Drezil/hgraph
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

made parallel

Browse Source 
some performance-data in main.hs.

Changes can be viewed in git.

TODO:
- Find out why there is overhed of >50%
This commit is contained in:
Stefan Dresselhaus
2013-12-09 01:51:32 +01:00
parent 9abbf0f508
commit 1ec8de8ee2
13 changed files with 1498 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

240
src/Main.hs
View File

@@ -42,7 +42,7 @@ import qualified Data.Stream as S
import qualified Data.Text as T
import Data.Text.Encoding
import qualified Data.Vector.Unboxed as V
import Debug.Trace
--import Debug.Trace
import System.Environment
import System.Exit (exitFailure, exitSuccess)
import Test.QuickCheck.All (quickCheckAll)
@@ -120,20 +120,21 @@ emptyLine a
doCalculation adj attr =
let
dens = 0.75
omega = (A.fromListUnboxed (ix1 3) [0.5,0.5,0.5])
omega = (A.fromListUnboxed (ix1 6) [0,5,3,300,5,10])
delta = 2
(adj_, graph_) = preprocess adj attr {--0.8--} omega delta
in
B.concat $
[
outputArray $ trace ("After: "++ show (sumAllS adj_)++"\n") adj_,
--outputArray $ --trace ("After: "++ show (sumAllS adj_)++"\n")
-- adj_,
outputGraph $ L.sort $ doAll graph_ adj_ attr dens omega delta
-- outputGraph $ L.sort $ (step graph_ adj attr dens omega delta)
-- ++ (step (step graph_ adj attr dens omega delta) adj attr dens omega delta)
]
where
-- don't print out seeds
doAll [] _ _ _ _ _ = []
doAll gs a b c d e = doAll' (step gs a b c d e) a b c d e
-- but everything in the following recursive calls
doAll' [] _ _ _ _ _ = []
doAll' gs a b c d e = gs ++ doAll' (step gs a b c d e) a b c d e
@@ -172,7 +173,7 @@ checkError a
-- | convinience debug-function. Needs to be
-- changed to return () to disable Debug.
debug a = putStrLn a
debug a = return () --putStrLn a
-- | The main-function to bootstrap our application
@@ -226,5 +227,230 @@ main = do
----- CALCULATE & OUTPUT
debug $ "Before: " ++ show (sumAllS graph)
--debug $ "Before: " ++ show (sumAllS graph)
B.putStr $ doCalculation graph attr
{---TIMINGS
SINGLE CORE
===========
./hgraph +RTS -s >result
197,751,229,488 bytes allocated in the heap
290,034,880 bytes copied during GC
11,061,600 bytes maximum residency (10 sample(s))
1,513,488 bytes maximum slop
33 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 381500 colls, 0 par 2.24s 2.20s 0.0000s 0.0005s
Gen 1 10 colls, 0 par 0.05s 0.05s 0.0054s 0.0154s
TASKS: 3 (1 bound, 2 peak workers (2 total), using -N1)
SPARKS: 6266 (0 converted, 0 overflowed, 0 dud, 15 GC'd, 6251 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 74.11s ( 74.11s elapsed)
GC time 2.30s ( 2.25s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 76.42s ( 76.36s elapsed)
Alloc rate 2,668,183,845 bytes per MUT second
Productivity 97.0% of total user, 97.1% of total elapsed
gc_alloc_block_sync: 0
whitehole_spin: 0
gen[0].sync: 0
gen[1].sync: 0
4 CORES:
========
./hgraph +RTS -s -N4 >result
197,754,645,560 bytes allocated in the heap
293,083,624 bytes copied during GC
11,061,264 bytes maximum residency (10 sample(s))
1,555,576 bytes maximum slop
34 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 380952 colls, 380952 par 15.25s 3.92s 0.0000s 0.0255s
Gen 1 10 colls, 9 par 0.22s 0.06s 0.0056s 0.0181s
Parallel GC work balance: 1.68% (serial 0%, perfect 100%)
TASKS: 6 (1 bound, 5 peak workers (5 total), using -N4)
SPARKS: 6266 (6228 converted, 0 overflowed, 0 dud, 30 GC'd, 8 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 105.25s ( 86.11s elapsed)
GC time 15.47s ( 3.98s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 120.73s ( 90.09s elapsed)
Alloc rate 1,878,861,647 bytes per MUT second
Productivity 87.2% of total user, 116.8% of total elapsed
gc_alloc_block_sync: 661438
whitehole_spin: 0
gen[0].sync: 655
gen[1].sync: 1347
parallel preprocessing (Adj, Seeds)
==================================
./hgraph +RTS -s -N4 >result
Building hgraph-0.0.1...
Preprocessing executable 'hgraph' for hgraph-0.0.1...
[4 of 5] Compiling DCB.DCB ( src/DCB/DCB.hs, dist/build/hgraph/hgraph-tmp/DCB/DCB.o )
Linking dist/build/hgraph/hgraph ...
197,755,802,848 bytes allocated in the heap
289,986,840 bytes copied during GC
11,071,880 bytes maximum residency (10 sample(s))
1,566,376 bytes maximum slop
34 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 380919 colls, 380919 par 15.73s 3.93s 0.0000s 0.0112s
Gen 1 10 colls, 9 par 0.28s 0.07s 0.0073s 0.0335s
Parallel GC work balance: 1.69% (serial 0%, perfect 100%)
TASKS: 6 (1 bound, 5 peak workers (5 total), using -N4)
SPARKS: 7895 (7825 converted, 0 overflowed, 0 dud, 50 GC'd, 20 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 98.47s ( 81.37s elapsed)
GC time 16.01s ( 4.00s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 114.49s ( 85.37s elapsed)
Alloc rate 2,008,240,220 bytes per MUT second
Productivity 86.0% of total user, 115.3% of total elapsed
gc_alloc_block_sync: 757575
whitehole_spin: 0
gen[0].sync: 592
gen[1].sync: 510
parallel processing (primitive, too many sparks fizzled) - Speedup: 76.36/51.51 = 1.48
========================================================
./hgraph +RTS -s -N4 >result
Building hgraph-0.0.1...
Preprocessing executable 'hgraph' for hgraph-0.0.1...
[4 of 5] Compiling DCB.DCB ( src/DCB/DCB.hs, dist/build/hgraph/hgraph-tmp/DCB/DCB.o )
[5 of 5] Compiling Main ( src/Main.hs, dist/build/hgraph/hgraph-tmp/Main.o )
Linking dist/build/hgraph/hgraph ...
205,324,862,344 bytes allocated in the heap
224,224,264 bytes copied during GC
11,157,008 bytes maximum residency (9 sample(s))
1,559,568 bytes maximum slop
35 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 123063 colls, 123063 par 6.77s 1.67s 0.0000s 0.0074s
Gen 1 9 colls, 8 par 0.21s 0.06s 0.0061s 0.0190s
Parallel GC work balance: 8.15% (serial 0%, perfect 100%)
TASKS: 6 (1 bound, 5 peak workers (5 total), using -N4)
SPARKS: 1714681 (861196 converted, 0 overflowed, 0 dud, 78 GC'd, 853407 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 145.46s ( 49.78s elapsed)
GC time 6.99s ( 1.73s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 152.45s ( 51.51s elapsed)
Alloc rate 1,411,565,587 bytes per MUT second
Productivity 95.4% of total user, 282.4% of total elapsed
gc_alloc_block_sync: 378641
whitehole_spin: 0
gen[0].sync: 572
gen[1].sync: 609
parallel processing (monad-par, repa-stuff seqential) - Speedup: 76.36/34,05 = 2.243
=====================================================
./hgraph +RTS -N4 -s > result.txt
204,368,634,080 bytes allocated in the heap
306,058,720 bytes copied during GC
11,108,872 bytes maximum residency (10 sample(s))
1,597,088 bytes maximum slop
35 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 108838 colls, 108838 par 9.21s 2.29s 0.0000s 0.0020s
Gen 1 10 colls, 9 par 0.32s 0.08s 0.0083s 0.0245s
Parallel GC work balance: 29.41% (serial 0%, perfect 100%)
TASKS: 6 (1 bound, 5 peak workers (5 total), using -N4)
SPARKS: 15737 (14412 converted, 0 overflowed, 0 dud, 1251 GC'd, 74 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 124.37s ( 31.67s elapsed)
GC time 9.53s ( 2.37s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 133.91s ( 34.05s elapsed)
Alloc rate 1,643,242,747 bytes per MUT second
Productivity 92.9% of total user, 365.3% of total elapsed
gc_alloc_block_sync: 531144
whitehole_spin: 0
gen[0].sync: 758
gen[1].sync: 17
ADDITIONAL OVERHEAD (running on 1 Core with parallel stuff): Slowdown: 76.36/123.39 = 0.62885
============================================================
./hgraph +RTS -N1 -s > result.txt
204,364,490,096 bytes allocated in the heap
291,824,120 bytes copied during GC
10,081,664 bytes maximum residency (11 sample(s))
1,545,536 bytes maximum slop
30 MB total memory in use (0 MB lost due to fragmentation)
Tot time (elapsed) Avg pause Max pause
Gen 0 393965 colls, 0 par 6.77s 6.72s 0.0000s 0.0017s
Gen 1 11 colls, 0 par 0.09s 0.09s 0.0079s 0.0217s
TASKS: 3 (1 bound, 2 peak workers (2 total), using -N1)
SPARKS: 15737 (0 converted, 0 overflowed, 0 dud, 166 GC'd, 15571 fizzled)
INIT time 0.00s ( 0.00s elapsed)
MUT time 116.53s (116.52s elapsed)
GC time 6.85s ( 6.80s elapsed)
EXIT time 0.00s ( 0.00s elapsed)
Total time 123.39s (123.32s elapsed)
Alloc rate 1,753,707,727 bytes per MUT second
Productivity 94.4% of total user, 94.5% of total elapsed
gc_alloc_block_sync: 0
whitehole_spin: 0
gen[0].sync: 0
gen[1].sync: 0
REAL SPEEDUP AGAINST OVERHEAD-VARIANT: 129.39/34.05 = 3.8
=========================================================
-}