Spannified internals of BigInteger #35565

sakno · 2020-04-28T10:24:57Z

Proposed refactoring according with issue: #22609. The implementation plan consists of the following steps:

Replace unmanaged pointers with managed ones as well as remove unsafe code and pinning
Beautify stack allocation
Use spans wherever possible, especially for memory slicing
Simplify (or probably remove at all) BitsBuffer value type
Spannify FastReducer value type
Attempt to replace some array allocations with span slicing
Square, Multiply and bitwise operations use heap-based allocation of arrays if their length is greater than or equal to stack allocation threshold. Maybe replace it with array pooling using shared ArrayPool<T>?
~~BMI intrinsics~~ (moved to separated issue BigInteger performance improvements #41495)

Spannified versions of internal and private static methods look pretty nice. However, I'm not sure about performance of passing span to the method. If RyuJIT uses scalar replacement then it's good news. Otherwise, maybe pass length and managed pointer to the first element as separate arguments to ensure that they passed through registers. This version was implemented in the first commit. I need advice here as well as preliminary code review because further work fully based on signatures of spannified methods.

ghost · 2020-04-28T10:25:04Z

Tagging subscribers to this area: @tannergooding
Notify danmosemsft if you want to be subscribed.

gfoidl

You can get rid of some Unsafe.Add usage by letting the JIT do the work and have the safety back.
(Didn't comment on all places, please do one pass of clean-up.)

Plus some nits.

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigInteger.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.AddSub.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.BitsBuffer.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.DivRem.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.FastReducer.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.AddSub.cs

sakno · 2020-04-29T21:21:02Z

@gfoidl , could you please take a look at the next iteration (commit a1e5006). I tried to reduce memory allocation for bitwise operations. But I'm not satisfied with necessity of delegate type that is introduced to reduce code duplication. Moreover, this approach leads to delegate instance allocation on every call. AFAIK, the current version of C# still doesn't support method pointers.

ToUInt32Array is renamed to CopyTo and now working with the memory allocated by the immediate caller.

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigInteger.cs

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.DivRem.cs

danmoseley · 2021-08-26T14:07:08Z

@GrabYourPitchforks sounds good

danmoseley · 2021-08-26T19:55:43Z

@sakno I haven't used it myself, but I believe you can run this on the before and after results to get a better comparison:
https://github.com/dotnet/performance/blob/main/src/tools/ResultsComparer/README.md
Hopefully it looks at allocations as well.

stephentoub · 2021-08-26T20:17:43Z

I believe you can run this on the before and after results to get a better comparison

I just add this to my dotnet run benchmarkdotnet cmd line:

--corerun d:\coreclrtest\main\corerun.exe d:\coreclrtest\pr\corerun.exe

where those main and pr folders contain a copy of the contents of D:\repos\runtime\artifacts\bin\testhost\net6.0-windows-Release-x64\shared\Microsoft.NETCore.App\7.0.0 before and after the change.

danmoseley · 2021-08-30T20:10:15Z

Just checking in on this one since it's so old, seems like the next action is @GrabYourPitchforks offered to have another look through. And @sakno if you are interested you could try the trick above to get a single table.

GrabYourPitchforks · 2021-08-30T21:14:00Z

I had something come up which will make me busy for the next few days, but I can continue driving this once my schedule frees up. Should we un-milestone from 6.0 then?

danmoseley · 2021-08-30T21:51:01Z

Yes, they won't take this change into 6.0.

sakno · 2021-09-06T13:33:02Z

Method	Branch	numberString	arguments	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen 0	Allocated
Ctor_ByteArray	main	-2147483648	?	13.954 ns	0.1019 ns	0.0851 ns	13.941 ns	13.819 ns	14.118 ns	1.00	0.00	-	-
Ctor_ByteArray	PR	-2147483648	?	14.761 ns	0.0901 ns	0.0799 ns	14.760 ns	14.580 ns	14.865 ns	1.06	0.01	-	-

ToByteArray	main	-2147483648	?	20.727 ns	0.7729 ns	0.8900 ns	20.618 ns	19.637 ns	22.860 ns	1.00	0.00	0.0101	32 B
ToByteArray	PR	-2147483648	?	20.204 ns	0.3181 ns	0.2975 ns	20.061 ns	19.861 ns	20.859 ns	0.99	0.04	0.0102	32 B

Parse	main	-2147483648	?	194.382 ns	4.7403 ns	5.4589 ns	192.446 ns	189.305 ns	209.048 ns	1.00	0.00	0.0427	136 B
Parse	PR	-2147483648	?	185.266 ns	0.9011 ns	0.7525 ns	184.929 ns	184.252 ns	186.515 ns	0.94	0.03	0.0426	136 B

ToStringX	main	-2147483648	?	80.635 ns	3.3895 ns	3.7674 ns	78.761 ns	75.194 ns	88.298 ns	1.00	0.00	0.0126	40 B
ToStringX	PR	-2147483648	?	72.292 ns	0.6056 ns	0.5369 ns	72.084 ns	71.747 ns	73.678 ns	0.89	0.05	0.0126	40 B

ToStringD	main	-2147483648	?	76.453 ns	0.8156 ns	0.7230 ns	76.179 ns	75.101 ns	77.658 ns	1.00	0.00	0.0481	152 B
ToStringD	PR	-2147483648	?	71.113 ns	0.2899 ns	0.2712 ns	71.067 ns	70.690 ns	71.631 ns	0.93	0.01	0.0484	152 B

Add	main	?	1024,1024 bits	56.347 ns	1.1633 ns	1.1946 ns	55.965 ns	55.397 ns	59.876 ns	1.00	0.00	0.0510	160 B
Add	PR	?	1024,1024 bits	56.439 ns	0.2406 ns	0.2009 ns	56.427 ns	56.043 ns	56.775 ns	1.00	0.02	0.0509	160 B

Subtract	main	?	1024,1024 bits	57.248 ns	0.3050 ns	0.2853 ns	57.254 ns	56.701 ns	57.831 ns	1.00	0.00	0.0482	152 B
Subtract	PR	?	1024,1024 bits	57.372 ns	0.5303 ns	0.4960 ns	57.226 ns	56.800 ns	58.195 ns	1.00	0.01	0.0484	152 B

Multiply	main	?	1024,1024 bits	1,018.982 ns	5.7862 ns	5.4124 ns	1,017.372 ns	1,010.925 ns	1,029.580 ns	1.00	0.00	0.0857	280 B
Multiply	PR	?	1024,1024 bits	1,070.928 ns	4.9263 ns	4.6081 ns	1,070.226 ns	1,064.082 ns	1,078.841 ns	1.05	0.01	0.0859	280 B

GreatestCommonDivisor	main	?	1024,1024 bits	10,890.242 ns	76.4818 ns	67.7992 ns	10,885.738 ns	10,799.565 ns	10,999.333 ns	1.00	0.00	0.0870	304 B
GreatestCommonDivisor	PR	?	1024,1024 bits	11,022.761 ns	86.6214 ns	76.7876 ns	10,992.526 ns	10,948.845 ns	11,209.706 ns	1.01	0.01	-	-

ModPow	main	?	1024,1024,64 bits	176,924.900 ns	1,493.7503 ns	1,247.3492 ns	177,199.311 ns	174,384.102 ns	178,660.212 ns	1.00	0.00	-	305 B
ModPow	PR	?	1024,1024,64 bits	167,633.288 ns	803.3346 ns	627.1908 ns	167,786.488 ns	166,646.389 ns	168,750.669 ns	0.95	0.01	-	33 B

Divide	main	?	1024,512 bits	642.343 ns	4.2045 ns	3.7272 ns	642.788 ns	636.328 ns	648.605 ns	1.00	0.00	0.0286	96 B
Divide	PR	?	1024,512 bits	662.570 ns	7.4316 ns	6.2057 ns	661.077 ns	656.037 ns	678.460 ns	1.03	0.01	0.0291	96 B

Remainder	main	?	1024,512 bits	701.984 ns	3.9658 ns	3.7096 ns	701.717 ns	696.460 ns	708.885 ns	1.00	0.00	0.0748	240 B
Remainder	PR	?	1024,512 bits	657.863 ns	3.2204 ns	3.0123 ns	657.702 ns	652.353 ns	663.535 ns	0.94	0.00	0.0264	88 B

Ctor_ByteArray	main	123	?	9.109 ns	0.1340 ns	0.1253 ns	9.082 ns	8.944 ns	9.307 ns	1.00	0.00	-	-
Ctor_ByteArray	PR	123	?	8.949 ns	0.0736 ns	0.0688 ns	8.951 ns	8.853 ns	9.075 ns	0.98	0.02	-	-

ToByteArray	main	123	?	16.474 ns	0.1133 ns	0.1060 ns	16.497 ns	16.320 ns	16.688 ns	1.00	0.00	0.0101	32 B
ToByteArray	PR	123	?	16.809 ns	0.2037 ns	0.1905 ns	16.770 ns	16.528 ns	17.156 ns	1.02	0.01	0.0102	32 B

Parse	main	123	?	125.084 ns	1.7400 ns	1.5425 ns	125.059 ns	123.403 ns	129.047 ns	1.00	0.00	0.0328	104 B
Parse	PR	123	?	125.753 ns	1.0728 ns	0.9510 ns	125.476 ns	124.542 ns	127.676 ns	1.01	0.02	0.0328	104 B

ToStringX	main	123	?	60.686 ns	0.4545 ns	0.4251 ns	60.796 ns	59.957 ns	61.458 ns	1.00	0.00	0.0100	32 B
ToStringX	PR	123	?	61.624 ns	0.3370 ns	0.3153 ns	61.519 ns	61.216 ns	62.294 ns	1.02	0.01	0.0100	32 B

ToStringD	main	123	?	43.681 ns	0.1716 ns	0.1521 ns	43.717 ns	43.313 ns	43.854 ns	1.00	0.00	0.0101	32 B
ToStringD	PR	123	?	45.596 ns	0.3587 ns	0.3180 ns	45.640 ns	45.087 ns	46.047 ns	1.04	0.01	0.0101	32 B

Ctor_ByteArray	main	123456789012(...)901234567890 [200]	?	144.698 ns	1.3993 ns	1.3089 ns	144.616 ns	142.887 ns	147.656 ns	1.00	0.00	0.0352	112 B
Ctor_ByteArray	PR	123456789012(...)901234567890 [200]	?	143.912 ns	0.6425 ns	0.6010 ns	144.048 ns	143.003 ns	144.906 ns	0.99	0.01	0.0355	112 B

ToByteArray	main	123456789012(...)901234567890 [200]	?	69.283 ns	0.5334 ns	0.4454 ns	69.176 ns	68.703 ns	70.031 ns	1.00	0.00	0.0356	112 B
ToByteArray	PR	123456789012(...)901234567890 [200]	?	70.821 ns	0.3920 ns	0.3667 ns	70.853 ns	70.167 ns	71.335 ns	1.02	0.01	0.0357	112 B

Parse	main	123456789012(...)901234567890 [200]	?	1,743.566 ns	7.1283 ns	5.9525 ns	1,742.759 ns	1,734.400 ns	1,755.052 ns	1.00	0.00	0.3071	984 B
Parse	PR	123456789012(...)901234567890 [200]	?	1,779.592 ns	8.3089 ns	7.3656 ns	1,779.426 ns	1,770.266 ns	1,794.882 ns	1.02	0.00	0.3131	984 B

ToStringX	main	123456789012(...)901234567890 [200]	?	464.817 ns	2.9414 ns	2.6075 ns	464.115 ns	461.699 ns	469.537 ns	1.00	0.00	0.1137	360 B
ToStringX	PR	123456789012(...)901234567890 [200]	?	456.938 ns	2.6321 ns	2.4621 ns	456.267 ns	453.956 ns	461.539 ns	0.98	0.01	0.1135	360 B

ToStringD	main	123456789012(...)901234567890 [200]	?	1,083.860 ns	6.4113 ns	5.6835 ns	1,083.582 ns	1,072.836 ns	1,094.347 ns	1.00	0.00	0.3151	992 B
ToStringD	PR	123456789012(...)901234567890 [200]	?	1,041.487 ns	9.7771 ns	8.6671 ns	1,038.352 ns	1,032.884 ns	1,058.836 ns	0.96	0.01	0.3155	992 B

Add	main	?	16,16 bits	11.229 ns	0.0470 ns	0.0393 ns	11.229 ns	11.162 ns	11.305 ns	1.00	0.00	-	-
Add	PR	?	16,16 bits	7.047 ns	0.0570 ns	0.0506 ns	7.045 ns	6.973 ns	7.149 ns	0.63	0.00	-	-

Subtract	main	?	16,16 bits	11.216 ns	0.0828 ns	0.0774 ns	11.204 ns	11.092 ns	11.341 ns	1.00	0.00	-	-
Subtract	PR	?	16,16 bits	6.494 ns	0.0715 ns	0.0634 ns	6.489 ns	6.424 ns	6.664 ns	0.58	0.01	-	-

Multiply	main	?	16,16 bits	10.973 ns	0.1227 ns	0.1147 ns	10.972 ns	10.802 ns	11.209 ns	1.00	0.00	-	-
Multiply	PR	?	16,16 bits	7.742 ns	0.1315 ns	0.1230 ns	7.731 ns	7.602 ns	8.052 ns	0.71	0.01	-	-

GreatestCommonDivisor	main	?	16,16 bits	76.718 ns	0.4069 ns	0.3607 ns	76.750 ns	76.264 ns	77.468 ns	1.00	0.00	-	-
GreatestCommonDivisor	PR	?	16,16 bits	76.764 ns	0.4106 ns	0.3640 ns	76.640 ns	76.164 ns	77.423 ns	1.00	0.01	-	-

ModPow	main	?	16,16,16 bits	179.341 ns	0.7725 ns	0.7226 ns	179.333 ns	177.896 ns	180.873 ns	1.00	0.00	-	-
ModPow	PR	?	16,16,16 bits	200.235 ns	1.0669 ns	0.9458 ns	200.308 ns	198.739 ns	202.131 ns	1.12	0.01	-	-

Divide	main	?	16,8 bits	9.324 ns	0.0411 ns	0.0343 ns	9.323 ns	9.277 ns	9.385 ns	1.00	0.00	-	-
Divide	PR	?	16,8 bits	10.364 ns	0.0747 ns	0.0662 ns	10.370 ns	10.247 ns	10.496 ns	1.11	0.01	-	-

Remainder	main	?	16,8 bits	9.433 ns	0.0795 ns	0.0705 ns	9.420 ns	9.358 ns	9.602 ns	1.00	0.00	-	-
Remainder	PR	?	16,8 bits	10.091 ns	0.0647 ns	0.0605 ns	10.081 ns	9.978 ns	10.178 ns	1.07	0.01	-	-

ModPow	main	?	16384,16384,64 bits	2,982,284.055 ns	20,304.5000 ns	18,992.8419 ns	2,981,140.681 ns	2,951,121.619 ns	3,024,813.569 ns	1.00	0.00	-	2,236 B
ModPow	PR	?	16384,16384,64 bits	3,149,539.852 ns	52,327.1029 ns	48,946.8045 ns	3,144,966.418 ns	3,064,405.190 ns	3,231,942.696 ns	1.06	0.02	-	47 B

Divide	main	?	65536,32768 bits	5,211,368.132 ns	20,965.3161 ns	18,585.2067 ns	5,207,763.708 ns	5,186,675.948 ns	5,253,423.531 ns	1.00	0.00	-	12,363 B
Divide	PR	?	65536,32768 bits	5,469,820.721 ns	23,984.5013 ns	22,435.1174 ns	5,474,290.333 ns	5,426,964.667 ns	5,501,347.646 ns	1.05	0.01	-	4,160 B

Remainder	main	?	65536,32768 bits	5,274,423.949 ns	40,466.6923 ns	35,872.6687 ns	5,260,983.656 ns	5,228,020.708 ns	5,351,705.146 ns	1.00	0.00	-	12,355 B
Remainder	PR	?	65536,32768 bits	5,327,585.356 ns	21,686.8215 ns	20,285.8663 ns	5,329,391.406 ns	5,293,440.948 ns	5,360,807.240 ns	1.01	0.01	-	4,146 B

Add	main	?	65536,65536 bits	2,261.169 ns	17.0800 ns	15.1409 ns	2,261.419 ns	2,236.812 ns	2,289.504 ns	1.00	0.00	2.6132	8,224 B
Add	PR	?	65536,65536 bits	2,474.616 ns	12.8627 ns	11.4024 ns	2,476.329 ns	2,454.378 ns	2,490.396 ns	1.09	0.01	2.6163	8,224 B

Subtract	main	?	65536,65536 bits	2,298.021 ns	25.6683 ns	22.7543 ns	2,294.305 ns	2,268.096 ns	2,338.891 ns	1.00	0.00	2.6141	8,216 B
Subtract	PR	?	65536,65536 bits	2,523.440 ns	28.7476 ns	24.0056 ns	2,519.823 ns	2,496.089 ns	2,586.340 ns	1.10	0.01	2.6145	8,216 B

Multiply	main	?	65536,65536 bits	927,272.318 ns	18,900.2201 ns	21,765.5216 ns	928,409.371 ns	902,183.483 ns	974,650.134 ns	1.00	0.00	47.7941	153,523 B
Multiply	PR	?	65536,65536 bits	1,004,700.576 ns	3,988.3908 ns	3,730.7432 ns	1,004,817.846 ns	998,809.113 ns	1,011,070.479 ns	1.08	0.02	4.1667	16,414 B

GreatestCommonDivisor	main	?	65536,65536 bits	5,591,038.422 ns	34,612.6535 ns	32,376.6975 ns	5,589,624.021 ns	5,530,768.417 ns	5,649,730.417 ns	1.00	0.00	-	16,451 B
GreatestCommonDivisor	PR	?	65536,65536 bits	6,993,940.330 ns	78,301.4450 ns	73,243.2202 ns	6,972,483.861 ns	6,905,103.250 ns	7,155,806.472 ns	1.25	0.02	-	32 B

jeffhandley · 2021-10-06T00:17:36Z

@sakno Thanks for sharing those performance numbers. We're finally down to the final stages of getting this merged!

@GrabYourPitchforks, @bartonjs, @tannergooding, and I all spent time reviewing the performance results, looking over the changes again, and discussing with each other what the risks and benefits are with these changes. With such a large refactoring, I needed to ensure we're looking at this from all angles, including potential future improvements. We've concluded that the improvements you've made here do indeed put us in a better position overall. Thank you!

I've asked @GrabYourPitchforks to resolve the 2 conflicts that exist right now; he'll do that later this week and push directly to the PR's branch. After that, assuming the CI shows up as green, we'll finally merge this in. These changes will then be part of .NET 7.0 Preview 1.

Thank you so much for your persistence and all of your effort on this, @sakno! This is a significant contribution to .NET.

sakno · 2021-10-06T09:28:14Z

@jeffhandley , @GrabYourPitchforks , thanks for the feedback! I can resolve the conflict by myself without any problem if you want. However, I see the usage of index expression (xd[^1]) that was previously proposed by Tanner and then rejected by Stephen. If we decided to use index expressions then we need to use it everywhere across BigInteger code base. Otherwise, it should be transformed to xd[xd.Length - 1].

stephentoub · 2021-10-06T10:33:53Z

However, I see the usage of index expression (xd[^1]) that was previously proposed by Tanner and then rejected by Stephen.

Where did I reject it?

sakno · 2021-10-06T11:01:29Z

@stephentoub , here is reverting PR: #57297. It was also applied to this PR.
Comment by @tannergooding

Did some brief additional investigation and it looks like the only changes that really need to be reverted are the ones using C# indexing expressions.

The motivation was:

But to your specific question: yes, consider removing range expressions. We don't use them frequently within the base libraries because they tend to interfere with assembly trimming.

stephentoub · 2021-10-06T11:40:19Z

Right, that wasn't me rejecting their usage, that was their usage in those cases showing to regress performance.

jeffhandley · 2021-10-07T04:39:19Z

@sakno Thanks! We'd happily take you up on the offer to resolve the 2 conflicts. And good catch on the range expressions having ~~snuck~~ made it in. In your conflict resolution, let's change those instances back to xd[xd.Length - 1].

sakno · 2021-10-07T10:41:17Z

@jeffhandley , done.

GrabYourPitchforks

We missed one thing in earlier reviews, but otherwise commit 65485fa LGTM! 🥳

GrabYourPitchforks · 2021-10-07T22:16:58Z

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.DivRem.cs

-            uint divHi = right[rightLength - 1];
-            uint divLo = rightLength > 1 ? right[rightLength - 2] : 0;
+            uint divHi = right[right.Length - 1];
+            uint divLo = right.Length > 1 ? right[right.Length - 2] : 0;

            // We measure the leading zeros of the divisor
            int shift = LeadingZeros(divHi);


Cleanup opportunity we didn't spot earlier: get rid of this LeadingZeros method and prefer BitOperations.LeadingZeroCount instead.

GrabYourPitchforks · 2021-10-07T22:24:27Z

@sakno We're at the final step! If you want to address #35565 (comment) (it's just deleting some dead code) as part of this PR, please feel free to do so, otherwise we can tackle it in a follow-up PR. Whatever you decide, I think we're good to merge right after we kick CI a bit.

jeffhandley · 2021-10-08T00:35:01Z

The test failure is #60119, which was fixed in #60140.

sakno · 2021-10-08T07:07:16Z

@GrabYourPitchforks , we have one more potential cleanup in BigIntegerCalculator.Utils file. Here is public static int Compare(ReadOnlySpan<uint> left, ReadOnlySpan<uint> right) method. It can be replaced with MemoryExtensions.SequenceCompareTo method. It uses IComparable<T>.CompareTo constrained call to compare elements. However, the span element is of type unit but I expect that JIT will be able to do inlining for CompareTo implementation

Oh sorry, not applicable. SequenceCompareTo starts from zero element, which is not suitable for BigInteger purposes.

Also, ActualLength can be replaced with MemoryExtensions.TrimEnd everywhere.

jeffhandley

Thank you, @sakno! 💯

dakersnar · 2022-08-15T23:37:47Z

@sakno I'm skimming these changes trying to determine if this PR is the cause of #70330. If you happen to have any insight, let me know.

sakno added 4 commits April 28, 2020 02:31

Unsafe pointers replaced with managed pointers

51260fd

Duplicate code replaced with static local

27cd3b6

Spannified internals

f6d1f89

Spannified compare

d8af4fd

Dotnet-GitSync-Bot added the area-System.Numerics label Apr 28, 2020

sakno added 2 commits April 28, 2020 15:00

Spannified div/rem operations

ca1cf5c

Minus two array allocations for each bitwise operation

d89cdea

monojenkins mentioned this pull request Apr 29, 2020

[WIP] Spannified internals of BigInteger mono/mono#19672

Closed

8 tasks

sakno force-pushed the issue-22609 branch from e8910bf to d89cdea Compare April 29, 2020 10:06

sakno added 2 commits April 29, 2020 13:08

Removed redundant locals

7a6a0d8

Sync with upstream

eb64ad1

gfoidl reviewed Apr 29, 2020

View reviewed changes

Code review feedback (code style)

ba2a01e

tannergooding reviewed Apr 29, 2020

View reviewed changes

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.AddSub.cs Outdated Show resolved Hide resolved

tannergooding reviewed Apr 29, 2020

View reviewed changes

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.AddSub.cs Outdated Show resolved Hide resolved

tannergooding reviewed Apr 29, 2020

View reviewed changes

src/libraries/System.Runtime.Numerics/src/System/Numerics/BigIntegerCalculator.AddSub.cs Outdated Show resolved Hide resolved

ericstj mentioned this pull request Apr 29, 2020

Test failed: System.Text.RegularExpressions.Tests.RegexCacheTests.Ctor_Cache_Promote_entries fails with Timeout #13610

Closed

sakno added 4 commits April 29, 2020 22:00

Code review feedback

49317c5

LeadingZeroes replaced with BitOperations

fa47aea

stackalloc in Divide without remainder

ea7c29b

Reduced memory allocation for bitwise operations

a1e5006

gfoidl reviewed Apr 29, 2020

View reviewed changes

sakno added 6 commits April 30, 2020 02:32

Managed pointer replaced with Span

f19ff53

Removed namespace imports

6e8705f

Removed delegate type

34893a3

Reduced memory allocation in the end of each bitwise operation

5992406

Code review feedback

2b7b7ba

Removed redundant array allocation

03313c1

danmoseley modified the milestones: 6.0.0, 7.0.0 Aug 30, 2021

Sync with upstream

65485fa

This was referenced Oct 7, 2021

System.Tests.TimeZoneInfoTests.TestWindowsNlsDisplayNames is failing on rolling builds #60119

Closed

System.Net.Sockets.Tests.SendPacketsAsync.SendPacketsElement_FileZeroCount_Success sometimes fails #60017

Closed

GrabYourPitchforks approved these changes Oct 7, 2021

View reviewed changes

Sync with upstream

fe39909

Review feedback

602574f

jeffhandley approved these changes Oct 8, 2021

View reviewed changes

jeffhandley merged commit 22b516c into dotnet:main Oct 8, 2021

EgorBo mentioned this pull request Oct 12, 2021

Regressions in System.Numerics.Tests.Perf_BigInteger #60293

Closed

ghost locked as resolved and limited conversation to collaborators Nov 7, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Spannified internals of BigInteger #35565

Spannified internals of BigInteger #35565

sakno commented Apr 28, 2020 •

edited

Loading

ghost commented Apr 28, 2020

gfoidl left a comment •

edited

Loading

sakno commented Apr 29, 2020

danmoseley commented Aug 26, 2021

danmoseley commented Aug 26, 2021

stephentoub commented Aug 26, 2021 •

edited

Loading

danmoseley commented Aug 30, 2021

GrabYourPitchforks commented Aug 30, 2021

danmoseley commented Aug 30, 2021

sakno commented Sep 6, 2021

jeffhandley commented Oct 6, 2021

sakno commented Oct 6, 2021

stephentoub commented Oct 6, 2021

sakno commented Oct 6, 2021

stephentoub commented Oct 6, 2021 •

edited

Loading

jeffhandley commented Oct 7, 2021

sakno commented Oct 7, 2021

GrabYourPitchforks left a comment

GrabYourPitchforks Oct 7, 2021

GrabYourPitchforks commented Oct 7, 2021

jeffhandley commented Oct 8, 2021

sakno commented Oct 8, 2021 •

edited

Loading

jeffhandley left a comment

dakersnar commented Aug 15, 2022

Spannified internals of BigInteger #35565

Spannified internals of BigInteger #35565

Conversation

sakno commented Apr 28, 2020 • edited Loading

ghost commented Apr 28, 2020

gfoidl left a comment • edited Loading

Choose a reason for hiding this comment

sakno commented Apr 29, 2020

danmoseley commented Aug 26, 2021

danmoseley commented Aug 26, 2021

stephentoub commented Aug 26, 2021 • edited Loading

danmoseley commented Aug 30, 2021

GrabYourPitchforks commented Aug 30, 2021

danmoseley commented Aug 30, 2021

sakno commented Sep 6, 2021

jeffhandley commented Oct 6, 2021

sakno commented Oct 6, 2021

stephentoub commented Oct 6, 2021

sakno commented Oct 6, 2021

stephentoub commented Oct 6, 2021 • edited Loading

jeffhandley commented Oct 7, 2021

sakno commented Oct 7, 2021

GrabYourPitchforks left a comment

Choose a reason for hiding this comment

GrabYourPitchforks Oct 7, 2021

Choose a reason for hiding this comment

GrabYourPitchforks commented Oct 7, 2021

jeffhandley commented Oct 8, 2021

sakno commented Oct 8, 2021 • edited Loading

jeffhandley left a comment

Choose a reason for hiding this comment

dakersnar commented Aug 15, 2022

sakno commented Apr 28, 2020 •

edited

Loading

gfoidl left a comment •

edited

Loading

stephentoub commented Aug 26, 2021 •

edited

Loading

stephentoub commented Oct 6, 2021 •

edited

Loading

sakno commented Oct 8, 2021 •

edited

Loading