w*s
发帖数: 7227 | 1 Q:
Alignment issue in Arm is like this, correct me if i'm wrong,
if i have
char a[40];
int *b = a;
later on b could cause crash
From web,
For example, on ARM-based systems you cannot address a 32-bit word that
is not aligned to a 4-byte boundary. Doing so will result in an access
violation exception. On x86 you can access such non-aligned data, though the
performance suffers a little since two words have to fetched from memory
instead of just one.
So how to fix this easier, any compile options ? |
G*****h
发帖数: 33134 | 2 compiler should already handle it.
and *(int *)(&a[0]) should be fine
*(int *)(&a[1]) will crash for sure.
that
the
【在 w*s 的大作中提到】
: Q:
: Alignment issue in Arm is like this, correct me if i'm wrong,
:
: if i have
: char a[40];
: int *b = a;
: later on b could cause crash
:
: From web,
: For example, on ARM-based systems you cannot address a 32-bit word that
|
w*s
发帖数: 7227 | 3 well, we're porting code from CE to linux,
so we do have the
*(int *)(&a[1])
in CE which works fine, guess CE compiler is better.
any quick solution for us ?
【在 G*****h 的大作中提到】
: compiler should already handle it.
: and *(int *)(&a[0]) should be fine
: *(int *)(&a[1]) will crash for sure.
:
: that
: the
|
G*****h
发帖数: 33134 | 4 easy
rewrite the code wherever crashes.
though ARM usually has config register that you can enable byte boundary
access
【在 w*s 的大作中提到】
: well, we're porting code from CE to linux,
: so we do have the
: *(int *)(&a[1])
: in CE which works fine, guess CE compiler is better.
: any quick solution for us ?
|
w*s
发帖数: 7227 | 5 rewrite code will take time. i'm looking for some compiler options.
【在 G*****h 的大作中提到】
: easy
: rewrite the code wherever crashes.
: though ARM usually has config register that you can enable byte boundary
: access
|
G*****h
发帖数: 33134 | 6 there is CPU option
Compiler cannot stop you using *(int *)(&a[1])
【在 w*s 的大作中提到】
: rewrite code will take time. i'm looking for some compiler options.
|
w*s
发帖数: 7227 | 7 curious how winCE made it work, still the same ARM cpu ...
【在 G*****h 的大作中提到】
: there is CPU option
: Compiler cannot stop you using *(int *)(&a[1])
|
w*s
发帖数: 7227 | 8 sorry more questions on byte boundary access,
http://www.atmel.com/Images/doc6438.pdf
9.4.4 Memory Access
The ARM9EJ-S core supports byte (8-bit), half-word (16-bit) and word (32-bit
) access. Words
must be aligned to four-byte boundaries, half-words must be aligned to two-
byte boundaries and
bytes can be placed on any byte boundary.
so for the case of (int *)(&array[1]),
it's still treated as int, which must be aligned to four-byte boundaries.
does this conflict to what you are saying on "ARM usually has config
register that you can enable byte boundary access" ?
【在 G*****h 的大作中提到】
: easy
: rewrite the code wherever crashes.
: though ARM usually has config register that you can enable byte boundary
: access
|
h****d
发帖数: 1305 | 9 is it the limitation of gcc?
that
the
【在 w*s 的大作中提到】
: Q:
: Alignment issue in Arm is like this, correct me if i'm wrong,
:
: if i have
: char a[40];
: int *b = a;
: later on b could cause crash
:
: From web,
: For example, on ARM-based systems you cannot address a 32-bit word that
|
