[SwiftCompilerSources] Bridged in-IR testing. #68813
Conversation
nate-chandler
force-pushed
the
swiftcompilersources/test-bridging
branch
from
82c8a63 to
487dc11
Compare
|
@swift-ci please test |
| self.name._withStringRef { nameStr in | ||
| registerFunctionTest( | ||
| nameStr, | ||
| unsafeBitCast(impl, to: UnsafeMutableRawPointer.self)) |
There was a problem hiding this comment.
Can you define the closure as @convention(c) to avoid the unsafe bitcast? Like it's done for the run-function in registerPass
There was a problem hiding this comment.
Yep, with the FunctionTest as a global struct whose member can be called which does the unwrapping, that works. Fixed.
| @_semantics("arc.immortal") | ||
| public class FunctionTest { | ||
| public var name: String | ||
| init(name: String, impl: @convention(thin) (Function, TestArguments, FunctionTest) -> ()) { |
There was a problem hiding this comment.
I tried to follow the convention to not use abbreviations in function API names (which includes argument labels). Therefore I suggest impl: -> implementation:
| import SILBridging | ||
|
|
||
| @_semantics("arc.immortal") | ||
| public class FunctionTest { |
There was a problem hiding this comment.
Does this really need to be a class? And does it need to bridge to C++ (via the swift object header)?
On the C++ side, a new FunctionTest is created anyway.
I think it can be a struct, similar to FunctionPass
There was a problem hiding this comment.
Thanks! Switched over to that.
| { bridgedFunction, bridgedTestArguments, bridgedFunctionTest, ctx in | ||
| typealias Impl = @convention(thin) (Function, TestArguments, FunctionTest) -> () | ||
| let impl = unsafeBitCast(ctx, to: Impl.self) | ||
| impl(bridgedFunction.function, bridgedTestArguments.native, bridgedFunctionTest.native) |
There was a problem hiding this comment.
IIUC, the reason to pass the test to the closure is to get things like dominance info. In the spirit of swift passes, it's better to pass a TestContext which can provide everything which is potentially needed by a test.
Though, it will be a problem that this context provides an analysis, like the dominator tree, because this is defined in the optimizer module. Anyway, that can be solved.
There was a problem hiding this comment.
Right, it's to make analyses available. Added a Bridged/TestContext that wraps a SwiftPassInvocation. Later on I'll add extensions to make the analyses available.
| } | ||
|
|
||
| @_semantics("arc.immortal") | ||
| public class TestArguments { |
There was a problem hiding this comment.
again, does this need to be a bridged class?
nate-chandler
force-pushed
the
swiftcompilersources/test-bridging
branch
from
487dc11 to
076c520
Compare
|
@swift-ci please test |
nate-chandler
force-pushed
the
swiftcompilersources/test-bridging
branch
from
076c520 to
48035ab
Compare
|
@swift-ci please test |
There's no advantage to using an function_ref and it might be confusing.
Added the bridging types involved and the basic functionality.
nate-chandler
force-pushed
the
swiftcompilersources/test-bridging
branch
from
48035ab to
c9389e1
Compare
Yep! Done (with appropriate translation). |
|
@swift-ci please test |
To prove out that the bridging works.
nate-chandler
force-pushed
the
swiftcompilersources/test-bridging
branch
from
c9389e1 to
36805c8
Compare
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
|
@swift-ci please clean test linux platform |
|
@swift-ci please smoke test linux platform |
|
@swift-ci please test linux platform |
| // | ||
| // CASE STUDY: | ||
| // Here's an example of how it works: | ||
| // 0) A source file, NeatUtils.cpp contains |
There was a problem hiding this comment.
I guess this is supposed to be NeatUtils.swift
There was a problem hiding this comment.
Oops! #68924 .
Added the bridging types involved and the basic functionality.