-
Notifications
You must be signed in to change notification settings - Fork 467
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[MINOR] Serialize benchmark #1880
Conversation
Just for reference at this point in time: Serialize Repetitions: 100 ConstMatrix ( Rows:10000, Cols:100, Spar:1.0, Unique: 4) threads: 16
Serialize, 3.721+- 0.206 ms, 2149908260+- 110873533 Byte/s, 2149869831+- 110871552 Byte/s
CustomDisk, 47.684+- 2.071 ms, 167774123+- 6593177 Byte/s, 167771124+- 6593059 Byte/s
StandardDisk, 16.028+- 1.260 ms, 499138069+- 34424829 Byte/s, 499168267+- 34426911 Byte/s
Compress Normal, 6.666+- 1.528 ms, 1200100947+- 256015622 Byte/s, 106348525+- 22687161 Byte/s
Compress Normal Serialize, 6.681+- 1.386 ms, 1197475373+- 202342062 Byte/s, 105588528+- 17841703 Byte/s
Compress Normal CustomDisk, 10.557+- 1.771 ms, 757809826+- 120157276 Byte/s, 66820601+- 10594982 Byte/s
Compress StandardIO, 10.725+- 1.853 ms, 745909565+- 127345784 Byte/s, 47687091+- 8055640 Byte/s
Update&Apply Scheme Fused, 3.400+- 0.856 ms, 2352963868+- 437720801 Byte/s, 134982266+- 25110690 Byte/s
Update&Apply Scheme Fused Serialize, 4.040+- 0.078 ms, 1980194678+- 34782513 Byte/s, 112725638+- 1980048 Byte/s
Update&Apply Scheme Fused Disk, 5.351+- 0.500 ms, 1495041723+- 118443086 Byte/s, 85107558+- 6742555 Byte/s
Update&Apply Standard IO, 7.662+- 2.114 ms, 1044126089+- 292755327 Byte/s, 33623017+- 9343659 Byte/s |
Updates Serialize Repetitions: 100 ConstMatrix ( Rows:10000, Cols:100, Spar:1.0, Unique: 4) threads: 16
Serialize, 4.108+- 0.229 ms, 1947284921+- 98408634 Byte/s, 1947250114+- 98406875 Byte/s
StandardDisk, 16.738+- 0.998 ms, 477962478+- 27594735 Byte/s, 477991394+- 27596404 Byte/s
Compress Normal, 6.439+- 0.989 ms, 1242413398+- 160468203 Byte/s, 114911724+- 14843580 Byte/s
Compress Normal Serialize, 6.185+- 0.671 ms, 1293438291+- 112369549 Byte/s, 119063421+- 10343828 Byte/s
Compress StandardIO, 10.349+- 1.463 ms, 773046144+- 104125613 Byte/s, 48442725+- 6462132 Byte/s
Update&Apply Scheme Fused, 3.145+- 0.132 ms, 2544027113+- 108157032 Byte/s, 243158579+- 10337669 Byte/s
Update&Apply Scheme Fused Serialize, 3.977+- 0.065 ms, 2011507633+- 32403630 Byte/s, 191374469+- 3082875 Byte/s
Update&Apply Standard IO, 6.085+- 1.305 ms, 1314730595+- 188225568 Byte/s, 81849890+- 11689032 Byte/s |
With better IO start, isolated dictionary writing, slicing of sub rows inside writing tasks, and single pass update and encoding. Serialize Repetitions: 1000 ConstMatrix ( Rows:10000, Cols:100, Spar:1.0, Unique: 4) threads: 16
Serialize, 3.555+- 0.115 ms, 2250210702+- 70104470 Byte/s, 2250170480+- 70103217 Byte/s
StandardDisk, 15.334+- 0.860 ms, 521730319+- 27725252 Byte/s, 521761883+- 27726929 Byte/s
Compress Normal, 6.664+- 2.232 ms, 1200494359+- 300677755 Byte/s, 106383388+- 26644955 Byte/s
Compress Normal Serialize, 5.912+- 0.698 ms, 1353258662+- 148101799 Byte/s, 119324868+- 13059017 Byte/s
Compress StandardIO, 8.020+- 0.658 ms, 997483498+- 73930038 Byte/s, 61658745+- 4483419 Byte/s
Update&Apply Scheme Fused, 1.965+- 0.049 ms, 4071051957+- 100537669 Byte/s, 233543671+- 5767535 Byte/s
Update&Apply Scheme Fused Serialize, 2.327+- 0.038 ms, 3438309852+- 56253829 Byte/s, 195731095+- 3202336 Byte/s
Update&Apply Standard IO, 3.823+- 0.148 ms, 2092482764+- 76584230 Byte/s, 57112163+- 3120851 Byte/s |
And now with single threaded execution fixes, we can report these values as well, futhermore the values now also correctly indicate how much is written to disk a sec. Single thread: 669MiB a sec compressed writing Serialize Repetitions: 1500 ConstMatrix ( Rows:10000, Cols:100, Spar:1.0, Unique: 4) threads: 1
Serialize, 4.085+- 0.066 ms, 1958553044+- 31007119 Byte/s, 1958518036+- 31006564 Byte/s
StandardDisk, 16.699+- 0.741 ms, 479074510+- 20665474 Byte/s, 491385354+- 21196518 Byte/s
Compress Normal, 18.321+- 0.214 ms, 436663603+- 5002905 Byte/s, 38695520+- 443339 Byte/s
Compress Normal Serialize, 18.481+- 0.173 ms, 432879634+- 3966299 Byte/s, 38169573+- 349732 Byte/s
Compress StandardIO, 23.159+- 0.234 ms, 345447203+- 3439765 Byte/s, 30803115+- 306720 Byte/s
Update&Apply Scheme Fused, 9.887+- 0.401 ms, 809155714+- 33166438 Byte/s, 46418763+- 1902656 Byte/s
Update&Apply Scheme Fused Serialize, 10.837+- 0.206 ms, 738199877+- 14382973 Byte/s, 42023167+- 818773 Byte/s
Update&Apply Standard IO, 11.953+- 0.521 ms, 669319430+- 29564362 Byte/s, 38764906+- 1712276 Byte/s Parallel: 2GiB a sec compressed writing. These results a a bit slower than the last reported values because my computer was hot: Serialize Repetitions: 1500 ConstMatrix ( Rows:10000, Cols:100, Spar:1.0, Unique: 4) threads: 16
Serialize, 5.189+- 0.568 ms, 1541824285+- 155223818 Byte/s, 1541796726+- 155221043 Byte/s
StandardDisk, 18.312+- 0.992 ms, 436891339+- 23294883 Byte/s, 441838821+- 23558681 Byte/s
Compress Normal, 6.357+- 0.612 ms, 1258542122+- 118458337 Byte/s, 111527367+- 10497341 Byte/s
Compress Normal Serialize, 6.710+- 0.585 ms, 1192213605+- 99070533 Byte/s, 105124567+- 8735638 Byte/s
Compress StandardIO, 10.455+- 1.115 ms, 765215442+- 71066049 Byte/s, 67206398+- 6157964 Byte/s
Update&Apply Scheme Fused, 2.169+- 0.121 ms, 3689126824+- 215759110 Byte/s, 211633807+- 12377433 Byte/s
Update&Apply Scheme Fused Serialize, 2.590+- 0.084 ms, 3089128610+- 102204861 Byte/s, 175853414+- 5818169 Byte/s
Update&Apply Standard IO, 3.918+- 0.159 ms, 2042115006+- 77696920 Byte/s, 109587347+- 4860114 Byte/s |
c61a31a
to
9b426a5
Compare
Optimize compression writing IDictionary Compressed save dictionaries separately. Optimize readers Slice inside task Optimization for single thread Spark Read Compressed Matrix Nan Values valid to compress. fix n unique in Nan case make the IOTest repeat if it fails
f241678
to
8f7eb0f
Compare
No description provided.