MM_ALIGNMENT- 1
to size?

Thanks,

--
brain cancer update:
http://richardlynch.blogspot.com/search/label/brain%20tumor
Donate:
https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=FS9NLTNEEKWBE