Perf: Consolidate implementation of LeadingZeroCount #22497

grant-d · 2019-02-09T01:22:32Z

There are several implementations of LeadingZeroCount in the stack, with differing or branch-heavy implementations.

This PR consolidates them into a central implementation that uses intrinsics & optimized software fallbacks.
Also fixes an issue in a related PR where explicit conversions to (IntPtr) was missing, causing some units to fail. Note the idiom used to convert from long to IntPtr inexpensively.
All BitOps methods pass units

cc @tannergooding

src/System.Private.CoreLib/shared/System/Buffers/Text/FormattingHelpers.CountDigits.cs

src/System.Private.CoreLib/shared/System/Decimal.DecCalc.cs

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs

src/System.Private.CoreLib/shared/System/SpanHelpers.Char.cs

src/System.Private.CoreLib/shared/System/SpanHelpers.Byte.cs

src/System.Private.CoreLib/shared/System/BitOps.cs

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs

src/System.Private.CoreLib/shared/System/BitOps.cs

src/System.Private.CoreLib/shared/System/Buffers/Utilities.cs

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs

src/System.Private.CoreLib/shared/System/SpanHelpers.Byte.cs

src/System.Private.CoreLib/shared/System/BitOps.cs

dotnet-bot · 2019-02-12T01:26:47Z

Welcome to the dotnet/coreclr Perf help

The following is a list of valid commands on this PR. To invoke a command, comment the indicated phrase on the PR

The following commands are valid for all PRs and repositories.

Click to expand

Comment Phrase	Action
@dotnet-bot test this please	Re-run all legs. Use sparingly
@dotnet-bot test ci please	Generates (but does not run) jobs based on changes to the groovy job definitions in this branch
@dotnet-bot help	Print this help message

The following jobs are launched by default for each PR against dotnet/coreclr:master.

Click to expand

Comment Phrase	Job Launched
@dotnet-bot test \QWindows_NT x64 full_opt ryujit CoreCLR Perf Tests Correctness\E	Windows_NT x64 full_opt ryujit CoreCLR Perf Tests Correctness
@dotnet-bot test \QWindows_NT x64 min_opt ryujit CoreCLR Perf Tests Correctness\E	Windows_NT x64 min_opt ryujit CoreCLR Perf Tests Correctness
@dotnet-bot test \QWindows_NT x86 full_opt ryujit CoreCLR Perf Tests Correctness\E	Windows_NT x86 full_opt ryujit CoreCLR Perf Tests Correctness
@dotnet-bot test \QWindows_NT x86 min_opt ryujit CoreCLR Perf Tests Correctness\E	Windows_NT x86 min_opt ryujit CoreCLR Perf Tests Correctness

The following optional jobs are available in PRs against dotnet/coreclr:master.

Click to expand

Comment Phrase	Job Launched
@dotnet-bot test Windows_NT x64 illink	Queues Windows_NT x64 full_opt ryujit IlLink Tests
@dotnet-bot test linux perf flow	Queues Linux Perf Test Flow
@dotnet-bot test Windows_NT x64 perf	Queues Windows_NT x64 full_opt ryujit CoreCLR Perf Tests
@dotnet-bot test Windows_NT x64 min_opts perf	Queues Windows_NT x64 min_opt ryujit CoreCLR Perf Tests
@dotnet-bot test Windows_NT x86 perf	Queues Windows_NT x86 full_opt ryujit CoreCLR Perf Tests
@dotnet-bot test Windows_NT x86 min_opts perf	Queues Windows_NT x86 min_opt ryujit CoreCLR Perf Tests
@dotnet-bot test Windows_NT x64 perf scenarios	Queues Windows_NT x64 full_opt ryujit Performance Scenarios Tests
@dotnet-bot test Windows_NT x64 min_opts perf scenarios	Queues Windows_NT x64 min_opt ryujit Performance Scenarios Tests
@dotnet-bot test Windows_NT x64 perf scenarios	Queues Windows_NT x64 tiered ryujit Performance Scenarios Tests
@dotnet-bot test Windows_NT x86 perf scenarios	Queues Windows_NT x86 full_opt ryujit Performance Scenarios Tests
@dotnet-bot test Windows_NT x86 min_opts perf scenarios	Queues Windows_NT x86 min_opt ryujit Performance Scenarios Tests
@dotnet-bot test Windows_NT x86 perf scenarios	Queues Windows_NT x86 tiered ryujit Performance Scenarios Tests
@dotnet-bot test linux throughput flow	Queues Linux arm Throughput Perf Test Flow
@dotnet-bot test linux throughput flow	Queues Linux x64 Throughput Perf Test Flow
@dotnet-bot test Windows_NT x64 throughput	Queues Windows_NT x64 full_opt ryujit nopgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x64 nopgo throughput	Queues Windows_NT x64 full_opt ryujit pgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x64 min_opts throughput	Queues Windows_NT x64 min_opt ryujit nopgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x64 min_opts nopgo throughput	Queues Windows_NT x64 min_opt ryujit pgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x86 throughput	Queues Windows_NT x86 full_opt ryujit nopgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x86 nopgo throughput	Queues Windows_NT x86 full_opt ryujit pgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x86 min_opts throughput	Queues Windows_NT x86 min_opt ryujit nopgo CoreCLR Throughput Perf Tests
@dotnet-bot test Windows_NT x86 min_opts nopgo throughput	Queues Windows_NT x86 min_opt ryujit pgo CoreCLR Throughput Perf Tests

Have a nice day!

src/System.Private.CoreLib/shared/System/BitOps.cs

jkotas

Thanks

safern · 2019-02-13T17:58:05Z

@safern /azp commands works also for coreclr-ci? dotnet/corefx#35121 (comment)

The outerloop builds shouldn't since those are corefx specific, but the concept of everything else is the same.

grant-d · 2019-02-13T19:18:58Z

@dotnet-bot test OSX10.12 x64 Checked Innerloop Build and Test please
@dotnet-bot test Windows_NT x64 Release CoreFX Tests please
@dotnet-bot test Ubuntu x64 Checked Innerloop Build and Test (Jit - TieredCompilation=0) please

grant-d · 2019-02-13T21:41:53Z

@dotnet-bot test OSX10.12 x64 Checked Innerloop Build and Test please

src/System.Private.CoreLib/shared/System/BitOps.cs

tannergooding · 2019-02-13T22:17:24Z

Thanks for the PR here @grant-d.

I'm fine with the two nits I had being looked at separately.

…22497) * Perf: BitOps.LeadingZeroCount * CR fix * CR fixes * Optimization * Revert * Confirmed fix * Simplify * Return int * Fixes * Simplify * CR fixes * CR fixes * Fix unit failure on Linux-musl x64 Debug * CR fixes * CR fixes * Cleanup Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>

grant-d · 2019-02-13T22:23:32Z

Thanks @tannergooding, @jkotas and @davidfowl
Appreciate the help and pointers

…22497) * Perf: BitOps.LeadingZeroCount * CR fix * CR fixes * Optimization * Revert * Confirmed fix * Simplify * Return int * Fixes * Simplify * CR fixes * CR fixes * Fix unit failure on Linux-musl x64 Debug * CR fixes * CR fixes * Cleanup Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>

…22497) * Perf: BitOps.LeadingZeroCount * CR fix * CR fixes * Optimization * Revert * Confirmed fix * Simplify * Return int * Fixes * Simplify * CR fixes * CR fixes * Fix unit failure on Linux-musl x64 Debug * CR fixes * CR fixes * Cleanup Commit migrated from dotnet/coreclr@eee8c24

Commit migrated from dotnet/coreclr@361e7dc

Perf: BitOps.LeadingZeroCount

3011abd

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Buffers/Text/FormattingHelpers.CountDigits.cs Show resolved Hide resolved

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Decimal.DecCalc.cs Show resolved Hide resolved

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs Outdated Show resolved Hide resolved

CR fix

334c816

grant-d changed the title ~~Perf: BitOps.LeadingZeroCount~~ Perf: Consolidate implementation of LeadingZeroCount Feb 9, 2019

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/SpanHelpers.Char.cs Show resolved Hide resolved

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/SpanHelpers.Byte.cs Show resolved Hide resolved

omariom reviewed Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Outdated Show resolved Hide resolved

Grant Dickinson added 4 commits February 8, 2019 18:08

CR fixes

fb57241

Optimization

c2b5219

Revert

457730c

Confirmed fix

f4bbf4d

grant-d commented Feb 9, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs Outdated Show resolved Hide resolved

Grant Dickinson added 4 commits February 8, 2019 19:04

Simplify

098bc6e

Return int

10d3794

Fixes

21dfb03

Simplify

2118e0c

davidfowl reviewed Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Outdated Show resolved Hide resolved

CR fixes

a1cb73d

grant-d commented Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Buffers/Utilities.cs Show resolved Hide resolved

jkotas reviewed Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/Number.BigInteger.cs Outdated Show resolved Hide resolved

jkotas reviewed Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/SpanHelpers.Byte.cs Outdated Show resolved Hide resolved

jkotas reviewed Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/SpanHelpers.Byte.cs Show resolved Hide resolved

Grant Dickinson added 2 commits February 10, 2019 20:05

CR fixes

406ee5f

Fix unit failure on Linux-musl x64 Debug

bff472e

grant-d commented Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Outdated Show resolved Hide resolved

CR fixes

df100cf

grant-d commented Feb 11, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Show resolved Hide resolved

grant-d commented Feb 13, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Show resolved Hide resolved

grant-d commented Feb 13, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Show resolved Hide resolved

Cleanup

2ccb5b3

jkotas approved these changes Feb 13, 2019

View reviewed changes

tannergooding reviewed Feb 13, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Show resolved Hide resolved

tannergooding reviewed Feb 13, 2019

View reviewed changes

src/System.Private.CoreLib/shared/System/BitOps.cs Show resolved Hide resolved

tannergooding approved these changes Feb 13, 2019

View reviewed changes

tannergooding merged commit eee8c24 into dotnet:master Feb 13, 2019

grant-d deleted the grant-d.lzcnt branch February 13, 2019 22:21

grant-d mentioned this pull request Feb 13, 2019

Consolidate implementation of Rotate and PopCount #22584

Merged

This was referenced Feb 14, 2019

Consolidate implementation of ExtractBit #22225

Closed

Units for BitOps.TrailingZeroCount dotnet/corefx#35193

Merged

Fix perf issue related to LeadingZeroCount #22612

Merged

jkotas pushed a commit that referenced this pull request Feb 15, 2019

Missed optimization from #22497

361e7dc

grant-d mentioned this pull request Feb 1, 2020

Proposal: Expose existing BitOps methods dotnet/runtime#28725

Closed

picenka21 pushed a commit to picenka21/runtime that referenced this pull request Feb 18, 2022

Missed optimization from dotnet/coreclr#22497

ea71ca7

Commit migrated from dotnet/coreclr@361e7dc

Perf: Consolidate implementation of LeadingZeroCount #22497

Perf: Consolidate implementation of LeadingZeroCount #22497

Uh oh!

Conversation

grant-d commented Feb 9, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dotnet-bot commented Feb 12, 2019

Uh oh!

Uh oh!

Uh oh!

jkotas left a comment

Choose a reason for hiding this comment

Uh oh!

safern commented Feb 13, 2019

Uh oh!

grant-d commented Feb 13, 2019

Uh oh!

grant-d commented Feb 13, 2019

Uh oh!

Uh oh!

Uh oh!

tannergooding commented Feb 13, 2019

Uh oh!

grant-d commented Feb 13, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

9 participants

grant-d commented Feb 9, 2019 •

edited

Loading