linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Al Viro <viro@zeniv.linux.org.uk>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	David Sterba <dsterba@suse.com>,
	Miklos Szeredi <miklos@szeredi.hu>,
	Anton Altaparmakov <anton@tuxera.com>,
	David Howells <dhowells@redhat.com>,
	Matthew Wilcox <willy@infradead.org>,
	Pavel Begunkov <asml.silence@gmail.com>
Subject: [RFC PATCH 23/37] iov_iter: massage iterate_iovec and iterate_kvec to logics similar to iterate_bvec
Date: Sun,  6 Jun 2021 19:10:37 +0000	[thread overview]
Message-ID: <20210606191051.1216821-23-viro@zeniv.linux.org.uk> (raw)
In-Reply-To: <20210606191051.1216821-1-viro@zeniv.linux.org.uk>

Premature optimization is the root of all evil...  Trying
to unroll the first pass through the loop makes it harder
to follow and not just for readers - compiler ends up
generating worse code than it would on a "non-optimized"
loop.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 lib/iov_iter.c | 91 +++++++++++++++++++++++-----------------------------------
 1 file changed, 36 insertions(+), 55 deletions(-)

diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index ebb907c6393c..578f40788943 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -16,55 +16,44 @@
 
 #define PIPE_PARANOIA /* for now */
 
-#define iterate_iovec(i, n, __v, __p, skip, STEP) {	\
-	size_t left;					\
-	size_t wanted = n;				\
-	__p = i->iov;					\
-	__v.iov_len = min(n, __p->iov_len - skip);	\
-	if (likely(__v.iov_len)) {			\
-		__v.iov_base = __p->iov_base + skip;	\
-		left = (STEP);				\
-		__v.iov_len -= left;			\
-		skip += __v.iov_len;			\
-		n -= __v.iov_len;			\
-	} else {					\
-		left = 0;				\
-	}						\
-	while (unlikely(!left && n)) {			\
-		__p++;					\
-		__v.iov_len = min(n, __p->iov_len);	\
-		if (unlikely(!__v.iov_len))		\
-			continue;			\
-		__v.iov_base = __p->iov_base;		\
-		left = (STEP);				\
-		__v.iov_len -= left;			\
-		skip = __v.iov_len;			\
-		n -= __v.iov_len;			\
-	}						\
-	n = wanted - n;					\
+#define iterate_iovec(i, n, __v, __p, skip, STEP) {		\
+	size_t left;						\
+	size_t wanted = n;					\
+	__p = i->iov;						\
+	do {							\
+		__v.iov_len = min(n, __p->iov_len - skip);	\
+		if (likely(__v.iov_len)) {			\
+			__v.iov_base = __p->iov_base + skip;	\
+			left = (STEP);				\
+			__v.iov_len -= left;			\
+			skip += __v.iov_len;			\
+			n -= __v.iov_len;			\
+			if (skip < __p->iov_len)		\
+				break;				\
+		}						\
+		__p++;						\
+		skip = 0;					\
+	} while (n);						\
+	n = wanted - n;						\
 }
 
-#define iterate_kvec(i, n, __v, __p, skip, STEP) {	\
-	size_t wanted = n;				\
-	__p = i->kvec;					\
-	__v.iov_len = min(n, __p->iov_len - skip);	\
-	if (likely(__v.iov_len)) {			\
-		__v.iov_base = __p->iov_base + skip;	\
-		(void)(STEP);				\
-		skip += __v.iov_len;			\
-		n -= __v.iov_len;			\
-	}						\
-	while (unlikely(n)) {				\
-		__p++;					\
-		__v.iov_len = min(n, __p->iov_len);	\
-		if (unlikely(!__v.iov_len))		\
-			continue;			\
-		__v.iov_base = __p->iov_base;		\
-		(void)(STEP);				\
-		skip = __v.iov_len;			\
-		n -= __v.iov_len;			\
-	}						\
-	n = wanted;					\
+#define iterate_kvec(i, n, __v, __p, skip, STEP) {		\
+	size_t wanted = n;					\
+	__p = i->kvec;						\
+	do {							\
+		__v.iov_len = min(n, __p->iov_len - skip);	\
+		if (likely(__v.iov_len)) {			\
+			__v.iov_base = __p->iov_base + skip;	\
+			(void)(STEP);				\
+			skip += __v.iov_len;			\
+			n -= __v.iov_len;			\
+			if (skip < __p->iov_len)		\
+				break;				\
+		}						\
+		__p++;						\
+		skip = 0;					\
+	} while (n);						\
+	n = wanted - n;						\
 }
 
 #define iterate_bvec(i, n, __v, __bi, skip, STEP) {	\
@@ -123,10 +112,6 @@
 			const struct iovec *iov;		\
 			struct iovec v;				\
 			iterate_iovec(i, n, v, iov, skip, (I))	\
-			if (skip == iov->iov_len) {		\
-				iov++;				\
-				skip = 0;			\
-			}					\
 			i->nr_segs -= iov - i->iov;		\
 			i->iov = iov;				\
 		} else if (iov_iter_is_bvec(i)) {		\
@@ -141,10 +126,6 @@
 			const struct kvec *kvec;		\
 			struct kvec v;				\
 			iterate_kvec(i, n, v, kvec, skip, (K))	\
-			if (skip == kvec->iov_len) {		\
-				kvec++;				\
-				skip = 0;			\
-			}					\
 			i->nr_segs -= kvec - i->kvec;		\
 			i->kvec = kvec;				\
 		} else if (iov_iter_is_xarray(i)) {		\
-- 
2.11.0


  parent reply	other threads:[~2021-06-06 19:12 UTC|newest]

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-06 19:07 [RFC][PATCHSET] iov_iter work Al Viro
2021-06-06 19:10 ` [RFC PATCH 01/37] ntfs_copy_from_user_iter(): don't bother with copying iov_iter Al Viro
2021-06-06 19:10   ` [RFC PATCH 02/37] generic_perform_write()/iomap_write_actor(): saner logics for short copy Al Viro
2021-06-06 19:10   ` [RFC PATCH 03/37] fuse_fill_write_pages(): don't bother with iov_iter_single_seg_count() Al Viro
2021-06-06 19:10   ` [RFC PATCH 04/37] iov_iter: Remove iov_iter_for_each_range() Al Viro
2021-06-06 19:10   ` [RFC PATCH 05/37] teach copy_page_to_iter() to handle compound pages Al Viro
2021-06-06 19:10   ` [RFC PATCH 06/37] copy_page_to_iter(): fix ITER_DISCARD case Al Viro
2021-06-06 19:10   ` [RFC PATCH 07/37] [xarray] iov_iter_fault_in_readable() should do nothing in xarray case Al Viro
2021-06-06 19:10   ` [RFC PATCH 08/37] iov_iter_advance(): use consistent semantics for move past the end Al Viro
2021-06-06 19:10   ` [RFC PATCH 09/37] iov_iter: switch ..._full() variants of primitives to use of iov_iter_revert() Al Viro
2021-06-06 19:10   ` [RFC PATCH 10/37] iov_iter: reorder handling of flavours in primitives Al Viro
2021-06-06 19:10   ` [RFC PATCH 11/37] iov_iter_advance(): don't modify ->iov_offset for ITER_DISCARD Al Viro
2021-06-06 19:10   ` [RFC PATCH 12/37] iov_iter: separate direction from flavour Al Viro
2021-06-06 19:10   ` [RFC PATCH 13/37] iov_iter: optimize iov_iter_advance() for iovec and kvec Al Viro
2021-06-06 19:10   ` [RFC PATCH 14/37] sanitize iov_iter_fault_in_readable() Al Viro
2021-06-06 19:10   ` [RFC PATCH 15/37] iov_iter_alignment(): don't bother with iterate_all_kinds() Al Viro
2021-06-06 19:10   ` [RFC PATCH 16/37] iov_iter_gap_alignment(): get rid of iterate_all_kinds() Al Viro
2021-06-09 13:01     ` Qian Cai
2021-06-09 18:06       ` Al Viro
2021-06-06 19:10   ` [RFC PATCH 17/37] get rid of iterate_all_kinds() in iov_iter_get_pages()/iov_iter_get_pages_alloc() Al Viro
2021-06-06 19:10   ` [RFC PATCH 18/37] iov_iter_npages(): don't bother with iterate_all_kinds() Al Viro
2021-06-06 19:10   ` [RFC PATCH 19/37] [xarray] iov_iter_npages(): just use DIV_ROUND_UP() Al Viro
2021-06-06 19:10   ` [RFC PATCH 20/37] iov_iter: replace iov_iter_copy_from_user_atomic() with iterator-advancing variant Al Viro
2021-06-06 19:10   ` [RFC PATCH 21/37] csum_and_copy_to_iter(): massage into form closer to csum_and_copy_from_iter() Al Viro
2021-06-06 19:10   ` [RFC PATCH 22/37] iterate_and_advance(): get rid of magic in case when n is 0 Al Viro
2021-06-06 19:10   ` Al Viro [this message]
2021-06-06 19:10   ` [RFC PATCH 24/37] iov_iter: unify iterate_iovec and iterate_kvec Al Viro
2021-06-06 19:10   ` [RFC PATCH 25/37] iterate_bvec(): expand bvec.h macro forest, massage a bit Al Viro
2021-06-06 19:10   ` [RFC PATCH 26/37] iov_iter: teach iterate_{bvec,xarray}() about possible short copies Al Viro
2021-06-06 19:10   ` [RFC PATCH 27/37] iov_iter: get rid of separate bvec and xarray callbacks Al Viro
2021-06-06 19:10   ` [RFC PATCH 28/37] iov_iter: make the amount already copied available to iterator callbacks Al Viro
2021-06-06 19:10   ` [RFC PATCH 29/37] iov_iter: make iterator callbacks use base and len instead of iovec Al Viro
2021-06-06 19:10   ` [RFC PATCH 30/37] pull handling of ->iov_offset into iterate_{iovec,bvec,xarray} Al Viro
2021-06-06 19:10   ` [RFC PATCH 31/37] iterate_xarray(): only of the first iteration we might get offset != 0 Al Viro
2021-06-06 19:10   ` [RFC PATCH 32/37] copy_page_to_iter(): don't bother with kmap_atomic() for bvec/kvec cases Al Viro
2021-06-06 19:10   ` [RFC PATCH 33/37] copy_page_from_iter(): don't need kmap_atomic() for kvec/bvec cases Al Viro
2021-06-06 19:10   ` [RFC PATCH 34/37] iov_iter: clean csum_and_copy_...() primitives up a bit Al Viro
2021-06-06 19:10   ` [RFC PATCH 35/37] pipe_zero(): we don't need no stinkin' kmap_atomic() Al Viro
2021-06-06 19:10   ` [RFC PATCH 36/37] clean up copy_mc_pipe_to_iter() Al Viro
2021-06-06 19:10   ` [RFC PATCH 37/37] csum_and_copy_to_pipe_iter(): leave handling of csum_state to caller Al Viro
2021-06-06 22:05 ` [RFC][PATCHSET] iov_iter work Linus Torvalds
2021-06-06 22:46   ` Linus Torvalds
2021-06-07  9:28     ` Christoph Hellwig
2021-06-07 14:43       ` Al Viro
2021-06-07 15:59         ` Christoph Hellwig
2021-06-07 21:07           ` Al Viro
2021-06-07 22:01             ` Linus Torvalds
2021-06-07 23:35               ` Linus Torvalds
2021-06-08  5:25                 ` Christoph Hellwig
2021-06-08 11:27                 ` Al Viro
2021-06-06 23:29   ` Al Viro
2021-06-07 10:38     ` Pavel Begunkov
2021-06-08 14:43 ` David Laight
2021-06-10 14:29 ` Qian Cai
2021-06-10 15:35   ` Al Viro
2021-06-10 15:48     ` Al Viro
2021-06-10 19:08     ` Qian Cai

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=20210606191051.1216821-23-viro@zeniv.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=anton@tuxera.com \
    --cc=asml.silence@gmail.com \
    --cc=dhowells@redhat.com \
    --cc=dsterba@suse.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=torvalds@linux-foundation.org \
    --cc=willy@infradead.org \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).