Add tape type mechanisms with parent_job capability #734
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Codecov ReportAttention: Patch coverage is
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## main #734 +/- ##
==========================================
- Coverage 77.34% 77.08% -0.26%
==========================================
Files 18 18
Lines 7377 7407 +30
==========================================
+ Hits 5706 5710 +4
- Misses 1671 1697 +26 ☔ View full report in Codecov by Sentry. |
vchuravy
reviewed
Apr 23, 2023
| @@ -2774,8 +2778,8 @@ end | |||
| subfunc = nothing | |||
| if mode == API.DEM_ForwardMode | |||
| if fwdmodenm === nothing | |||
| etarget = Compiler.EnzymeTarget() | |||
| eparams = Compiler.EnzymeCompilerParams(Tuple{(dupClosure ? Duplicated : Const){funcT}, e_tt.parameters...}, API.DEM_ForwardMode, width, Const{Nothing}, #=runEnzyme=#true, #=abiwrap=#true, modifiedBetween, #=returnPrimal=#false, #=shadowInit=#false, UnknownTapeType) | |||
| etarget = Compiler.EnzymeTarget(job.config.target.parent_target) | |||
There was a problem hiding this comment.
Suggested change
| etarget = Compiler.EnzymeTarget(job.config.target.parent_target) | |
| etarget = job.config.target |
vchuravy
reviewed
Apr 23, 2023
| @@ -6197,10 +6200,25 @@ end | |||
|
|
|||
| # Define EnzymeTarget | |||
| Base.@kwdef struct EnzymeTarget <: AbstractCompilerTarget | |||
| parent_target::Union{Nothing, AbstractCompilerTarget} | |||
| end | |||
| GPUCompiler.llvm_triple(T::EnzymeTarget) = Sys.MACHINE ? T.parent_target === nothing : GPUCompiler.llvm_triple(T.parent_target) | |||
There was a problem hiding this comment.
Suggested change
| GPUCompiler.llvm_triple(T::EnzymeTarget) = Sys.MACHINE ? T.parent_target === nothing : GPUCompiler.llvm_triple(T.parent_target) | |
| GPUCompiler.llvm_triple(T::EnzymeTarget) = T.parent_target === nothing ? Sys.MACHINE : GPUCompiler.llvm_triple(T.parent_target) |
vchuravy
reviewed
Apr 23, 2023
|
|
||
| function GPUCompiler.process_module!(@nospecialize(job::CompilerJob{EnzymeTarget}), mod::LLVM.Module) | ||
| if job.target.parent_target !== nothing | ||
| # process_module!(similar(job, job.target.parent_target), mod) |
vchuravy
reviewed
Apr 23, 2023
vchuravy
reviewed
Apr 23, 2023
| @@ -8782,8 +8805,8 @@ import GPUCompiler: deferred_codegen_jobs | |||
| @generated function deferred_codegen(::Val{World}, ::Type{FA}, ::Val{tt}, ::Val{rt},::Val{Mode}, | |||
| ::Val{width}, ::Val{ModifiedBetween}, ::Val{ReturnPrimal}=Val(false),::Val{ShadowInit}=Val(false),::Type{ExpectedTapeType}=UnknownTapeType) where {World, FA<:Annotation,tt, rt, Mode, width, ModifiedBetween, ReturnPrimal, ShadowInit,ExpectedTapeType} | |||
| mi = fspec(eltype(FA), tt, World) | |||
| target = EnzymeTarget() | |||
| params = EnzymeCompilerParams(Tuple{FA, tt.parameters...}, Mode, width, remove_innerty(rt), true, #=abiwrap=#true, ModifiedBetween, ReturnPrimal, ShadowInit,ExpectedTapeType) | |||
| target = EnzymeTarget(nothing) | |||
There was a problem hiding this comment.
Make this NativeTarget then we don't need to worry about nothing vs defined?
vchuravy
reviewed
Apr 24, 2023
| etarget = Compiler.EnzymeTarget() | ||
| eparams = Compiler.EnzymeCompilerParams(Tuple{(dupClosure ? Duplicated : Const){funcT}, e_tt.parameters...}, API.DEM_ForwardMode, width, Const{Nothing}, #=runEnzyme=#true, #=abiwrap=#true, modifiedBetween, #=returnPrimal=#false, #=shadowInit=#false, UnknownTapeType) | ||
| etarget = Compiler.EnzymeTarget(job.config.target.parent_target) | ||
| eparams = Compiler.EnzymeCompilerParams(Tuple{(dupClosure ? Duplicated : Const){funcT}, e_tt.parameters...}, API.DEM_ForwardMode, width, Const{Nothing}, #=runEnzyme=#true, #=abiwrap=#true, modifiedBetween, #=returnPrimal=#false, #=shadowInit=#false, UnknownTapeType, GPUCompiler.method_table(job)) | ||
| ejob = Compiler.CompilerJob(mi2, CompilerConfig(etarget, eparams; kernel=false), world) |
There was a problem hiding this comment.
Btw there is a copy-constructor https://github.com/JuliaGPU/GPUCompiler.jl/blob/d5086fb3d93bbc4795a96f6f1457898af46a24cb/src/interface.jl#L111-L115
|
Rebase after #1104 ? there still seems to be some useful things in here. |
|
Superceeded by present functionality |
|
hm @vchuravy you're right maybe there is something still useful here |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.