Rusty Russell [Wed, 1 Dec 2010 13:22:43 +0000 (23:52 +1030)]
tdb2: shrink free header from 32 to 24 bytes.
This reduces our minimum key+data length to 8 bytes; we do this by packing
the prev pointer where we used to put the flist pointer, and storing the
flist as an 8 bit index (meaning we can only have 256 free tables).
Note that this has a perverse result on the size of the database, as our
4-byte key and 4-byte data now fit perfectly in a minimal record, so
appeding causes us to allocate new records which are 50% larger,
since we detect growing.
Current results of speed test:
$ ./speed
1000000
Adding
1000000 records: 23210 ns (
59193360 bytes)
Finding
1000000 records: 2387 ns (
59193360 bytes)
Traversing
1000000 records: 2150 ns (
59193360 bytes)
Deleting
1000000 records: 13392 ns (
59193360 bytes)
Re-adding
1000000 records: 11546 ns (
59193360 bytes)
Appending
1000000 records: 29327 ns (
91193360 bytes)
Churning
1000000 records: 33026 ns (
91193360 bytes)
Previous:
$ ./speed
1000000
Adding
1000000 records: 28324 ns (
67232528 bytes)
Finding
1000000 records: 2468 ns (
67232528 bytes)
Traversing
1000000 records: 2200 ns (
67232528 bytes)
Deleting
1000000 records: 13083 ns (
67232528 bytes)
Re-adding
1000000 records: 16433 ns (
67232528 bytes)
Appending
1000000 records: 2511 ns (
67232528 bytes)
Churning
1000000 records: 31068 ns (
67570448 bytes)
Rusty Russell [Tue, 23 Nov 2010 03:10:59 +0000 (13:40 +1030)]
tdb2: rename set_header to the more appropriate set_used_header.
Rusty Russell [Wed, 1 Dec 2010 12:52:55 +0000 (23:22 +1030)]
tdb2: Add speed test to tdb and tdb2
Current results of speed test:
$ ./speed
1000000
Adding
1000000 records: 14726 ns (
67244816 bytes)
Finding
1000000 records: 2844 ns (
67244816 bytes)
Missing
1000000 records: 2528 ns (
67244816 bytes)
Traversing
1000000 records: 2572 ns (
67244816 bytes)
Deleting
1000000 records: 5358 ns (
67244816 bytes)
Re-adding
1000000 records: 9176 ns (
67244816 bytes)
Appending
1000000 records: 3035 ns (
67244816 bytes)
Churning
1000000 records: 18139 ns (
67565840 bytes)
$ ./speed 100000
Adding 100000 records: 13270 ns (
14349584 bytes)
Finding 100000 records: 2769 ns (
14349584 bytes)
Missing 100000 records: 2422 ns (
14349584 bytes)
Traversing 100000 records: 2595 ns (
14349584 bytes)
Deleting 100000 records: 5331 ns (
14349584 bytes)
Re-adding 100000 records: 5875 ns (
14349584 bytes)
Appending 100000 records: 2751 ns (
14349584 bytes)
Churning 100000 records: 20666 ns (
25771280 bytes)
vs tdb1 (with hashsize 100003):
$ ./speed
1000000
Adding
1000000 records: 8547 ns (
44306432 bytes)
Finding
1000000 records: 5595 ns (
44306432 bytes)
Missing
1000000 records: 3469 ns (
44306432 bytes)
Traversing
1000000 records: 4571 ns (
44306432 bytes)
Deleting
1000000 records: 12115 ns (
44306432 bytes)
Re-adding
1000000 records: 10505 ns (
44306432 bytes)
Appending
1000000 records: 10610 ns (
44306432 bytes)
Churning
1000000 records: 28697 ns (
44306432 bytes)
$ ./speed 100000
Adding 100000 records: 6030 ns (
4751360 bytes)
Finding 100000 records: 3141 ns (
4751360 bytes)
Missing 100000 records: 3143 ns (
4751360 bytes)
Traversing 100000 records: 4659 ns (
4751360 bytes)
Deleting 100000 records: 7891 ns (
4751360 bytes)
Re-adding 100000 records: 5913 ns (
4751360 bytes)
Appending 100000 records: 4242 ns (
4751360 bytes)
Churning 100000 records: 15300 ns (
4751360 bytes)
Rusty Russell [Wed, 1 Dec 2010 12:36:56 +0000 (23:06 +1030)]
tdb2: fix tdb_check() return when free table entries missing.
It mistakenly returned -1 meaning "success".
Rusty Russell [Wed, 1 Dec 2010 12:26:35 +0000 (22:56 +1030)]
tdb2: make tdb_check call check() function.
Rusty Russell [Wed, 1 Dec 2010 13:21:41 +0000 (23:51 +1030)]
tdb2: make summary command handle recovery "dead zone"
We can run summary with a recovery area, or a dead zone.
Rusty Russell [Wed, 1 Dec 2010 13:18:42 +0000 (23:48 +1030)]
tdb2: cancel transactions on tdb_close
Otherwise we leak memory.
Rusty Russell [Tue, 23 Nov 2010 01:44:31 +0000 (12:14 +1030)]
tdb2: enable transactions in tdbtorture
Rusty Russell [Tue, 23 Nov 2010 01:43:59 +0000 (12:13 +1030)]
tdb2: stricter ordering on expansion lock
It's problematic for transaction commit to get the expansion lock, but
in fact we always grab a hash lock before the transaction lock, so it
doesn't really need it (the transaction locks the entire database).
Assert that this is true, and fix up a few lowlevel tests where it wasn't.
Rusty Russell [Tue, 23 Nov 2010 01:40:32 +0000 (12:10 +1030)]
tdb2: remove all the dead code
I left much tdb1 code in various files for inspiration, and in case I needed
it later. Now we have all the major features implemented, remove it.
Rusty Russell [Tue, 23 Nov 2010 01:38:21 +0000 (12:08 +1030)]
tdb2: transaction support
This adds transactions to tdb2; the code is taken from tdb1 with minimal
modifications, as are the unit
Rusty Russell [Tue, 23 Nov 2010 01:37:29 +0000 (12:07 +1030)]
tdb2: allow nesting of read locks on top of write locks.
If we have a write lock and ask for a read lock, that's OK, but not the
other way around. tdb_nest_lock() allowed both, tdb_allrecord_lock() allowed
neither.
Rusty Russell [Tue, 23 Nov 2010 01:36:31 +0000 (12:06 +1030)]
ccanlint: fix -x core dump
This wasn't fixed when we converted to ccan/opt in
8d70667887.
Unfortunately, unistd.h defines optarg, so the compiler didn't catch
this.
Rusty Russell [Wed, 17 Nov 2010 09:56:52 +0000 (20:26 +1030)]
tdb2: handle chains of free tables
This adds chains of free tables: we choose one at random at the start and
we iterate through them when they are exhausted.
Currently there is no code to actually add a new free table, but we test
that we can handle it if we add one in future.
Rusty Russell [Wed, 17 Nov 2010 09:55:41 +0000 (20:25 +1030)]
tdb2: get rid of zones
Zones were a bad idea. They mean we can't simply add stuff to the end
of the file (which transactions relied upon), and there's no good heuristic
in how to size them.
This patch reverts us to a single free table.
Rusty Russell [Wed, 17 Nov 2010 10:00:07 +0000 (20:30 +1030)]
tdb2: fix bucket search
We were previously jumping straight from the first bucket to the end.
Rusty Russell [Wed, 17 Nov 2010 09:52:19 +0000 (20:22 +1030)]
tdb2: only adjust size once when growing
We were adding 50% to datalen twice, so move it out of adjust_size and
make the callers do it.
We also add a test that the heuristic is working at all.
Rusty Russell [Wed, 17 Nov 2010 09:50:35 +0000 (20:20 +1030)]
tdb2: remove tailer
We don't actually need it.
Rusty Russell [Mon, 15 Nov 2010 07:24:15 +0000 (17:54 +1030)]
tdb2: fix tdb_chainlock
We can't enlarge the lock without risking deadlock, so tdb_chainlock() can't
simply grab a one-byte lock; it needs to grab the lock we'd need to protect
the hash.
In theory, tdb_chainlock_read() could use a one-byte lock though.
Rusty Russell [Mon, 15 Nov 2010 07:25:45 +0000 (17:55 +1030)]
tdb2: fix coalesce race #3
When we're coalescing, we need to drop the lock on the current free list, as
we've enlarged the block and it may now belong in a different list.
Unfortunately (as shown by repeated tdbtorture -n 8) another coalescing run
can do the coalescing while we've dropped the lock. So for this case, we
use the TDB_COALESCING_MAGIC flag so it doesn't look free.
Rusty Russell [Mon, 15 Nov 2010 07:25:34 +0000 (17:55 +1030)]
tdb2: add TDB_COALESCING_MAGIC to solve coalescing race.
A new special record marker to indicate coalescing is in progress.
Rusty Russell [Mon, 15 Nov 2010 08:32:42 +0000 (19:02 +1030)]
tdb2: fix coalesce race #2
When we find a free block, we need to mark it as used before we drop the
free lock, even though we've removed it from the list. Otherwise the
coalescing code can find it.
This means handing the information we need down to lock_and_alloc, which
also means we know when we're grabbing a "growing" entry, and can relax
the heuristics about finding a good-sized block in that case.
Rusty Russell [Mon, 15 Nov 2010 07:10:47 +0000 (17:40 +1030)]
tdb2: coalescing race fix #1
When coalescing, we check the adjacent entry then lock its free list: we
need to *recheck* after locking, to make sure it's still in that free list.
Rusty Russell [Mon, 15 Nov 2010 07:19:11 +0000 (17:49 +1030)]
tdb2: minor optimization for set_header
We actually only need the bottom 5 bits of the hash value, so don't waste
8 bytes passing it.
Rusty Russell [Mon, 15 Nov 2010 07:26:57 +0000 (17:56 +1030)]
tdb2: hoist adjust_size
We're going to want it in get_free() in the next patch, so move it upwards.
Trivial changes, too: add to size before min length check, and rename growing
to want_extra.
Rusty Russell [Mon, 15 Nov 2010 07:07:03 +0000 (17:37 +1030)]
tdb2: clean up makefile for tools
Rusty Russell [Mon, 15 Nov 2010 07:31:48 +0000 (18:01 +1030)]
tdb2: extra debugging checks
Rusty Russell [Wed, 17 Nov 2010 09:59:13 +0000 (20:29 +1030)]
ccanlint: add ccanlint section to _info
This supersedes the previous Fails: section, into a more general set of
lines of form:
<testname> <option>...
With the special <option> "FAIL" to mean we know we fail this test.
We accept options to valgrind-tests; in particular tdb2 wants
--partial-loads-ok=yes passed to valgrind.
Rusty Russell [Wed, 17 Nov 2010 09:57:37 +0000 (20:27 +1030)]
ccanlint: override _info's Fails: with --target
I wanted to see what happened with tdb2's valgrind test (suppressed in the
_info file).
Rusty Russell [Wed, 17 Nov 2010 09:53:30 +0000 (20:23 +1030)]
ccanlint: run tests in alphabetical order
This allows tests (like in tdb2) to run the simpler tests first and build up,
making it easier to spot the root cause of errors.
Rusty Russell [Mon, 15 Nov 2010 09:46:40 +0000 (20:16 +1030)]
config: add HAVE_BUILTIN_FFSLL
Rusty Russell [Mon, 15 Nov 2010 02:47:55 +0000 (13:17 +1030)]
nfs: remove trailing whitespace
ccanlint complained about it; it's trivial but fixing it adds one point to
our score.
Rusty Russell [Mon, 15 Nov 2010 02:47:12 +0000 (13:17 +1030)]
nfs: make headers idempotent
ccanlint did this automatically.
Rusty Russell [Mon, 15 Nov 2010 02:46:46 +0000 (13:16 +1030)]
nfs: Remove _U_, use ccan/compiler and UNUSED
Also, add Author and License lines to _info (ccanlint complained).
Now ccanlint can build it as a module.
Rusty Russell [Mon, 15 Nov 2010 02:45:36 +0000 (13:15 +1030)]
nfs: Add _info, remove -D_FILE_OFFSET_BITS=64, use nfs_off_t
This makes it closer to compiling under ccanlint.
Rusty Russell [Mon, 15 Nov 2010 02:00:48 +0000 (12:30 +1030)]
ccanlint: remove -Werror, use output of compile command to detect warnings.
This means we can pass various compulation tests, but deduct points if
they get warnings.
Rusty Russell [Sun, 14 Nov 2010 11:40:04 +0000 (22:10 +1030)]
ccanlint: make compile commands return output.
We want to distinguish between warnings and errors: the first step is to
return the output even if the command doesn't fail.
Rusty Russell [Mon, 15 Nov 2010 02:23:00 +0000 (12:53 +1030)]
ccanlint: fix idempotent handler
Test for inserting idempotent header was wrong way around, and we check
all headers at once, rather than finishing after one.
Also, turn - into _ rather than removing.
Rusty Russell [Sun, 14 Nov 2010 11:53:52 +0000 (22:23 +1030)]
ccanlint: only print 5 lines of output unless -vv
Rusty Russell [Mon, 15 Nov 2010 02:29:52 +0000 (12:59 +1030)]
ccanlint: print error information even if we pass.
This means we can see messages even if we don't fail; ie. for compiler warnings
once they are no longer fatal.
This means that various tests have to be more careful in not setting
score->error.
Rusty Russell [Sun, 14 Nov 2010 11:48:12 +0000 (22:18 +1030)]
ccanlint: minor print formatting cleanup.
Rusty Russell [Mon, 15 Nov 2010 02:48:57 +0000 (13:18 +1030)]
ccanlint: don't skip every second question
We only grabbed one character at a time from stdin, so "y<return>" answered
the next question as well.
Rusty Russell [Thu, 11 Nov 2010 02:18:09 +0000 (12:48 +1030)]
ciniparser: add LICENSE symlink
Rusty Russell [Wed, 10 Nov 2010 12:45:14 +0000 (23:15 +1030)]
foreach, iscsi, jbitset, jmap, opt, rbtree, sparse_bsearch, tally, tdb2: add LICENSE symlinks
Rusty Russell [Wed, 10 Nov 2010 12:44:30 +0000 (23:14 +1030)]
ccanlint: check license and LICENSE symlink.
Michael Ellerman noted that ccan/opt has no symlink, so add ccanlint test.
Rusty Russell [Wed, 10 Nov 2010 12:16:12 +0000 (22:46 +1030)]
ccanlint: clarify different -v levels.
Rusty Russell [Wed, 10 Nov 2010 12:08:02 +0000 (22:38 +1030)]
ccanlint: fix --target=examples-compile
Rusty Russell [Wed, 10 Nov 2010 11:51:06 +0000 (22:21 +1030)]
ccanlint: use isspace instead of isblank
I am told that CentOS 5.3 doesn't like isblank (it's ISO C).
Rusty Russell [Wed, 10 Nov 2010 11:40:03 +0000 (22:10 +1030)]
ccan_tokenizer, check_type, container_of, typesafe_cb: handle !HAVE_TYPEOF
Delio Brignoli points out that check_type fails when HAVE_TYPEOF is 0,
so turn that off here and see what else breaks...
Rusty Russell [Wed, 10 Nov 2010 11:38:15 +0000 (22:08 +1030)]
ccanlint: typo fix and fix errant description parsing.
"C has fairly weak typing:" from check_type/_info is not a new section heading!
Enforce that each word in multi-word sections must be caps.
Rusty Russell [Wed, 10 Nov 2010 06:38:29 +0000 (17:08 +1030)]
ccanlint: tweak example compilation.
We used to prefer to build up examples using previous ones, but this broke
for tests which were to be executed. But list.h relied on the buildup
behavior, so return to that unless we see "// Given" comments indicating
an execution test.
Rusty Russell [Wed, 10 Nov 2010 06:37:29 +0000 (17:07 +1030)]
list: fix uninitialized variable in list_add example.
gcc with optimization complained (correctly) about this.
Rusty Russell [Wed, 10 Nov 2010 06:25:24 +0000 (16:55 +1030)]
ccanlint: list file errors in order they are encountered.
Rusty Russell [Wed, 10 Nov 2010 06:24:59 +0000 (16:54 +1030)]
ccanlint: fix abort when -vv used and examples don't compile.
Rusty Russell [Wed, 10 Nov 2010 05:18:31 +0000 (15:48 +1030)]
Makefile: exclude Judy-depending libraries as well as nfs
(nfs gives warnings on build, make ccanlint unhappy).
Rusty Russell [Wed, 10 Nov 2010 05:17:12 +0000 (15:47 +1030)]
Merge branch 'ronnie'
Rusty Russell [Tue, 9 Nov 2010 22:35:58 +0000 (09:05 +1030)]
compiler: shorten names of attributes, add UNUSED
The long names were unwieldy in practice; at risk of clashing, replace
with shorter versions.
Rusty Russell [Tue, 9 Nov 2010 11:37:58 +0000 (22:07 +1030)]
nfs: add generates C files
ccanlint won't run rpcgen, so we need the C source in the repo anyway.
Rusty Russell [Tue, 9 Nov 2010 08:24:45 +0000 (18:54 +1030)]
nfs: ccanize a little more.
rename libnfs.h to nfs.h (CCAN expects "main" header to match module name).
tools move into tools/, .x file move into rpc/
Don't sed the rpcgen files, generate headers in rpc/
Make tools use <ccan/nfs/nfs.h> as per normal ccan usage.
Rusty Russell [Tue, 9 Nov 2010 08:04:21 +0000 (18:34 +1030)]
nfs: initial import.
Another Ronnie module!
Rusty Russell [Tue, 9 Nov 2010 02:44:39 +0000 (13:14 +1030)]
ccanlint: rework so checks have more structure.
Previously each check returned a void *, but in fact most of them fell into
similar patterns. So define 'struct score' and a helper to add files to it,
and use that.
Under these rules, you get 0/1 if you skip a test because a dependency failed
which in theory means your score (as a percentage) could drop if you fix
a test.
Rusty Russell [Mon, 8 Nov 2010 22:20:21 +0000 (08:50 +1030)]
ccanlint: fix up creation of test/run.c
Make test directory correctly, and #include <ccan/...> not #include "".
Rusty Russell [Mon, 8 Nov 2010 11:35:42 +0000 (22:05 +1030)]
likely: switch from using ccan/hashtable to ccan/htable
Rusty Russell [Mon, 8 Nov 2010 11:30:04 +0000 (22:00 +1030)]
rbtree: new module from Ronnie.
Rusty Russell [Mon, 8 Nov 2010 11:27:41 +0000 (21:57 +1030)]
hashtable: replaced by htable.
Rusty Russell [Mon, 8 Nov 2010 10:14:06 +0000 (20:44 +1030)]
talloc: define TALLOC_CTX
I'm not a fan of the SHOUTING, but it's upstream.
Rusty Russell [Mon, 8 Nov 2010 09:59:17 +0000 (20:29 +1030)]
iscsi: add LICENCE link
Rusty Russell [Mon, 8 Nov 2010 09:54:47 +0000 (20:24 +1030)]
iscsi: ccanize a little more, add silly simple test case.
Rusty Russell [Mon, 8 Nov 2010 09:19:41 +0000 (19:49 +1030)]
iscsi: new module from Ronnie.
Rusty Russell [Mon, 8 Nov 2010 03:02:51 +0000 (13:32 +1030)]
htable: push capacity limit from 66 to 75%
With the extra bits, long runs don't hurt our cache very much on search,
so we can pack quite a few in. Here are the runs at maximal density before
and after:
Before:
$ ./speed
3145000
Initial insert: 248 ns
Details: hash size
4194304, mask bits 9, perfect 63%
Initial lookup (match): 122 ns
Initial lookup (miss): 142 ns
Initial lookup (random): 170 ns
Initial delete all: 134 ns
Details: rehashes
3145000
Initial re-inserting: 149 ns
Deleting first half: 73 ns
Details: rehashes
1572500, delete markers
1572500
Adding (a different) half: 128 ns
Details: delete markers 0, perfect 62%
Lookup after half-change (match): 129 ns
Lookup after half-change (miss): 145 ns
Details: initial churn
Churning second time: 703 ns
Churning third time: 725 ns
Churning fourth time: 717 ns
Churning fifth time: 710 ns
Details: reinserting with spread
Details: delete markers 149261, perfect 57%
Details: worst run 254 (2 deleted)
Lookup after churn & spread (match): 132 ns
Lookup after churn & spread (miss): 159 ns
Lookup after churn & spread (random): 184 ns
Deleting half after churn & spread: 71 ns
Adding (a different) half after churn & spread: 129 ns
Details: delete markers 0, perfect 62%
After:
$ ./speed
3145727
Initial insert: 232 ns
Details: hash size
4194304, mask bits 9, perfect 63%
Initial lookup (match): 122 ns
Initial lookup (miss): 141 ns
Initial lookup (random): 234 ns
Initial delete all: 129 ns
Details: rehashes
3145727
Initial re-inserting: 153 ns
Deleting first half: 80 ns
Details: rehashes
1572864, delete markers
1572864
Adding (a different) half: 137 ns
Details: delete markers 0, perfect 62%
Lookup after half-change (match): 125 ns
Lookup after half-change (miss): 145 ns
Details: initial churn
Churning second time: 702 ns
Churning third time: 719 ns
Churning fourth time: 712 ns
Churning fifth time: 709 ns
Details: reinserting with spread
Details: delete markers 169474, perfect 56%
Details: worst run 248 (12 deleted)
Lookup after churn & spread (match): 129 ns
Lookup after churn & spread (miss): 159 ns
Lookup after churn & spread (random): 242 ns
Deleting half after churn & spread: 70 ns
Adding (a different) half after churn & spread: 133 ns
Details: delete markers 0, perfect 62%
Rusty Russell [Mon, 8 Nov 2010 05:54:31 +0000 (16:24 +1030)]
htable: restore perfect bit when resizing.
If the lower bit differs between pointers, we can lose our "perfect"
bit. Since we're rehashing the entire thing anyway, we can pick
another bit when we double the hash table.
Rusty Russell [Mon, 8 Nov 2010 05:54:06 +0000 (16:24 +1030)]
htable: use "perfect" bit to reduce rehashes.
Steal one bit to indicate an entry is in its perfect position. This
means we don't have to rehash it when we are rehashing the entire
table to get rid of deleted records, and we can still use it for
comparison during lookup.
Before: $ ./speed
50000000
Initial insert: 462 ns
Details: hash size
134217728, mask bits 5, perfect 81%
Initial lookup (match): 161 ns
Initial lookup (miss): 168 ns
Initial lookup (random): 326 ns
Initial delete all: 164 ns
Details: rehashes
50000000
Initial re-inserting: 167 ns
Deleting first half: 86 ns
Details: rehashes
25000000, delete markers
25000000
Adding (a different) half: 217 ns
Details: delete markers 0, perfect 81%
Lookup after half-change (match): 169 ns
Lookup after half-change (miss): 180 ns
Details: initial churn
Churning second time: 593 ns
Churning third time: 611 ns
Churning fourth time: 619 ns
Churning fifth time: 622 ns
Details: reinserting with spread
Details: delete markers
13800468, perfect 73%
Details: worst run 48 (4 deleted)
Lookup after churn & spread (match): 192 ns
Lookup after churn & spread (miss): 211 ns
Lookup after churn & spread (random): 373 ns
Deleting half after churn & spread: 103 ns
Adding (a different) half after churn & spread: 102 ns
Details: delete markers
29539689, perfect 72%
After:
Initial insert: 467 ns
Details: hash size
134217728, mask bits 5, perfect 81%
Initial lookup (match): 163 ns
Initial lookup (miss): 171 ns
Initial lookup (random): 326 ns
Initial delete all: 170 ns
Details: rehashes
50000000
Initial re-inserting: 169 ns
Deleting first half: 88 ns
Details: rehashes
25000000, delete markers
25000000
Adding (a different) half: 141 ns
Details: delete markers 0, perfect 81%
Lookup after half-change (match): 166 ns
Lookup after half-change (miss): 171 ns
Details: initial churn
Churning second time: 441 ns
Churning third time: 469 ns
Churning fourth time: 466 ns
Churning fifth time: 490 ns
Details: reinserting with spread
Details: delete markers
13800468, perfect 73%
Details: worst run 48 (4 deleted)
Lookup after churn & spread (match): 197 ns
Lookup after churn & spread (miss): 209 ns
Lookup after churn & spread (random): 369 ns
Deleting half after churn & spread: 98 ns
Adding (a different) half after churn & spread: 100 ns
Details: delete markers
29539689, perfect 72%
Rusty Russell [Mon, 8 Nov 2010 05:53:17 +0000 (16:23 +1030)]
htable: rehash to clear delete markers
Delete markers build up over time, even if we try to clean as we go.
So just garbage collect them when there are too many.
Before:
rusty@vivaldi:~/devel/cvs/ccan/ccan/htable/tools (htable)$ ./speed
50000000
Initial insert: 467 ns
Details: hash size
134217728, mask bits 5, perfect 81%
Initial lookup (match): 160 ns
Initial lookup (miss): 169 ns
Initial lookup (random): 328 ns
Initial delete all: 165 ns
Details: rehashes
50000000
Initial re-inserting: 317 ns
Deleting first half: 86 ns
Details: rehashes
25000000, delete markers
25000000
Adding (a different) half: 89 ns
Details: delete markers
18894324, perfect 76%
Lookup after half-change (match): 174 ns
Lookup after half-change (miss): 203 ns
Details: initial churn
Churning second time: 816 ns
Churning third time: 615 ns
Churning fourth time: 621 ns
Churning fifth time: 846 ns
Details: reinserting with spread
Details: delete markers
11078719, perfect 74%
Details: worst run 48 (4 deleted)
Lookup after churn & spread (match): 191 ns
Lookup after churn & spread (miss): 208 ns
Lookup after churn & spread (random): 374 ns
Deleting half after churn & spread: 102 ns
Adding (a different) half after churn & spread: 103 ns
Details: delete markers
27442234, perfect 73%
After:
Initial insert: 462 ns
Details: hash size
134217728, mask bits 5, perfect 81%
Initial lookup (match): 161 ns
Initial lookup (miss): 168 ns
Initial lookup (random): 326 ns
Initial delete all: 164 ns
Details: rehashes
50000000
Initial re-inserting: 167 ns
Deleting first half: 86 ns
Details: rehashes
25000000, delete markers
25000000
Adding (a different) half: 217 ns
Details: delete markers 0, perfect 81%
Lookup after half-change (match): 169 ns
Lookup after half-change (miss): 180 ns
Details: initial churn
Churning second time: 593 ns
Churning third time: 611 ns
Churning fourth time: 619 ns
Churning fifth time: 622 ns
Details: reinserting with spread
Details: delete markers
13800468, perfect 73%
Details: worst run 48 (4 deleted)
Lookup after churn & spread (match): 192 ns
Lookup after churn & spread (miss): 211 ns
Lookup after churn & spread (random): 373 ns
Deleting half after churn & spread: 103 ns
Adding (a different) half after churn & spread: 102 ns
Details: delete markers
29539689, perfect 72%
Rusty Russell [Mon, 8 Nov 2010 05:51:20 +0000 (16:21 +1030)]
htable: speed benchmark
Rusty Russell [Mon, 8 Nov 2010 05:52:19 +0000 (16:22 +1030)]
htable: first implementation
Rusty Russell [Mon, 8 Nov 2010 05:51:06 +0000 (16:21 +1030)]
jmap: a speed benchmark
Rusty Russell [Mon, 8 Nov 2010 05:51:58 +0000 (16:21 +1030)]
jmap: fix incorrect prototype in header documentation.
Rusty Russell [Sun, 7 Nov 2010 01:45:10 +0000 (12:15 +1030)]
jmap: just use unsigned long, not size_t
Judy.h uses "Word_t" which it defines to an "unsigned long", so just use that.
If I were writing Judy from scratch, I'd use size_t or uintptr_t.
Rusty Russell [Sun, 7 Nov 2010 01:44:55 +0000 (12:14 +1030)]
jbitset: just use unsigned long, not size_t
Judy.h uses "Word_t" which it defines to an "unsigned long", so just use that.
If I were writing Judy from scratch, I'd use size_t or uintptr_t.
Rusty Russell [Sun, 7 Nov 2010 01:42:31 +0000 (12:12 +1030)]
ccanlint: fix compilation of chained examples.
Broken in
5378c864f9c37f39d906f599285da25a7db0c9fe: we were trying to
recompile the previous failed version. And the message we printed out
was unclear.
Rusty Russell [Thu, 4 Nov 2010 10:22:42 +0000 (20:52 +1030)]
antithread, foreach, grab_file, hashtable, str_talloc: fix _info depends.
For str_talloc and grab_file, avoid using ccan/str (it's just for tests).
Rusty Russell [Thu, 4 Nov 2010 10:14:00 +0000 (20:44 +1030)]
ccanlint: add --target
Much easier to run just a single test across the tree.
Rusty Russell [Thu, 4 Nov 2010 10:13:17 +0000 (20:43 +1030)]
ccanlint: add test to look for modules we use but don't list in _info.
Rusty Russell [Thu, 4 Nov 2010 10:11:44 +0000 (20:41 +1030)]
ccanlint: fix wrong initialization in example compilation.
I moved the code around and missed this.
Rusty Russell [Thu, 4 Nov 2010 10:10:52 +0000 (20:40 +1030)]
Makefile: make non-optimized variant comment a bit clearer.
Rusty Russell [Thu, 4 Nov 2010 04:32:13 +0000 (15:02 +1030)]
alignof, array_size, crc: provide comments to allow testing examples.
Rusty Russell [Thu, 4 Nov 2010 03:25:15 +0000 (13:55 +1030)]
ccanlint: try running example code.
Comments of form "// [given x] outputs y" we can check the examples do as expected.
Rusty Russell [Thu, 4 Nov 2010 02:11:13 +0000 (12:41 +1030)]
opt: fix junk after string in opt_show_charp.
Rusty Russell [Wed, 3 Nov 2010 22:34:58 +0000 (09:04 +1030)]
cleanup: get rid of unwanted tap/ toplevel which crept in.
Rusty Russell [Wed, 3 Nov 2010 13:14:48 +0000 (23:44 +1030)]
web: point to the configurator to generate config.h
Rusty Russell [Wed, 3 Nov 2010 13:07:33 +0000 (23:37 +1030)]
.gitignore: ignore configurator
Rusty Russell [Wed, 3 Nov 2010 12:48:10 +0000 (23:18 +1030)]
makefiles: add "scores" target.
Rusty Russell [Wed, 3 Nov 2010 12:48:10 +0000 (23:18 +1030)]
alloc: fix missing header in compilation of example.
Rusty Russell [Wed, 3 Nov 2010 12:55:57 +0000 (23:25 +1030)]
ccanlint: make tests non-compulsory, always print score.
No longer abort ccanlint because a test fails, but you will get a low
score if there are no tests.
(Note: total_score is filled in, even if it's overridden when the test is
run. This means that if the prereq fails, that total counts so there is
still a penalty!)
Rusty Russell [Wed, 3 Nov 2010 12:51:18 +0000 (23:21 +1030)]
ccanlint: with -vv, give details on attempts to compile examples.
The heuristics can be hinky, so tell people what we're doing.
Rusty Russell [Wed, 3 Nov 2010 00:21:07 +0000 (10:51 +1030)]
tools: new "configurator" tool.
A simple C program to generate config.h. It also saves the compiler and
flags for use by ccanlint (though they have to accept -c and -o).
Rusty Russell [Wed, 3 Nov 2010 00:20:03 +0000 (10:50 +1030)]
tdb2: fix gcc -O3 warnings on test/layout.c
Warnings about ignored returns, and uninitialized len after case statement.
Rusty Russell [Wed, 3 Nov 2010 00:17:39 +0000 (10:47 +1030)]
tdb: fix warnings with gcc -O3, and one test crash.
unsigned char * is allowed to alias, so we use that for byte reversing
rather than uint32_t. This is portable.
Remove warnings about ignoring pread/pwrite return values.
And initialize tdb before setjmp: with optimization, gcc validly saw this
as NULL and crashed.
Rusty Russell [Wed, 3 Nov 2010 00:16:42 +0000 (10:46 +1030)]
talloc: fix gcc -O3 aliasing warnings
A void * could point to anything, but a void ** can't point to any pointer.
So we use a void * and memcpy, which I believe is safe.
Rusty Russell [Wed, 3 Nov 2010 00:16:02 +0000 (10:46 +1030)]
opt: fix gcc -O3 warnings.