allow multiple inlining again

Before #52415 we could do multiple rounds of inlining on
IR that had already been inlined, but this was no longer possible. By
removing the assertion that was introduced in #52415,
this commit makes it possible to do multi-inlining once more. Note that
to fully solve this, though, we need to enhance `ir_inline_linetable!`
so it can add new linetables to an inner linetable that's already been
inlined. This commit notes that enhancement as something we need to do
later, leaving it with a TODO comment.

Author: aviatesk

base/compiler/ssair/inlining.jl

@@ -300,7 +300,8 @@ function finish_cfg_inline!(state::CFGInliningState)
     end
 end
 
-function ir_inline_linetable!(debuginfo::DebugInfoStream, inlinee_debuginfo::DebugInfo, inlinee::MethodInstance)
+# TODO append `inlinee_debuginfo` to inner linetable when `inlined_at[2] ≠ 0`
+function ir_inline_linetable!(debuginfo::DebugInfoStream, inlinee_debuginfo::DebugInfo, inlined_at::NTuple{3,Int32})
     # Append the linetable of the inlined function to our edges table
     linetable_offset = 1
     while true
@@ -312,16 +313,14 @@ function ir_inline_linetable!(debuginfo::DebugInfoStream, inlinee_debuginfo::Deb
         end
         linetable_offset += 1
     end
-    return Int32(linetable_offset)
+    return (inlined_at[1], Int32(linetable_offset), Int32(0))
 end
 
 function ir_prepare_inlining!(insert_node!::Inserter, inline_target::Union{IRCode, IncrementalCompact},
-                              ir::IRCode, di::DebugInfo, mi::MethodInstance, inlined_at::Int32, argexprs::Vector{Any})
+                              ir::IRCode, di::DebugInfo, mi::MethodInstance, inlined_at::NTuple{3,Int32}, argexprs::Vector{Any})
     def = mi.def::Method
     debuginfo = inline_target isa IRCode ? inline_target.debuginfo : inline_target.ir.debuginfo
-
-    linetable_offset = ir_inline_linetable!(debuginfo, di, mi)
-    topline = (inlined_at, linetable_offset, Int32(0))
+    topline = new_inlined_at = ir_inline_linetable!(debuginfo, di, inlined_at)
     if should_insert_coverage(def.module, di)
         insert_node!(NewInstruction(Expr(:code_coverage_effect), Nothing, topline))
     end
@@ -343,7 +342,7 @@ function ir_prepare_inlining!(insert_node!::Inserter, inline_target::Union{IRCod
             NewInstruction(Expr(:call, GlobalRef(Core, :getfield), argexprs[1], QuoteNode(:captures)),
             ir.argtypes[1], topline))
     end
-    return SSASubstitute(mi, argexprs, spvals_ssa, (inlined_at, linetable_offset))
+    return SSASubstitute(mi, argexprs, spvals_ssa, new_inlined_at)
 end
 
 function adjust_boundscheck!(inline_compact::IncrementalCompact, idx′::Int, stmt::Expr, boundscheck::Symbol)
@@ -359,8 +358,7 @@ function ir_inline_item!(compact::IncrementalCompact, idx::Int, argexprs::Vector
                          item::InliningTodo, boundscheck::Symbol, todo_bbs::Vector{Tuple{Int, Int}})
     # Ok, do the inlining here
     inlined_at = compact.result[idx][:line]
-    @assert inlined_at[2] == 0 "already inlined this instruction"
-    ssa_substitute = ir_prepare_inlining!(InsertHere(compact), compact, item.ir, item.di, item.mi, inlined_at[1], argexprs)
+    ssa_substitute = ir_prepare_inlining!(InsertHere(compact), compact, item.ir, item.di, item.mi, inlined_at, argexprs)
     boundscheck = has_flag(compact.result[idx], IR_FLAG_INBOUNDS) ? :off : boundscheck
 
     # If the iterator already moved on to the next basic block,
@@ -1771,7 +1769,7 @@ struct SSASubstitute
     mi::MethodInstance
     arg_replacements::Vector{Any}
     spvals_ssa::Union{Nothing,SSAValue}
-    inlined_at::NTuple{2,Int32} # TODO: add a map also, so that ssaidx doesn't need to equal inlined_idx?
+    inlined_at::NTuple{3,Int32} # TODO: add a map also, so that ssaidx doesn't need to equal inlined_idx?
 end
 
 function insert_spval!(insert_node!::Inserter, spvals_ssa::SSAValue, spidx::Int, do_isdefined::Bool)

base/compiler/ssair/passes.jl

@@ -1516,7 +1516,7 @@ function try_inline_finalizer!(ir::IRCode, argexprs::Vector{Any}, idx::Int,
     add_inlining_backedge!(et, mi)
 
     # TODO: Should there be a special line number node for inlined finalizers?
-    inline_at = ir[SSAValue(idx)][:line][1]
+    inline_at = ir[SSAValue(idx)][:line]
     ssa_substitute = ir_prepare_inlining!(InsertBefore(ir, SSAValue(idx)), ir, src, di, mi, inline_at, argexprs)
 
     # TODO: Use the actual inliner here rather than open coding this special purpose inliner.

test/compiler/inline.jl

@@ -1818,23 +1818,55 @@ let src = code_typed1((AtomicMemoryRef{Int},)) do a
     @test count(isinvokemodify(:+), src.code) == 1
 end
 
-
-
 # apply `ssa_inlining_pass` multiple times
-let interp = Core.Compiler.NativeInterpreter()
+func_mul_int(a::Int, b::Int) = Core.Intrinsics.mul_int(a, b)
+multi_inlining1(a::Int, b::Int) = @noinline func_mul_int(a, b)
+let i::Int, continue_::Bool
+    interp = Core.Compiler.NativeInterpreter()
     # check if callsite `@noinline` annotation works
-    ir, = Base.code_ircode((Int,Int); optimize_until="inlining", interp) do a, b
-        @noinline a*b
-    end |> only
-    i = findfirst(isinvoke(:*), ir.stmts.stmt)
+    ir, = only(Base.code_ircode(multi_inlining1, (Int,Int); optimize_until="inlining", interp))
+    i = findfirst(isinvoke(:func_mul_int), ir.stmts.stmt)
     @test i !== nothing
-
-    # ok, now delete the callsite flag, and see the second inlining pass can inline the call
+    # now delete the callsite flag, and see the second inlining pass can inline the call
     @eval Core.Compiler $ir.stmts[$i][:flag] &= ~IR_FLAG_NOINLINE
     inlining = Core.Compiler.InliningState(interp)
     ir = Core.Compiler.ssa_inlining_pass!(ir, inlining, false)
-    @test count(isinvoke(:*), ir.stmts.stmt) == 0
-    @test count(iscall((ir, Core.Intrinsics.mul_int)), ir.stmts.stmt) == 1
+    @test findfirst(isinvoke(:func_mul_int), ir.stmts.stmt) === nothing
+    @test (i = findfirst(iscall((ir, Core.Intrinsics.mul_int)), ir.stmts.stmt)) !== nothing
+    lins = Base.IRShow.buildLineInfoNode(ir.debuginfo, nothing, i)
+    @test (continue_ = length(lins) == 2) # :multi_inlining1 -> :func_mul_int
+    if continue_
+        def1 = lins[1].method
+        @test def1 isa Core.MethodInstance && def1.def.name === :multi_inlining1
+        def2 = lins[2].method
+        @test def2 isa Core.MethodInstance && def2.def.name === :func_mul_int
+    end
+end
+
+call_func_mul_int(a::Int, b::Int) = @noinline func_mul_int(a, b)
+multi_inlining2(a::Int, b::Int) = call_func_mul_int(a, b)
+let i::Int, continue_::Bool
+    interp = Core.Compiler.NativeInterpreter()
+    # check if callsite `@noinline` annotation works
+    ir, = only(Base.code_ircode(multi_inlining2, (Int,Int); optimize_until="inlining", interp))
+    i = findfirst(isinvoke(:func_mul_int), ir.stmts.stmt)
+    @test i !== nothing
+    # now delete the callsite flag, and see the second inlining pass can inline the call
+    @eval Core.Compiler $ir.stmts[$i][:flag] &= ~IR_FLAG_NOINLINE
+    inlining = Core.Compiler.InliningState(interp)
+    ir = Core.Compiler.ssa_inlining_pass!(ir, inlining, false)
+    @test findfirst(isinvoke(:func_mul_int), ir.stmts.stmt) === nothing
+    @test (i = findfirst(iscall((ir, Core.Intrinsics.mul_int)), ir.stmts.stmt)) !== nothing
+    lins = Base.IRShow.buildLineInfoNode(ir.debuginfo, nothing, i)
+    @test_broken (continue_ = length(lins) == 3) # see TODO in `ir_inline_linetable!`
+    if continue_
+        def1 = lins[1].method
+        @test def1 isa Core.MethodInstance && def1.def.name === :multi_inlining2
+        def2 = lins[2].method
+        @test def2 isa Core.MethodInstance && def2.def.name === :call_func_mul_int
+        def3 = lins[3].method
+        @test def3 isa Core.MethodInstance && def3.def.name === :call_func_mul_int
+    end
 end
 
 # Test special purpose inliner for Core.ifelse