From: sbradshaw@micron.com (Sam Bradshaw)
Subject: [PATCH] NVMe: Reduce divide operations
Date: Thu, 20 Nov 2014 14:13:24 -0800 [thread overview]
Message-ID: <546E6784.6000609@micron.com> (raw)
There are several expensive divide operations in the submit and
completion paths that can be converted to less expensive arithmetic
and logical operations. Profiling shows significant drops in time
spent in nvme_alloc_iod() under common workloads as a result of this
change.
Patch is against Jens' for-3.19/drivers branch.
Signed-off-by: Sam Bradshaw <sbradshaw at micron.com>
---
diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index 9310fe5..a5e2ebc 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -360,8 +360,14 @@ static __le64 **iod_list(struct nvme_iod *iod)
*/
static int nvme_npages(unsigned size, struct nvme_dev *dev)
{
- unsigned nprps = DIV_ROUND_UP(size + dev->page_size, dev->page_size);
- return DIV_ROUND_UP(8 * nprps, dev->page_size - 8);
+ unsigned page_size = (1 << dev->page_shift);
+ unsigned nprps = (size >> dev->page_shift) + 1;
+
+ if (size & (page_size - 1))
+ nprps++;
+ if ((nprps << 3) < (page_size - 8))
+ return 1;
+ return DIV_ROUND_UP(nprps << 3, page_size - 8);
}
static struct nvme_iod *
@@ -384,7 +390,7 @@ nvme_alloc_iod(unsigned nseg, unsigned nbytes, struct nvme_dev *dev, gfp_t gfp)
void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
{
- const int last_prp = dev->page_size / 8 - 1;
+ const int last_prp = (1 << (dev->page_shift - 3)) - 1;
int i;
__le64 **list = iod_list(iod);
dma_addr_t prp_dma = iod->first_dma;
@@ -459,7 +465,7 @@ int nvme_setup_prps(struct nvme_dev *dev, struct nvme_iod *iod, int total_len,
__le64 **list = iod_list(iod);
dma_addr_t prp_dma;
int nprps, i;
- u32 page_size = dev->page_size;
+ u32 page_size = 1 << dev->page_shift;
length -= (page_size - offset);
if (length <= 0)
@@ -1416,7 +1422,7 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev)
aqa = nvmeq->q_depth - 1;
aqa |= aqa << 16;
- dev->page_size = 1 << page_shift;
+ dev->page_shift = page_shift;
dev->ctrl_config = NVME_CC_CSS_NVM;
dev->ctrl_config |= (page_shift - 12) << NVME_CC_MPS_SHIFT;
diff --git a/include/linux/nvme.h b/include/linux/nvme.h
index 258945f..f504fb8 100644
--- a/include/linux/nvme.h
+++ b/include/linux/nvme.h
@@ -100,9 +100,9 @@ struct nvme_dev {
char firmware_rev[8];
u32 max_hw_sectors;
u32 stripe_size;
- u32 page_size;
u16 oncs;
u16 abort_limit;
+ u8 page_shift;
u8 event_limit;
u8 vwc;
u8 initialized;
next reply other threads:[~2014-11-20 22:13 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-20 22:13 Sam Bradshaw [this message]
2014-11-20 23:41 ` [PATCH] NVMe: Reduce divide operations Keith Busch
2014-11-21 17:36 ` Sam Bradshaw (sbradshaw)
2014-11-21 13:50 ` Matthew Wilcox
2014-11-21 15:52 ` Keith Busch
2014-11-21 17:14 ` Matthew Wilcox
2014-11-21 17:38 ` Sam Bradshaw (sbradshaw)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=546E6784.6000609@micron.com \
--to=sbradshaw@micron.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.