On 7/20/2024 1:53 AM, Mateusz Guzik wrote:
On Wed, Jul 17, 2024 at 4:24 PM Yu Ma <yu.ma@xxxxxxxxx> wrote:
Skip 2-levels searching via find_next_zero_bit() when there is free slot in the
word contains next_fd, as:
(1) next_fd indicates the lower bound for the first free fd.
(2) There is fast path inside of find_next_zero_bit() when size<=64 to speed up
searching.
this is stale -- now the fast path searches up to 64 fds in the lower bitmap
Nope, this is still valid, as the searching size of the fast path inside
of find_next_fd() is always 64, it will execute the fast path inside of
find_next_zero_bit().
(3) After fdt is expanded (the bitmap size doubled for each time of expansion),
it would never be shrunk. The search size increases but there are few open fds
available here.
This fast path is proposed by Mateusz Guzik <mjguzik@xxxxxxxxx>, and agreed by
Jan Kara <jack@xxxxxxx>, which is more generic and scalable than previous
versions.
I think this paragraph is droppable. You already got an ack from Jan
below, so stating he agrees with the patch is redundant. As for me I
don't think this warrants mentioning. Just remove it, perhaps
Christian will be willing to massage it by himself to avoid another
series posting.
The idea of fast path for the word contains next_fd is from you,
although this patch is small, I think it is reasonable to record here
out of my respect. Appreciate for your guide and comments on this patch,
I've learned a lot on the way of resolving problems :)
Regards
Yu
And on top of patch 1 and 2, it improves pts/blogbench-1.1.0 read by
8% and write by 4% on Intel ICX 160 cores configuration with v6.10-rc7.
Reviewed-by: Jan Kara <jack@xxxxxxx>
Reviewed-by: Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx>
Signed-off-by: Yu Ma <yu.ma@xxxxxxxxx>
---
fs/file.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/fs/file.c b/fs/file.c
index 1be2a5bcc7c4..729c07a4fc28 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -491,6 +491,15 @@ static unsigned int find_next_fd(struct fdtable *fdt, unsigned int start)
unsigned int maxfd = fdt->max_fds; /* always multiple of BITS_PER_LONG */
unsigned int maxbit = maxfd / BITS_PER_LONG;
unsigned int bitbit = start / BITS_PER_LONG;
+ unsigned int bit;
+
+ /*
+ * Try to avoid looking at the second level bitmap
+ */
+ bit = find_next_zero_bit(&fdt->open_fds[bitbit], BITS_PER_LONG,
+ start & (BITS_PER_LONG - 1));
+ if (bit < BITS_PER_LONG)
+ return bit + bitbit * BITS_PER_LONG;
bitbit = find_next_zero_bit(fdt->full_fds_bits, maxbit, bitbit) * BITS_PER_LONG;
if (bitbit >= maxfd)
--
2.43.0
