Skip to content

[RyuJIT] Unroll single-iteration loops #43947

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
Merged
sandreenko merged 6 commits into from
Nov 12, 2020
Merged

[RyuJIT] Unroll single-iteration loops #43947

sandreenko merged 6 commits into from
Nov 12, 2020

Conversation

@EgorBo
Copy link
Member

@EgorBo EgorBo commented Oct 28, 2020 •
edited
Loading

JIT is able to unroll small constant-size loops via optUnrollLoops, but currently it only works if Vector_.Count is used (magic) as the upper limit. This PR extends that behavior to also handle single iteration loops (thus it won't regress anything)

static int GetCount() => 1; // e.g. static readonly field inited via an envvar

static int Test()
{
    int x = 0;
    for (int i = 0; i < GetCount(); i++)
        x++;
    return x;
}

Current codegen for Test:

G_M58954_IG01:
G_M58954_IG02:
       xor      eax, eax
       xor      edx, edx
G_M58954_IG03:
       inc      eax
       inc      edx
       test     edx, edx
       jle      SHORT G_M58954_IG03
G_M58954_IG04:
       ret

New codegen for Test:

G_M58954_IG02:
       mov      eax, 1
G_M58954_IG03:
       ret

jit-diff (-f --cctors --pmi):

(surprisingly, it's not empty)

Total bytes of delta: -352 (-0.00% of base)
    diff is an improvement.

Top file improvements (bytes):
        -211 : System.Private.CoreLib.dasm (-0.00% of base)
        -103 : Microsoft.CodeAnalysis.dasm (-0.01% of base)
         -38 : Microsoft.Diagnostics.Tracing.TraceEvent.dasm (-0.00% of base)

3 total files with Code Size differences (3 improved, 0 regressed), 265 unchanged.

Top method improvements (bytes):
        -103 (-12.50% of base) : Microsoft.CodeAnalysis.dasm - SmallDictionary`2:LeftComplex(AvlNode):AvlNode (8 base, 7 diff methods)
         -66 (-12.69% of base) : System.Private.CoreLib.dasm - Vector64`1:Equals(Vector64`1):bool:this (6 methods)
         -66 (-11.21% of base) : System.Private.CoreLib.dasm - Vector64`1:GetHashCode():int:this (6 methods)
         -62 (-4.52% of base) : System.Private.CoreLib.dasm - Utf8Formatter:TryFormat(TimeSpan,Span`1,byref,StandardFormat):bool
         -17 (-5.28% of base) : System.Private.CoreLib.dasm - Utf8Parser:TryParseByteX(ReadOnlySpan`1,byref,byref):bool
         -14 (-24.56% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:get_FileDescription():String:this
         -13 (-1.53% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:ToXml(StringBuilder):StringBuilder:this
         -11 (-1.75% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:PayloadValue(int):Object:this

Top method improvements (percentages):
         -14 (-24.56% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:get_FileDescription():String:this
         -66 (-12.69% of base) : System.Private.CoreLib.dasm - Vector64`1:Equals(Vector64`1):bool:this (6 methods)
        -103 (-12.50% of base) : Microsoft.CodeAnalysis.dasm - SmallDictionary`2:LeftComplex(AvlNode):AvlNode (8 base, 7 diff methods)
         -66 (-11.21% of base) : System.Private.CoreLib.dasm - Vector64`1:GetHashCode():int:this (6 methods)
         -17 (-5.28% of base) : System.Private.CoreLib.dasm - Utf8Parser:TryParseByteX(ReadOnlySpan`1,byref,byref):bool
         -62 (-4.52% of base) : System.Private.CoreLib.dasm - Utf8Formatter:TryFormat(TimeSpan,Span`1,byref,StandardFormat):bool
         -11 (-1.75% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:PayloadValue(int):Object:this
         -13 (-1.53% of base) : Microsoft.Diagnostics.Tracing.TraceEvent.dasm - FileVersionTraceData:ToXml(StringBuilder):StringBuilder:this

8 total methods with Code Size differences (8 improved, 0 regressed), 259114 unchanged.
Completed analysis in 23.84

Example: Utf8Parser:TryParseByteX (ByteOverflowLengthHex = 2) - looks suspicious btw:

@Dotnet-GitSync-Bot Dotnet-GitSync-Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Oct 28, 2020
@EgorBo
Copy link
Member Author

EgorBo commented Nov 5, 2020

@dotnet/jit-contrib PTAL

@AndyAyersMS
Copy link
Member

AndyAyersMS commented Nov 5, 2020

cc @BruceForstall

BruceForstall
BruceForstall reviewed Nov 6, 2020
View reviewed changes
Copy link
Contributor

@BruceForstall BruceForstall left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Question

sandreenko
sandreenko approved these changes Nov 11, 2020
View reviewed changes
Copy link
Contributor

@sandreenko sandreenko left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, thank you.

@sandreenko sandreenko merged commit 51393be into dotnet:master Nov 12, 2020
@ghost ghost locked as resolved and limited conversation to collaborators Dec 12, 2020
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Reviewers

2 more reviewers

@BruceForstall BruceForstall BruceForstall left review comments

@sandreenko sandreenko sandreenko approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@EgorBo @AndyAyersMS @BruceForstall @sandreenko @Dotnet-GitSync-Bot