[Cluster-devel] [PATCH] libgfs2: Improve rgblocks2bitblocks()

Andrew Price anprice at redhat.com
Fri Nov 18 21:28:10 UTC 2011


On 18/11/11 20:52, Bob Peterson wrote:
> ----- Original Message -----
> | This patch reworks the rgblocks2bitblocks function which was
> | inefficient, difficult to read and generally unwieldy.
> |
> | As this is core code from the days of yore and fsck.gfs2 depends on
> | it,
> | I made sure to test the new function extensively, comparing its
> | outputs
> | with the original function over a large range of values for rgblocks
> | (up
> | to 195312500) and valid block sizes between 512 and 4096.
> |
> | All call points have been updated and, as a nice side effect, the run
> | time of the function is greatly reduced.
> |
> | Signed-off-by: Andrew Price<anprice at redhat.com>
> | +	while (blks_rgrp + blks_meta * bitblocks<  ((rgblocks - bitblocks)
>
> Hi,
>
> The patch looks good. There's only thing I'd do differently:
> I know the implied arithmetic operator order is correct, but
> I still prefer to see parens around statements like the above
> just for clarity. e.g.
>
> +	while (blks_rgrp + (blks_meta * bitblocks)<  ((rgblocks - bitblocks)
>
> I guess that's more of a style thing.

Yeah, it is more readable that way, I'll change it.

Andy




More information about the Cluster-devel mailing list