[PATCH 2.6.19] Adding branch to remove possible unnecessary inst

[PATCH 2.6.19] Adding branch to remove possible unnecessary inst

[yc_zhou]

Function blk_queue_bounce_limit using dma flag to determine whether
assigned a certain value for member of request_queue_t. But the
assignment is unconditionally after the flag is set. It introduce
possible unnecessary instructions.


Signed-Off-By: Yingchao Zhou <yc_zhou@ncic.ac.cn>

---
--- linux-2.6.19/block/ll_rw_blk.c.orig 2006-12-31 13:35:22.307742904 +0800
+++ linux-2.6.19/block/ll_rw_blk.c 2006-12-31 13:34:05.032490528 +0800
@@ -606,11 +606,13 @@ void blk_queue_bounce_limit(request_queu
know of a way to test this here. */
if (bounce_pfn < (min_t(u64,0xffffffff,BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
dma = 1;
- q->bounce_pfn = max_low_pfn;
+ else
+ q->bounce_pfn = max_low_pfn;
#else
if (bounce_pfn < blk_max_low_pfn)
dma = 1;
- q->bounce_pfn = bounce_pfn;
+ else
+ q->bounce_pfn = bounce_pfn;
#endif
if (dma) {
init_emergency_isa_pool();

Re: [PATCH 2.6.19] Adding branch to remove possible unnecessary inst

[Jens Axboe]

On Sun, Dec 31 2006, yc_zhou wrote:
> Function blk_queue_bounce_limit using dma flag to determine whether
> assigned a certain value for member of request_queue_t. But the
> assignment is unconditionally after the flag is set. It introduce
> possible unnecessary instructions.

Your patch is white space damaged, it makes it hard to read and
impossible to apply.

Note that blk_queue_bounce_limit() is an initialization function, so not
much gained from optimizing on that.

-- 
Jens Axboe