[RFC] Alignment of fields in struct dentry

[Jörn Engel]

After taking a look at struct dentry, Arnd noted an alignment
problem.  The first four fields currently are:
	atomic_t d_count;
	unsigned int d_flags;		/* protected by d_lock */
	spinlock_t d_lock;		/* per dentry lock */
	struct inode *d_inode;		/* Where the name belongs to - NULL is
					 * negative */
On 64bit architectures, the first three take 12 bytes and d_inode is
not naturally aligned, so it can be aligned to byte 16.  This grows a
struct dentry from 196 to 200 Bytes (assuming no funky config options
like DEBUG_*, PROFILING or PREEMT && SMP are set).

One possible solution would be to exchange d_inode with d_mounted, but
I fear that d_inode would move from a hot cacheline to a cold one,
reducing performance.  Could there be a good solution or would any
rearrangement here only cause regressions?

Also, both 196 and 200 bytes are fairly close to 192 bytes, so I could
imagine performance improvements on 64bit machines with 64 Byte
cachelines.  Might it make sense to trim DNAME_INLINE_LEN_MIN by 4 or
8 bytes for such machines?

Jörn

-- 
The wise man seeks everything in himself; the ignorant man tries to get
everything from somebody else.
-- unknown
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/