Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Append/Prepend some more optimizations #764

Merged
merged 7 commits into from
Jul 16, 2018
Merged

Append/Prepend some more optimizations #764

merged 7 commits into from
Jul 16, 2018

Conversation

quinmars
Copy link
Contributor

@quinmars quinmars commented Jul 5, 2018

In this PR I tried to follow the mprovements made to the ToObersable class. I also splitted the single AppendPrependMultiple class into two implementation. It made the code easier to understand. I also added a specialized version for the prepend loop. That way we save the allocation creation of the prepend array. Now all 3 or 4 classes of the AppendPrepend familiy have a similar structure and method names.

There are a some small things missing:

  • I need to merge the latest changes to AppendPrependSingle
  • I want to delay the append array creation until it is realy needed, i.e., after the source completion. So the array does not need to be created when the subscription is disposed before and it is fresh in memory at the moment when it is needed.

Measurements before:

BenchmarkDotNet=v0.10.14, OS=Windows 10.0.17134
Intel Core i5-5200U CPU 2.20GHz (Broadwell), 1 CPU, 4 logical and 2 physical cores
Frequency=2143476 Hz, Resolution=466.5319 ns, Timer=TSC
  [Host]     : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
  DefaultJob : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
Method N Mean Error StdDev Scaled ScaledSD Gen 0 Gen 1 Gen 2 Allocated
StartWithArray 1 3.528 us 0.0700 us 0.0934 us 1.00 0.00 0.9193 - - 1.42 KB
StartWithList 1 3.983 us 0.0837 us 0.2442 us 1.13 0.07 1.0223 - - 1.57 KB
StartWithLinkedList 1 3.717 us 0.0907 us 0.0891 us 1.05 0.04 1.0605 - - 1.64 KB
AppendPrepend 1 3.238 us 0.0636 us 0.0707 us 0.92 0.03 1.3046 - - 2.01 KB
StartWithArray 10 7.786 us 0.1843 us 0.2975 us 1.00 0.00 1.7853 - - 2.75 KB
StartWithList 10 8.027 us 0.0864 us 0.0722 us 1.03 0.04 2.1210 - - 3.27 KB
StartWithLinkedList 10 8.349 us 0.0853 us 0.0756 us 1.07 0.04 2.5024 - - 3.85 KB
AppendPrepend 10 16.528 us 0.2417 us 0.2261 us 2.13 0.08 4.6387 - - 7.16 KB
StartWithArray 100 46.883 us 0.5206 us 0.4870 us 1.00 0.00 10.4370 - - 16.12 KB
StartWithList 100 52.340 us 1.0122 us 1.2050 us 1.12 0.03 12.3901 - - 19.12 KB
StartWithLinkedList 100 52.390 us 0.5356 us 0.4748 us 1.12 0.01 17.1509 - - 26.37 KB
AppendPrepend 100 151.495 us 1.3802 us 1.2911 us 3.23 0.04 34.6680 - - 53.46 KB
StartWithArray 1000 444.947 us 8.8062 us 12.0540 us 1.00 0.00 97.1680 - - 149.73 KB
StartWithList 1000 590.252 us 11.0556 us 10.3414 us 1.33 0.04 112.3047 - - 173.92 KB
StartWithLinkedList 1000 505.393 us 6.3248 us 5.6068 us 1.14 0.03 151.3672 9.7656 - 251.38 KB
AppendPrepend 1000 1,591.257 us 16.0398 us 15.0037 us 3.58 0.10 332.0313 - - 512.74 KB
StartWithArray 10000 4,378.554 us 46.1767 us 38.5597 us 1.00 0.00 960.9375 - - 1485.71 KB
StartWithList 10000 15,953.991 us 182.8502 us 162.0919 us 3.64 0.05 1093.7500 31.2500 31.2500 1820.76 KB
StartWithLinkedList 10000 6,142.251 us 139.7508 us 130.7230 us 1.40 0.03 843.7500 257.8125 - 2501.78 KB
AppendPrepend 10000 17,522.780 us 238.5906 us 223.1778 us 4.00 0.06 968.7500 406.2500 - 5098.99 KB

After the switch to the light recursive scheduling:

BenchmarkDotNet=v0.10.14, OS=Windows 10.0.17134
Intel Core i5-5200U CPU 2.20GHz (Broadwell), 1 CPU, 4 logical and 2 physical cores
Frequency=2143476 Hz, Resolution=466.5319 ns, Timer=TSC
  [Host]     : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
  DefaultJob : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
Method N Mean Error StdDev Scaled ScaledSD Gen 0 Gen 1 Gen 2 Allocated
StartWithArray 1 4.283 us 0.0409 us 0.0383 us 1.00 0.00 0.9155 - - 1.42 KB
StartWithList 1 4.331 us 0.0402 us 0.0376 us 1.01 0.01 1.0223 - - 1.57 KB
StartWithLinkedList 1 4.724 us 0.0468 us 0.0415 us 1.10 0.01 1.0605 - - 1.64 KB
AppendPrepend 1 3.650 us 0.0202 us 0.0189 us 0.85 0.01 0.9575 - - 1.47 KB
StartWithArray 10 8.948 us 0.1027 us 0.0961 us 1.00 0.00 1.7853 - - 2.75 KB
StartWithList 10 9.637 us 0.0825 us 0.0732 us 1.08 0.01 2.1210 - - 3.27 KB
StartWithLinkedList 10 9.538 us 0.0900 us 0.0797 us 1.07 0.01 2.5024 - - 3.85 KB
AppendPrepend 10 10.712 us 0.0547 us 0.0512 us 1.20 0.01 3.2806 - - 5.05 KB
StartWithArray 100 56.349 us 0.4865 us 0.4063 us 1.00 0.00 10.4370 - - 16.12 KB
StartWithList 100 62.029 us 0.4953 us 0.4633 us 1.10 0.01 12.3291 - - 19.12 KB
StartWithLinkedList 100 62.506 us 0.5599 us 0.4371 us 1.11 0.01 17.0898 - - 26.37 KB
AppendPrepend 100 83.353 us 0.8382 us 0.7431 us 1.48 0.02 23.8037 - - 36.66 KB
StartWithArray 1000 530.851 us 5.2621 us 4.3941 us 1.00 0.00 96.6797 - - 149.73 KB
StartWithList 1000 706.448 us 4.5027 us 4.2118 us 1.33 0.01 112.3047 - - 173.92 KB
StartWithLinkedList 1000 608.962 us 8.6027 us 7.6260 us 1.15 0.02 162.1094 0.9766 - 251.5 KB
AppendPrepend 1000 758.401 us 14.8476 us 27.5209 us 1.43 0.05 229.4922 - - 353.13 KB
StartWithArray 10000 4,430.645 us 29.2578 us 27.3677 us 1.00 0.00 960.9375 - - 1485.73 KB
StartWithList 10000 16,078.175 us 81.1868 us 67.7946 us 3.63 0.03 1093.7500 31.2500 31.2500 1820.76 KB
StartWithLinkedList 10000 6,162.668 us 28.1868 us 24.9869 us 1.39 0.01 843.7500 257.8125 - 2501.78 KB
AppendPrepend 10000 8,515.175 us 73.1290 us 68.4049 us 1.92 0.02 718.7500 312.5000 - 3517.31 KB

With the specialized prepend loop:

BenchmarkDotNet=v0.10.14, OS=Windows 10.0.17134
Intel Core i5-5200U CPU 2.20GHz (Broadwell), 1 CPU, 4 logical and 2 physical cores
Frequency=2143476 Hz, Resolution=466.5319 ns, Timer=TSC
  [Host]     : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
  DefaultJob : .NET Framework 4.7.1 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3110.0
Method N Mean Error StdDev Scaled ScaledSD Gen 0 Gen 1 Gen 2 Allocated
StartWithArray 1 4.232 us 0.0847 us 0.0941 us 1.00 0.00 0.9155 - - 1.42 KB
StartWithList 1 4.189 us 0.0121 us 0.0095 us 0.99 0.02 1.0223 - - 1.57 KB
StartWithLinkedList 1 4.028 us 0.0217 us 0.0181 us 0.95 0.02 1.0605 - - 1.64 KB
AppendPrepend 1 3.466 us 0.0219 us 0.0205 us 0.82 0.02 0.9270 - - 1.43 KB
StartWithArray 10 8.760 us 0.0131 us 0.0109 us 1.00 0.00 1.7853 - - 2.75 KB
StartWithList 10 9.340 us 0.0129 us 0.0121 us 1.07 0.00 2.1210 - - 3.27 KB
StartWithLinkedList 10 9.678 us 0.0261 us 0.0231 us 1.10 0.00 2.5024 - - 3.85 KB
AppendPrepend 10 9.179 us 0.0530 us 0.0496 us 1.05 0.01 2.9449 - - 4.54 KB
StartWithArray 100 54.213 us 0.0596 us 0.0529 us 1.00 0.00 10.4370 - - 16.12 KB
StartWithList 100 60.062 us 0.1947 us 0.1821 us 1.11 0.00 12.3901 - - 19.12 KB
StartWithLinkedList 100 60.878 us 0.3439 us 0.3217 us 1.12 0.01 17.0898 - - 26.37 KB
AppendPrepend 100 64.398 us 0.3147 us 0.2789 us 1.19 0.01 20.5078 - - 31.67 KB
StartWithArray 1000 516.180 us 2.5220 us 2.2357 us 1.00 0.00 96.6797 - - 149.73 KB
StartWithList 1000 684.949 us 2.8728 us 2.5467 us 1.33 0.01 112.3047 - - 173.92 KB
StartWithLinkedList 1000 602.670 us 11.3540 us 10.6206 us 1.17 0.02 162.1094 0.9766 - 251.5 KB
AppendPrepend 1000 662.919 us 0.6880 us 0.5746 us 1.28 0.01 196.2891 - - 302.36 KB
StartWithArray 10000 5,141.553 us 36.8668 us 34.4852 us 1.00 0.00 960.9375 - - 1485.73 KB
StartWithList 10000 18,687.550 us 28.6423 us 23.9176 us 3.63 0.02 1093.7500 31.2500 31.2500 1820.76 KB
StartWithLinkedList 10000 7,104.952 us 56.7801 us 50.3341 us 1.38 0.01 843.7500 257.8125 - 2501.78 KB
AppendPrepend 10000 7,999.898 us 32.5722 us 30.4680 us 1.56 0.01 515.6250 234.3750 - 3009.51 KB

@danielcweber
Copy link
Collaborator

Watch out, you're working on a stale master, there have been merged changes by David for AppendPrepend. I hope you can rebase easily.

@quinmars quinmars force-pushed the appendprepend-moreoptimization branch from fc5326a to 1c43c5c Compare July 6, 2018 08:46
@quinmars
Copy link
Contributor Author

quinmars commented Jul 6, 2018

I have now rebased the branch on the master's tip. It was semi successfull not every single commit will lead to a compileable version but the tip compiles again. Maybe you could squash it when you commit it to the master.

It is not complete finished, because I want to fix some comments that are outdated. Once that is done I will remove the WIP tag.

@quinmars quinmars changed the title [WIP] Append/Prepend some more optimizations Append/Prepend some more optimizations Jul 6, 2018
@quinmars quinmars force-pushed the appendprepend-moreoptimization branch from ccd1073 to f91f449 Compare July 8, 2018 17:27
@danielcweber danielcweber merged commit 03691a8 into dotnet:master Jul 16, 2018
@akarnokd akarnokd mentioned this pull request Jul 24, 2018
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@quinmars @danielcweber