[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
// Copyright 2021 The Go Authors. All rights reserved.
|
|
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
|
|
|
|
|
package noder
|
|
|
|
|
|
|
|
|
|
import (
|
2021-01-22 13:29:59 -08:00
|
|
|
"go/constant"
|
|
|
|
|
|
|
|
|
|
"cmd/compile/internal/base"
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
"cmd/compile/internal/ir"
|
|
|
|
|
"cmd/compile/internal/typecheck"
|
|
|
|
|
"cmd/compile/internal/types"
|
|
|
|
|
"cmd/internal/src"
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
// Helpers for constructing typed IR nodes.
|
|
|
|
|
//
|
|
|
|
|
// TODO(mdempsky): Move into their own package so they can be easily
|
|
|
|
|
// reused by iimport and frontend optimizations.
|
|
|
|
|
|
2021-01-22 13:29:59 -08:00
|
|
|
type ImplicitNode interface {
|
|
|
|
|
ir.Node
|
|
|
|
|
SetImplicit(x bool)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Implicit returns n after marking it as Implicit.
|
|
|
|
|
func Implicit(n ImplicitNode) ImplicitNode {
|
|
|
|
|
n.SetImplicit(true)
|
|
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-20 12:54:23 -08:00
|
|
|
// typed returns n after setting its type to typ.
|
|
|
|
|
func typed(typ *types.Type, n ir.Node) ir.Node {
|
|
|
|
|
n.SetType(typ)
|
|
|
|
|
n.SetTypecheck(1)
|
|
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
// Values
|
|
|
|
|
|
|
|
|
|
func Const(pos src.XPos, typ *types.Type, val constant.Value) ir.Node {
|
2021-01-20 12:54:23 -08:00
|
|
|
return typed(typ, ir.NewBasicLit(pos, val))
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func Nil(pos src.XPos, typ *types.Type) ir.Node {
|
2021-01-20 12:54:23 -08:00
|
|
|
return typed(typ, ir.NewNilExpr(pos))
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Expressions
|
|
|
|
|
|
2021-01-22 13:29:59 -08:00
|
|
|
func Addr(pos src.XPos, x ir.Node) *ir.AddrExpr {
|
2021-03-10 17:27:30 -08:00
|
|
|
n := typecheck.NodAddrAt(pos, x)
|
|
|
|
|
switch x.Op() {
|
|
|
|
|
case ir.OARRAYLIT, ir.OMAPLIT, ir.OSLICELIT, ir.OSTRUCTLIT:
|
|
|
|
|
n.SetOp(ir.OPTRLIT)
|
|
|
|
|
}
|
2021-01-22 13:29:59 -08:00
|
|
|
typed(types.NewPtr(x.Type()), n)
|
|
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
func Assert(pos src.XPos, x ir.Node, typ *types.Type) ir.Node {
|
2021-01-20 12:54:23 -08:00
|
|
|
return typed(typ, ir.NewTypeAssertExpr(pos, x, nil))
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
func Binary(pos src.XPos, op ir.Op, typ *types.Type, x, y ir.Node) ir.Node {
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
switch op {
|
|
|
|
|
case ir.OANDAND, ir.OOROR:
|
2021-01-20 12:54:23 -08:00
|
|
|
return typed(x.Type(), ir.NewLogicalExpr(pos, op, x, y))
|
|
|
|
|
case ir.OADD:
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
n := ir.NewBinaryExpr(pos, op, x, y)
|
|
|
|
|
if x.Type().HasTParam() || y.Type().HasTParam() {
|
|
|
|
|
// Delay transformAdd() if either arg has a type param,
|
|
|
|
|
// since it needs to know the exact types to decide whether
|
|
|
|
|
// to transform OADD to OADDSTR.
|
|
|
|
|
n.SetType(typ)
|
|
|
|
|
n.SetTypecheck(3)
|
|
|
|
|
return n
|
2021-01-20 12:54:23 -08:00
|
|
|
}
|
2021-03-24 14:50:02 -07:00
|
|
|
typed(typ, n)
|
|
|
|
|
return transformAdd(n)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
default:
|
2021-01-20 12:54:23 -08:00
|
|
|
return typed(x.Type(), ir.NewBinaryExpr(pos, op, x, y))
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-30 08:43:58 -08:00
|
|
|
func Call(pos src.XPos, typ *types.Type, fun ir.Node, args []ir.Node, dots bool) ir.Node {
|
cmd/compile: getting more built-ins to work with generics
For Builtin ops, we currently stay with using the old
typechecker to transform the call to a more specific expression
and possibly use more specific ops. However, for a bunch of the
ops, we delay calling the old typechecker if any of the args have
type params, for a variety of reasons.
In the near future, we will start creating separate functions that do
the same transformations as the old typechecker for calls, builtins,
indexing, comparisons, etc. These functions can then be called at noder
time for nodes with no type params, and at stenciling time for nodes
with type params.
Remove unnecessary calls to types1 typechecker for most kinds of
statements (still need it for SendStmt, AssignStmt, ReturnStmt, and
SelectStmt). In particular, we don't need it for RangeStmt, and this
avoids some complaints by the types1 typechecker on generic code.
Other small changes:
- Fix check on whether to delay calling types1-typechecker on type
conversions. Should check if HasTParam is true, rather than if the
type is directly a TYPEPARAM.
- Don't call types1-typechecker on an indexing operation if the left
operand has a typeparam in its type and is not obviously a TMAP,
TSLICE, or TARRAY. As above, we will eventually have to create a new
function that can do the required transformations (for complicated
cases) at noder time or stenciling time.
- Copy n.BuiltinOp in subster.node()
- The complex arithmetic example in absdiff.go now works.
- Added new tests double.go and append.go
- Added new example with a new() call in settable.go
Change-Id: I8f377afb6126cab1826bd3c2732aa8cdf1f7e0b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/301951
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-12 11:36:02 -08:00
|
|
|
n := ir.NewCallExpr(pos, ir.OCALL, fun, args)
|
|
|
|
|
n.IsDDD = dots
|
|
|
|
|
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
if fun.Op() == ir.OTYPE {
|
2021-01-30 08:43:58 -08:00
|
|
|
// Actually a type conversion, not a function call.
|
cmd/compile: getting more built-ins to work with generics
For Builtin ops, we currently stay with using the old
typechecker to transform the call to a more specific expression
and possibly use more specific ops. However, for a bunch of the
ops, we delay calling the old typechecker if any of the args have
type params, for a variety of reasons.
In the near future, we will start creating separate functions that do
the same transformations as the old typechecker for calls, builtins,
indexing, comparisons, etc. These functions can then be called at noder
time for nodes with no type params, and at stenciling time for nodes
with type params.
Remove unnecessary calls to types1 typechecker for most kinds of
statements (still need it for SendStmt, AssignStmt, ReturnStmt, and
SelectStmt). In particular, we don't need it for RangeStmt, and this
avoids some complaints by the types1 typechecker on generic code.
Other small changes:
- Fix check on whether to delay calling types1-typechecker on type
conversions. Should check if HasTParam is true, rather than if the
type is directly a TYPEPARAM.
- Don't call types1-typechecker on an indexing operation if the left
operand has a typeparam in its type and is not obviously a TMAP,
TSLICE, or TARRAY. As above, we will eventually have to create a new
function that can do the required transformations (for complicated
cases) at noder time or stenciling time.
- Copy n.BuiltinOp in subster.node()
- The complex arithmetic example in absdiff.go now works.
- Added new tests double.go and append.go
- Added new example with a new() call in settable.go
Change-Id: I8f377afb6126cab1826bd3c2732aa8cdf1f7e0b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/301951
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-12 11:36:02 -08:00
|
|
|
if fun.Type().HasTParam() || args[0].Type().HasTParam() {
|
2021-02-09 15:13:19 -08:00
|
|
|
// For type params, don't typecheck until we actually know
|
|
|
|
|
// the type.
|
|
|
|
|
return typed(typ, n)
|
|
|
|
|
}
|
2021-03-24 14:50:02 -07:00
|
|
|
typed(typ, n)
|
|
|
|
|
return transformConvCall(n)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if fun, ok := fun.(*ir.Name); ok && fun.BuiltinOp != 0 {
|
cmd/compile: getting more built-ins to work with generics
For Builtin ops, we currently stay with using the old
typechecker to transform the call to a more specific expression
and possibly use more specific ops. However, for a bunch of the
ops, we delay calling the old typechecker if any of the args have
type params, for a variety of reasons.
In the near future, we will start creating separate functions that do
the same transformations as the old typechecker for calls, builtins,
indexing, comparisons, etc. These functions can then be called at noder
time for nodes with no type params, and at stenciling time for nodes
with type params.
Remove unnecessary calls to types1 typechecker for most kinds of
statements (still need it for SendStmt, AssignStmt, ReturnStmt, and
SelectStmt). In particular, we don't need it for RangeStmt, and this
avoids some complaints by the types1 typechecker on generic code.
Other small changes:
- Fix check on whether to delay calling types1-typechecker on type
conversions. Should check if HasTParam is true, rather than if the
type is directly a TYPEPARAM.
- Don't call types1-typechecker on an indexing operation if the left
operand has a typeparam in its type and is not obviously a TMAP,
TSLICE, or TARRAY. As above, we will eventually have to create a new
function that can do the required transformations (for complicated
cases) at noder time or stenciling time.
- Copy n.BuiltinOp in subster.node()
- The complex arithmetic example in absdiff.go now works.
- Added new tests double.go and append.go
- Added new example with a new() call in settable.go
Change-Id: I8f377afb6126cab1826bd3c2732aa8cdf1f7e0b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/301951
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-12 11:36:02 -08:00
|
|
|
// For Builtin ops, we currently stay with using the old
|
|
|
|
|
// typechecker to transform the call to a more specific expression
|
|
|
|
|
// and possibly use more specific ops. However, for a bunch of the
|
|
|
|
|
// ops, we delay doing the old typechecker if any of the args have
|
|
|
|
|
// type params, for a variety of reasons:
|
|
|
|
|
//
|
|
|
|
|
// OMAKE: hard to choose specific ops OMAKESLICE, etc. until arg type is known
|
|
|
|
|
// OREAL/OIMAG: can't determine type float32/float64 until arg type know
|
|
|
|
|
// OLEN/OCAP: old typechecker will complain if arg is not obviously a slice/array.
|
|
|
|
|
// OAPPEND: old typechecker will complain if arg is not obviously slice, etc.
|
|
|
|
|
//
|
|
|
|
|
// We will eventually break out the transforming functionality
|
|
|
|
|
// needed for builtin's, and call it here or during stenciling, as
|
|
|
|
|
// appropriate.
|
|
|
|
|
switch fun.BuiltinOp {
|
|
|
|
|
case ir.OMAKE, ir.OREAL, ir.OIMAG, ir.OLEN, ir.OCAP, ir.OAPPEND:
|
|
|
|
|
hasTParam := false
|
|
|
|
|
for _, arg := range args {
|
|
|
|
|
if arg.Type().HasTParam() {
|
|
|
|
|
hasTParam = true
|
|
|
|
|
break
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if hasTParam {
|
|
|
|
|
return typed(typ, n)
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2021-03-24 14:50:02 -07:00
|
|
|
typed(typ, n)
|
|
|
|
|
return transformBuiltin(n)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
2021-01-19 13:54:33 -08:00
|
|
|
// Add information, now that we know that fun is actually being called.
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
switch fun := fun.(type) {
|
|
|
|
|
case *ir.ClosureExpr:
|
|
|
|
|
fun.Func.SetClosureCalled(true)
|
|
|
|
|
case *ir.SelectorExpr:
|
|
|
|
|
if fun.Op() == ir.OCALLPART {
|
|
|
|
|
op := ir.ODOTMETH
|
|
|
|
|
if fun.X.Type().IsInterface() {
|
|
|
|
|
op = ir.ODOTINTER
|
|
|
|
|
}
|
|
|
|
|
fun.SetOp(op)
|
2021-01-19 13:54:33 -08:00
|
|
|
// Set the type to include the receiver, since that's what
|
|
|
|
|
// later parts of the compiler expect
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
fun.SetType(fun.Selection.Type)
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2021-03-23 10:19:11 -07:00
|
|
|
n.Use = ir.CallUseExpr
|
|
|
|
|
if fun.Type().NumResults() == 0 {
|
|
|
|
|
n.Use = ir.CallUseStmt
|
|
|
|
|
}
|
|
|
|
|
|
2021-02-08 14:33:51 -08:00
|
|
|
if fun.Op() == ir.OXDOT {
|
2021-03-10 17:27:30 -08:00
|
|
|
if !fun.(*ir.SelectorExpr).X.Type().HasTParam() {
|
2021-02-08 14:33:51 -08:00
|
|
|
base.FatalfAt(pos, "Expecting type param receiver in %v", fun)
|
|
|
|
|
}
|
|
|
|
|
// For methods called in a generic function, don't do any extra
|
|
|
|
|
// transformations. We will do those later when we create the
|
|
|
|
|
// instantiated function and have the correct receiver type.
|
|
|
|
|
typed(typ, n)
|
|
|
|
|
return n
|
|
|
|
|
}
|
2021-02-02 13:04:16 -08:00
|
|
|
if fun.Op() != ir.OFUNCINST {
|
2021-03-23 10:19:11 -07:00
|
|
|
// If no type params, do the normal call transformations. This
|
|
|
|
|
// will convert OCALL to OCALLFUNC.
|
|
|
|
|
typed(typ, n)
|
2021-03-24 14:50:02 -07:00
|
|
|
transformCall(n)
|
2021-01-30 08:43:58 -08:00
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
2021-03-13 22:41:51 -08:00
|
|
|
// Leave the op as OCALL, which indicates the call still needs typechecking.
|
2021-01-30 08:43:58 -08:00
|
|
|
typed(typ, n)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func Compare(pos src.XPos, typ *types.Type, op ir.Op, x, y ir.Node) ir.Node {
|
2021-01-20 12:54:23 -08:00
|
|
|
n := ir.NewBinaryExpr(pos, op, x, y)
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
if x.Type().HasTParam() || y.Type().HasTParam() {
|
|
|
|
|
// Delay transformCompare() if either arg has a type param, since
|
|
|
|
|
// it needs to know the exact types to decide on any needed conversions.
|
|
|
|
|
n.SetType(typ)
|
|
|
|
|
n.SetTypecheck(3)
|
|
|
|
|
return n
|
2021-01-20 12:54:23 -08:00
|
|
|
}
|
2021-03-24 14:50:02 -07:00
|
|
|
typed(typ, n)
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
transformCompare(n)
|
2021-03-24 14:50:02 -07:00
|
|
|
return n
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
2021-01-22 13:29:59 -08:00
|
|
|
func Deref(pos src.XPos, x ir.Node) *ir.StarExpr {
|
|
|
|
|
n := ir.NewStarExpr(pos, x)
|
|
|
|
|
typed(x.Type().Elem(), n)
|
|
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func DotField(pos src.XPos, x ir.Node, index int) *ir.SelectorExpr {
|
|
|
|
|
op, typ := ir.ODOT, x.Type()
|
|
|
|
|
if typ.IsPtr() {
|
|
|
|
|
op, typ = ir.ODOTPTR, typ.Elem()
|
|
|
|
|
}
|
|
|
|
|
if !typ.IsStruct() {
|
|
|
|
|
base.FatalfAt(pos, "DotField of non-struct: %L", x)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// TODO(mdempsky): This is the backend's responsibility.
|
|
|
|
|
types.CalcSize(typ)
|
|
|
|
|
|
|
|
|
|
field := typ.Field(index)
|
|
|
|
|
return dot(pos, field.Type, op, x, field)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func DotMethod(pos src.XPos, x ir.Node, index int) *ir.SelectorExpr {
|
|
|
|
|
method := method(x.Type(), index)
|
|
|
|
|
|
|
|
|
|
// Method value.
|
|
|
|
|
typ := typecheck.NewMethodType(method.Type, nil)
|
|
|
|
|
return dot(pos, typ, ir.OCALLPART, x, method)
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-24 09:59:20 -08:00
|
|
|
// MethodExpr returns a OMETHEXPR node with the indicated index into the methods
|
|
|
|
|
// of typ. The receiver type is set from recv, which is different from typ if the
|
|
|
|
|
// method was accessed via embedded fields. Similarly, the X value of the
|
|
|
|
|
// ir.SelectorExpr is recv, the original OTYPE node before passing through the
|
|
|
|
|
// embedded fields.
|
|
|
|
|
func MethodExpr(pos src.XPos, recv ir.Node, embed *types.Type, index int) *ir.SelectorExpr {
|
|
|
|
|
method := method(embed, index)
|
|
|
|
|
typ := typecheck.NewMethodType(method.Type, recv.Type())
|
|
|
|
|
// The method expression T.m requires a wrapper when T
|
|
|
|
|
// is different from m's declared receiver type. We
|
|
|
|
|
// normally generate these wrappers while writing out
|
|
|
|
|
// runtime type descriptors, which is always done for
|
|
|
|
|
// types declared at package scope. However, we need
|
|
|
|
|
// to make sure to generate wrappers for anonymous
|
|
|
|
|
// receiver types too.
|
|
|
|
|
if recv.Sym() == nil {
|
|
|
|
|
typecheck.NeedRuntimeType(recv.Type())
|
|
|
|
|
}
|
|
|
|
|
return dot(pos, typ, ir.OMETHEXPR, recv, method)
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-22 13:29:59 -08:00
|
|
|
func dot(pos src.XPos, typ *types.Type, op ir.Op, x ir.Node, selection *types.Field) *ir.SelectorExpr {
|
|
|
|
|
n := ir.NewSelectorExpr(pos, op, x, selection.Sym)
|
|
|
|
|
n.Selection = selection
|
|
|
|
|
typed(typ, n)
|
|
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// TODO(mdempsky): Move to package types.
|
|
|
|
|
func method(typ *types.Type, index int) *types.Field {
|
|
|
|
|
if typ.IsInterface() {
|
|
|
|
|
return typ.Field(index)
|
|
|
|
|
}
|
|
|
|
|
return types.ReceiverBaseType(typ).Methods().Index(index)
|
|
|
|
|
}
|
|
|
|
|
|
cmd/compile: getting more built-ins to work with generics
For Builtin ops, we currently stay with using the old
typechecker to transform the call to a more specific expression
and possibly use more specific ops. However, for a bunch of the
ops, we delay calling the old typechecker if any of the args have
type params, for a variety of reasons.
In the near future, we will start creating separate functions that do
the same transformations as the old typechecker for calls, builtins,
indexing, comparisons, etc. These functions can then be called at noder
time for nodes with no type params, and at stenciling time for nodes
with type params.
Remove unnecessary calls to types1 typechecker for most kinds of
statements (still need it for SendStmt, AssignStmt, ReturnStmt, and
SelectStmt). In particular, we don't need it for RangeStmt, and this
avoids some complaints by the types1 typechecker on generic code.
Other small changes:
- Fix check on whether to delay calling types1-typechecker on type
conversions. Should check if HasTParam is true, rather than if the
type is directly a TYPEPARAM.
- Don't call types1-typechecker on an indexing operation if the left
operand has a typeparam in its type and is not obviously a TMAP,
TSLICE, or TARRAY. As above, we will eventually have to create a new
function that can do the required transformations (for complicated
cases) at noder time or stenciling time.
- Copy n.BuiltinOp in subster.node()
- The complex arithmetic example in absdiff.go now works.
- Added new tests double.go and append.go
- Added new example with a new() call in settable.go
Change-Id: I8f377afb6126cab1826bd3c2732aa8cdf1f7e0b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/301951
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-12 11:36:02 -08:00
|
|
|
func Index(pos src.XPos, typ *types.Type, x, index ir.Node) ir.Node {
|
|
|
|
|
n := ir.NewIndexExpr(pos, x, index)
|
2021-03-23 10:19:11 -07:00
|
|
|
if x.Type().HasTParam() {
|
|
|
|
|
// transformIndex needs to know exact type
|
|
|
|
|
n.SetType(typ)
|
|
|
|
|
n.SetTypecheck(3)
|
cmd/compile: getting more built-ins to work with generics
For Builtin ops, we currently stay with using the old
typechecker to transform the call to a more specific expression
and possibly use more specific ops. However, for a bunch of the
ops, we delay calling the old typechecker if any of the args have
type params, for a variety of reasons.
In the near future, we will start creating separate functions that do
the same transformations as the old typechecker for calls, builtins,
indexing, comparisons, etc. These functions can then be called at noder
time for nodes with no type params, and at stenciling time for nodes
with type params.
Remove unnecessary calls to types1 typechecker for most kinds of
statements (still need it for SendStmt, AssignStmt, ReturnStmt, and
SelectStmt). In particular, we don't need it for RangeStmt, and this
avoids some complaints by the types1 typechecker on generic code.
Other small changes:
- Fix check on whether to delay calling types1-typechecker on type
conversions. Should check if HasTParam is true, rather than if the
type is directly a TYPEPARAM.
- Don't call types1-typechecker on an indexing operation if the left
operand has a typeparam in its type and is not obviously a TMAP,
TSLICE, or TARRAY. As above, we will eventually have to create a new
function that can do the required transformations (for complicated
cases) at noder time or stenciling time.
- Copy n.BuiltinOp in subster.node()
- The complex arithmetic example in absdiff.go now works.
- Added new tests double.go and append.go
- Added new example with a new() call in settable.go
Change-Id: I8f377afb6126cab1826bd3c2732aa8cdf1f7e0b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/301951
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-12 11:36:02 -08:00
|
|
|
return n
|
|
|
|
|
}
|
2021-03-23 10:19:11 -07:00
|
|
|
typed(typ, n)
|
|
|
|
|
// transformIndex will modify n.Type() for OINDEXMAP.
|
|
|
|
|
transformIndex(n)
|
|
|
|
|
return n
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func Slice(pos src.XPos, typ *types.Type, x, low, high, max ir.Node) ir.Node {
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
op := ir.OSLICE
|
|
|
|
|
if max != nil {
|
|
|
|
|
op = ir.OSLICE3
|
|
|
|
|
}
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
n := ir.NewSliceExpr(pos, op, x, low, high, max)
|
|
|
|
|
if x.Type().HasTParam() {
|
|
|
|
|
// transformSlice needs to know if x.Type() is a string or an array or a slice.
|
|
|
|
|
n.SetType(typ)
|
|
|
|
|
n.SetTypecheck(3)
|
|
|
|
|
return n
|
|
|
|
|
}
|
2021-03-24 14:50:02 -07:00
|
|
|
typed(typ, n)
|
cmd/compile: replace calls to typecheck with transform functions
For additions, compares, and slices, create transform functions that do
just the transformations for those nodes by the typecheck package (given
that the code has been fully typechecked by types2). For nodes that have
no args with typeparams, we call these transform functions directly in
noder2. But for nodes that have args with typeparams, we have to delay
and call the tranform functions during stenciling, since we don't know
the specific types involved.
We indicate that a node still needs transformation by setting Typecheck
to a new value 3. This value means the current type of the node has been
set (via types2), but the node may still need transformation.
Had to export typcheck.IsCmp and typecheck.Assignop from the typecheck
package.
Added new tests list2.go (required delaying compare typecheck/transform
because of != compare in checkList) and adder.go (requires delaying add
typecheck/transform, since it can do addition for numbers or strings).
There are several more transformation functions needed for expressions
(indexing, calls, etc.) and several more complicated ones needed for
statements (mainly various kinds of assignments).
Change-Id: I7d89d13a4108308ea0304a4b815ab60b40c59b0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/303091
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2021-03-18 14:36:39 -07:00
|
|
|
transformSlice(n)
|
2021-03-24 14:50:02 -07:00
|
|
|
return n
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func Unary(pos src.XPos, op ir.Op, x ir.Node) ir.Node {
|
|
|
|
|
switch op {
|
|
|
|
|
case ir.OADDR:
|
2021-01-22 13:29:59 -08:00
|
|
|
return Addr(pos, x)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
case ir.ODEREF:
|
2021-01-22 13:29:59 -08:00
|
|
|
return Deref(pos, x)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
typ := x.Type()
|
|
|
|
|
if op == ir.ORECV {
|
|
|
|
|
typ = typ.Elem()
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
2021-01-22 13:29:59 -08:00
|
|
|
return typed(typ, ir.NewUnaryExpr(pos, op, x))
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Statements
|
|
|
|
|
|
|
|
|
|
var one = constant.MakeInt64(1)
|
|
|
|
|
|
2021-03-23 10:19:11 -07:00
|
|
|
func IncDec(pos src.XPos, op ir.Op, x ir.Node) *ir.AssignOpStmt {
|
|
|
|
|
assert(x.Type() != nil)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
return ir.NewAssignOpStmt(pos, op, x, typecheck.DefaultLit(ir.NewBasicLit(pos, one), x.Type()))
|
|
|
|
|
}
|