Age | Commit message (Collapse) | Author |
|
Similar to the previous CL to suppress escape analysis diagnostics for
method wrappers, suppress liveness analysis diagnostics too. It's
hardly useful to know that all of a wrapper method's arguments are
live at entry.
Change-Id: I0d1e44552c6334ee3b454adc107430232abcb56a
Reviewed-on: https://go-review.googlesource.com/c/go/+/330749
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
|
|
Currently, for stack objects, the compiler emits metadata that
includes the offset and type descriptor for each object. The type
descriptor symbol has many fields, and it references many other
symbols, e.g. field/element types, equality functions, names.
Observe that what we actually need at runtime is only the GC
metadata that are needed to scan the object, and the GC metadata
are "leaf" symbols (which doesn't reference other symbols). Emit
only the GC data instead. This avoids bringing live the type
descriptor as well as things referenced by it (if it is not
otherwise live).
This reduces binary sizes:
old new
hello (println) 1187776 1133856 (-4.5%)
hello (fmt) 1902448 1844416 (-3.1%)
cmd/compile 22670432 22438576 (-1.0%)
cmd/link 6346272 6225408 (-1.9%)
No significant change in compiler speed.
name old time/op new time/op delta
Template 184ms ± 2% 186ms ± 5% ~ (p=0.905 n=9+10)
Unicode 78.4ms ± 5% 76.3ms ± 3% -2.60% (p=0.009 n=10+10)
GoTypes 1.09s ± 1% 1.08s ± 1% -0.73% (p=0.027 n=10+8)
Compiler 85.6ms ± 3% 84.6ms ± 4% ~ (p=0.143 n=10+10)
SSA 7.23s ± 1% 7.25s ± 1% ~ (p=0.780 n=10+9)
Flate 116ms ± 5% 115ms ± 6% ~ (p=0.912 n=10+10)
GoParser 201ms ± 4% 195ms ± 1% ~ (p=0.089 n=10+10)
Reflect 455ms ± 1% 458ms ± 2% ~ (p=0.050 n=9+9)
Tar 155ms ± 2% 155ms ± 3% ~ (p=0.436 n=10+10)
XML 202ms ± 2% 200ms ± 2% ~ (p=0.053 n=10+9)
Change-Id: I33a7f383d79afba1a482cac6da0cf5b7de9c0ec4
Reviewed-on: https://go-review.googlesource.com/c/go/+/313514
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
callMethod's ABI wrappers
reflect.callReflect and reflect.callMethod are called from special
functions makeFuncStub and methodValueCall. The runtime expects
that it can find the first argument (ctxt) at 0(SP) in
makeFuncStub and methodValueCall's frame. Normally callReflect and
callMethod already do not modify the argument, and keep it alive.
But the compiler-generated ABI wrappers don't do that. Special
case the wrappers to not clobber its arguments.
Change-Id: I1769f49b81c38eabe452d561001c418352814d86
Reviewed-on: https://go-review.googlesource.com/c/go/+/310889
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
There are stack slots that are kept live for defers, which are
tracked separately. Don't clobber them.
Change-Id: Ib558345758b5a4fd89c7ff8a3fe08087059add21
Reviewed-on: https://go-review.googlesource.com/c/go/+/310329
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
In CL 307909 we generate code that spills pointer-typed argument
registers if it is part of an SSA-able aggregate. The current
code spill the register unconditionally. Sometimes it is
unnecessary, because it is already spilled, or it is never live.
This CL reworks the spill generation. We move it to the end of
compilation, after liveness analysis, so we have information about
if a spill is necessary, and only generate spills for the
necessary ones.
Change-Id: I8d60be9b2c47651aeda14f5e2d1bbd207c134b26
Reviewed-on: https://go-review.googlesource.com/c/go/+/309331
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
function
This fixes the build crash for
GOEXPERIMENT=regabi,regabiargs GOOS=windows go build syscall
Updates #40724.
Change-Id: I4400f6ff2e83e7e7e93ad5e58c6063b327532504
Reviewed-on: https://go-review.googlesource.com/c/go/+/309110
Trust: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
replaced old type-based logic with new abi-based logic;
earlier versions of this CL compared them for equality.
For not-in-a-register, they match everywhere tested.
also modified GetFrameOffset to make it more like the one it replaces;
the LocalsOffset is subtracted.
Change-Id: I65ce7f0646c493c277df6b6f46e4839a0d886ac9
Reviewed-on: https://go-review.googlesource.com/c/go/+/302072
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This CL resurrects the clobberdead debugging mode (CL 23924).
When -clobberdead flag is set (TODO: make it GOEXPERIMENT?), the
compiler inserts code that clobbers all dead stack slots that
contains pointers.
Mark windows syscall functions cgo_unsafe_args, as the code
actually does that, by taking the address of one argument and
passing it to cgocall.
Change-Id: Ie09a015f4bd14ae6053cc707866e30ae509b9d6f
Reviewed-on: https://go-review.googlesource.com/c/go/+/301791
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Repair of CL 300749.
ALSO:
found evidence that stack maps for bodyless methods are wrong.
gofmt in test/abi
removed never-executed code in types/size.go
Updates #44816.
Updates #40724.
Change-Id: Ifeb5fee60f60e7c7b58ee0457f58a3265d6cf3f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/302071
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This reverts commit 8ed438c077d82c4b4662c327dbbdb3c64e7547ca, CL 300749.
Reason for revert: Looks like it crashes on link-register architectures
Change-Id: I0c261df58900008cada3359889d2a87508158447
Reviewed-on: https://go-review.googlesource.com/c/go/+/302053
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
ALSO:
found evidence that stack maps for bodyless methods are wrong.
gofmt in test/abi
removed never-executed code in types/size.go
Updates #44816.
Change-Id: I658c33f049337fb6f1e625f0c55430d25bfa877e
Reviewed-on: https://go-review.googlesource.com/c/go/+/300749
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Verified with test and with single step watching changes to register
values across morestack calls, after reload.
Also added stack-growth test with pointer parameters of varying lifetime.
For #40724.
Change-Id: Idb5fe27786ac5c6665a734d41e68d3d39de2f4da
Reviewed-on: https://go-review.googlesource.com/c/go/+/294429
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
|
|
Calculating and emitting stack objects are essentially part of
liveness analysis, so move the code from ssagen to liveness. Allows
unexporting liveness.ShouldTrack.
Passes toolstash -cmp.
Change-Id: I88b5b2e75b8dfb46b8b03a2fa09a9236865cbf3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/284413
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
recent renames
Went in a semi-automated way through the clearest renames of functions,
and updated comments and error messages where it made sense.
Change-Id: Ied8e152b562b705da7f52f715991a77dab60da35
Reviewed-on: https://go-review.googlesource.com/c/go/+/284216
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
This CL moves almost all PAUTOHEAP handling code to SSA construction.
Instead of changing Names to PAUTOHEAP, escape analysis now only sets
n.Esc() to ir.EscHeap, and SSA handles creating the "&x"
pseudo-variables and associating them via Heapaddr.
This CL also gets rid of n.Stackcopy, which was used to distinguish
the heap copy of a parameter used within a function from the stack
copy used in the function calling convention. In practice, this is
always obvious from context: liveness and function prologue/epilogue
want to know about the stack copies, and everywhere else wants the
heap copy.
Hopefully moving all parameter/result handling into SSA helps with
making the register ABI stuff easier.
Also, the only remaining uses of PAUTOHEAP are now for closure
variables, so I intend to rename it to PCLOSUREVAR or get rid of those
altogether too. But this CL is already big and scary enough.
Change-Id: Ief5ef6205041b9d0ee445314310c0c5a98187e77
Reviewed-on: https://go-review.googlesource.com/c/go/+/283233
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: David Chase <drchase@google.com>
|
|
The compiler currently has two modes for compilation: one where it
compiles each function as it sees them, and another where it enqueues
them all into a work queue. A subsequent CL is going to reorder
function compilation to ensure that functions are always compiled
before any non-trivial function literals they enclose, and this will
be easier if we always use the compile work queue.
Also, fewer compilation modes makes things simpler to reason about.
Change-Id: Ie090e81f7476c49486296f2b90911fa0a466a5dd
Reviewed-on: https://go-review.googlesource.com/c/go/+/283313
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
These aren't part of the Node interface anymore, so no need to keep
them around.
Passes toolstash -cmp.
[git-generate]
cd src/cmd/compile/internal/ir
: Fix one off case that causes trouble for rf.
sed -i -e 's/n.SetClass(ir.PAUTO)/n.Class_ = ir.PAUTO/' ../ssa/export_test.go
pkgs=$(go list . ../...)
rf '
ex '"$(echo $pkgs)"' {
var n *Name
var c Class
n.Class() -> n.Class_
n.SetClass(c) -> n.Class_ = c
}
rm Name.Class
rm Name.SetClass
mv Name.Class_ Name.Class
'
Change-Id: Ifb304bf4691a8c455456aabd8aa77178d4a49500
Reviewed-on: https://go-review.googlesource.com/c/go/+/281294
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
|
|
[git-generate]
cd src/cmd/compile/internal/ir
pkgs=$(grep -l -w Name ../*/*.go | xargs dirname | sort -u | grep -v '/ir$')
rf '
ex . '"$(echo $pkgs)"' {
var n *Name
n.Name() -> n
}
'
Change-Id: I6bfce6417a6dba833d2f652ae212a32c11bc5ef6
Reviewed-on: https://go-review.googlesource.com/c/go/+/280972
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
This CL shuffles a couple functions around to help flatten the package
dependency graph somewhat:
1. ssa.LosesStmtMark is only ever used in associated with an
objw.Prog, so we might as well move it to that package. This removes a
dependency from objw (a relatively low-level utility package that
wraps cmd/internal/obj) on ssa (a large and relatively high-level
package).
2. Moves liveness.SetTypeBits into a new package typebits. A
single-function package is a bit on the silly side, but reflectdata
shouldn't need to depend on liveness (nor vice versa).
[git-generate]
cd src/cmd/compile/internal/ssa
rf '
mv LosesStmtMark prog.go
mv prog.go cmd/compile/internal/objw
'
cd ../liveness
rf '
mv SetTypeBits Set
mv Set typebits.go
rm typebits.go:/Copyright/+4,/^package/-0
mv typebits.go cmd/compile/internal/typebits
'
Change-Id: Ic9a983f0ad6c0cf1a537f99889699a8444699e6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/280447
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
|
|
All function parameters and return values in liveness have explicit
*ir.Name type, so use it directly instead of casting from ir.Node. While
at it, rename "affectedNode" to "affectedVar" to reflect this change.
Passes buildall w/ toolstash -cmp.
Change-Id: Id927e817a92ddb551a029064a2a54e020ca27074
Reviewed-on: https://go-review.googlesource.com/c/go/+/279434
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# AutoVar is essentially an ssa helper; move it there.
mv AutoVar value.go
mv value.go cmd/compile/internal/ssa
# Export liveness API and unexport non-API.
mv LivenessMap Map
mv Map.vals Map.Vals
mv Map.deferreturn Map.DeferReturn
mv livenessShouldTrack ShouldTrack
mv onebitwalktype1 SetTypeBits
mv allUnsafe IsUnsafe
mv liveness Compute
mv BlockEffects blockEffects
mv Liveness liveness
mv liveness _liveness # make room for import
mv emitptrargsmap WriteFuncMap
mv WriteFuncMap plive.go
mv bvset.go plive.go cmd/compile/internal/liveness
'
cd ../liveness
rf '
mv _liveness liveness
'
Change-Id: I3b86e5025bd9d32a7e19f44714fa16be4125059e
Reviewed-on: https://go-review.googlesource.com/c/go/+/279311
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|