Not used any more.
Fixes#53860
Change-Id: Id0b1c3ed30b576d6c5f08f064d1262de337262b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/418374
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Performance is kind of hard to exactly quantify.
One big difference between jump tables and the old binary search
scheme is that there's only 1 branch statement instead of O(n) of
them. That can be both a blessing and a curse, and can make evaluating
jump tables very hard to do.
The single branch can become a choke point for the hardware branch
predictor. A branch table jump must fit all of its state in a single
branch predictor entry (technically, a branch target predictor entry).
With binary search that predictor state can be spread among lots of
entries. In cases where the case selection is repetitive and thus
predictable, binary search can perform better.
The big win for a jump table is that it doesn't consume so much of the
branch predictor's resources. But that benefit is essentially never
observed in microbenchmarks, because the branch predictor can easily
keep state for all the binary search branches in a microbenchmark. So
that benefit is really hard to measure.
So predictable switch microbenchmarks are ~useless - they will almost
always favor the binary search scheme. Fully unpredictable switch
microbenchmarks are better, as they aren't lying to us quite so
much. In a perfectly unpredictable situation, a jump table will expect
to incur 1-1/N branch mispredicts, where a binary search would incur
lg(N)/2 of them. That makes the crossover point at about N=4. But of
course switches in real programs are seldom fully unpredictable, so
we'll use a higher crossover point.
Beyond the branch predictor, jump tables tend to execute more
instructions per switch but have no additional instructions per case,
which also argues for a larger crossover.
As far as code size goes, with this CL cmd/go has a slightly smaller
code segment and a slightly larger overall size (from the jump tables
themselves which live in the data segment).
This is a case where some FDO (feedback-directed optimization) would
be really nice to have. #28262
Some large-program benchmarks might help make the case for this
CL. Especially if we can turn on branch mispredict counters so we can
see how much using jump tables can free up branch prediction resources
that can be gainfully used elsewhere in the program.
name old time/op new time/op delta
Switch8Predictable 1.89ns ± 2% 1.27ns ± 3% -32.58% (p=0.000 n=9+10)
Switch8Unpredictable 9.33ns ± 1% 7.50ns ± 1% -19.60% (p=0.000 n=10+9)
Switch32Predictable 2.20ns ± 2% 1.64ns ± 1% -25.39% (p=0.000 n=10+9)
Switch32Unpredictable 10.0ns ± 2% 7.6ns ± 2% -24.04% (p=0.000 n=10+10)
Fixes#5496
Update #34381
Change-Id: I3ff56011d02be53f605ca5fd3fb96b905517c34f
Reviewed-on: https://go-review.googlesource.com/c/go/+/357330
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Keith Randall <khr@google.com>
[This CL is part of a sequence implementing the proposal #51082.
The design doc is at https://go.dev/s/godocfmt-design.]
Run the updated gofmt, which reformats doc comments,
on the main repository. Vendored files are excluded.
For #51082.
Change-Id: I7332f099b60f716295fb34719c98c04eb1a85407
Reviewed-on: https://go-review.googlesource.com/c/go/+/384268
Reviewed-by: Jonathan Amsterdam <jba@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
For certain type of method wrappers we used to generate a tail
call. That was disabled in CL 307234 when register ABI is used,
because with the current IR it was difficult to generate a tail
call with the arguments in the right places. The problem was that
the IR does not contain a CALL-like node with arguments; instead,
it contains an OAS node that adjusts the receiver, than an
OTAILCALL node that just contains the target, but no argument
(with the assumption that the OAS node will put the adjusted
receiver in the right place). With register ABI, putting
arguments in registers are done in SSA. The assignment (OAS)
doesn't put the receiver in register.
This CL changes the IR of a tail call to take an actual OCALL
node. Specifically, a tail call is represented as
OTAILCALL (OCALL target args...)
This way, the call target and args are connected through the OCALL
node. So the call can be analyzed in SSA and the args can be passed
in the right places.
(Alternatively, we could have OTAILCALL node directly take the
target and the args, without the OCALL node. Using an OCALL node is
convenient as there are existing code that processes OCALL nodes
which do not need to be changed. Also, a tail call is similar to
ORETURN (OCALL target args...), except it doesn't preserve the
frame. I did the former but I'm open to change.)
The SSA representation is similar. Previously, the IR lowers to
a Store the receiver then a BlockRetJmp which jumps to the target
(without putting the arg in register). Now we use a TailCall op,
which takes the target and the args. The call expansion pass and
the register allocator handles TailCall pretty much like a
StaticCall, and it will do the right ABI analysis and put the args
in the right places. (Args other than the receiver are already in
the right places. For register args it generates no code for them.
For stack args currently it generates a self copy. I'll work on
optimize that out.) BlockRetJmp is still used, signaling it is a
tail call. The actual call is made in the TailCall op so
BlockRetJmp generates no code (we could use BlockExit if we like).
This slightly reduces binary size:
old new
cmd/go 14003088 13953936
cmd/link 6275552 6271456
Change-Id: I2d16d8d419fe1f17554916d317427383e17e27f0
Reviewed-on: https://go-review.googlesource.com/c/go/+/350145
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: David Chase <drchase@google.com>
CL 330671 move rewriting method call to method expression to escape
analysis. This CL move the rewriting up further, into typecheck. It
helps simplify the code for dowstream passes, as they now only have to
deal with OCALLFUNC.
There're two notes:
- For -G=3, we can't rewrite d.M() where d is an instantiated receiver
in transformCall, but let irgen.stencil to rewrite it.
- Escape analysis still have to check for rewriting method calls, as
the devirtualization pass can still generate OCALLMETH.
Does not pass toolstash, since when the export data now contains method
expression calls instead of method calls.
Change-Id: I77465ef04d50dc4efedddca7eb55b3fc9483db0e
Reviewed-on: https://go-review.googlesource.com/c/go/+/330831
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>
This CL moves two bits of related code from order.go to escape
analysis:
1. The recognition of "unsafe uintptr" arguments passed to
syscall-like functions.
2. The wrapping of go/defer function calls in parameter-free function
literals.
As with previous CLs, it would be nice to push this logic even further
forward, but for now escape analysis seems most pragmatic.
A couple side benefits:
1. It allows getting rid of the uintptrEscapesHack kludge.
2. When inserting wrappers, we can move some expressions into the
wrapper and escape analyze them better. For example, the test
expectation changes are all due to slice literals in go/defer calls
where the slice is now constructed at the call site, and can now be
stack allocated.
Change-Id: I73679bcad7fa8d61d2fc52d4cea0dc5ff0de8c0c
Reviewed-on: https://go-review.googlesource.com/c/go/+/330330
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Currently ORECOVER is a single operation that both (1) calculates
the (logical) caller frame pointer and (2) calls runtime.gorecover.
This is normally fine, but it's inconvenient for regabi, which wants
to wrap "defer recover()" into "defer func() { recover() }" and
needs (1) and (2) to happen at different times.
The current solution is to apply walkRecover early to split it into
the two steps, but calling it during order is a minor layering
violation. It works well today because the order and walk phases are
closely related anyway and walkRecover is relatively simple, but it
won't work for go/defer wrapping earlier into the frontend.
This CL adds a new, lower-level ORECOVERFP primitive, which represents
just part (2); and OGETCALLER{PC,SP} primitives, which provide a way
to compute (1) in the frontend too.
OGETCALLERPC isn't needed/used today, but it seems worth including for
completeness. Maybe it will be useful at some point for intrinsifying
runtime.getcaller{pc,sp}, like we already do for runtime.getg.
Change-Id: Iaa8ae51e09306c45c147b6759a5b7c24dcc317ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/330192
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
This flag is only needed to prevent the directClosureCall optimization
in walkCall, when called for walkGoDefer. But walkGoDefer don't need
to call walkCall: at this point in the compile, the call expression
isn't a real call anymore.
Instead, we just need to walkExpr on the function expression.
Change-Id: I8a5176cfe1bff53700cbd21ed1b479ebd9a839ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/330271
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Matthew Dempsky <mdempsky@google.com>
order already takes care of wrapping all go/defer function calls, so
there's no need for walk to duplicate that logic: it's never going to
be used.
Change-Id: I54e545404e52ab8f9d60151d1bd2aff4b9bd8b72
Reviewed-on: https://go-review.googlesource.com/c/go/+/330270
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Following CL 309029, this CL does the same thing for map
functions (mapaccess, mapassign, mapdelete).
For simplicity, always wrap "defer delete(m, k)". With
regabidefers enabled, this call is wrapped in a closure and the
rewriting happens automatically. Without regabidefers, it may not
be wrapped for certain key types, and then we'd need special
handling of the delete (because earlier the order pass does not
handle it). And we will turn on defer wrapping by default anyway.
Change-Id: I30663b1aa8e1d6f98e1fb81bf8c0c0ce607ab80b
Reviewed-on: https://go-review.googlesource.com/c/go/+/309510
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>
Adds code to the compiler's "order" phase to rewrite go and defer
statements to always be argument-less. E.g.
defer f(x,y) => x1, y1 := x, y
defer func() { f(x1, y1) }
This transformation is not beneficial on its own, but it helps
simplify runtime defer handling for the new register ABI (when
invoking deferred functions on the panic path, the runtime doesn't
need to manage the complexity of determining which args to pass in
register vs memory).
This feature is currently enabled by default if GOEXPERIMENT=regabi or
GOEXPERIMENT=regabidefer is in effect.
Included in this CL are some workarounds in the runtime to insure that
"go" statement targets in the runtime are argument-less already (since
wrapping them can potentially introduce heap-allocated closures, which
are currently not allowed). The expectation is that these workarounds
will be temporary, and can go away once we either A) change the rules
about heap-allocated closures, or B) implement some other scheme for
handling go statements.
Change-Id: I01060d79a6b140c6f0838d6e6813f807ccdca319
Reviewed-on: https://go-review.googlesource.com/c/go/+/298669
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
The explanatory comment and the associated version counter variable
for the helper routine "wrapCall" seem to have been left behind in
walk.go during the big refactoring -- move it back to where it should
be, next to wrapCall in stmt.go. Also fix a small buglet in the
comment itself.
Change-Id: I8637a838214b216581be59e01149a72282a46526
Reviewed-on: https://go-review.googlesource.com/c/go/+/303729
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
In CL 253457, we did the same fix for direct function calls. But for
method calls, the receiver argument also need to be passed through the
wrapper function, which we are not doing so the compiler crashes with
the code in #44415.
It will be nicer if we can rewrite OCALLMETHOD to normal OCALLFUNC, but
that will be for future CL. The passing receiver argument to wrapper
function is easier for backporting to go1.16 branch.
Fixes#44415
Change-Id: I03607a64429042c6066ce673931db9769deb3124
Reviewed-on: https://go-review.googlesource.com/c/go/+/296490
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This reverts commit CL 294849.
Reason for revert: this doesn't actually fix the issue, as revealed
by the noopt builder's failures.
Change-Id: Ib4ea9ceb4d75e46b3b91ec348b365fd8c83316ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/296629
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
In CL 253457, we did the same fix for direct function calls. But for
method calls, the receiver argument also need to be passed through the
wrapper function, which we are not doing so the compiler crashes with
the code in #44415.
As we already rewrite t.M(...) into T.M(t, ...) during walkCall1, to fix
this, we can do the same trick in wrapCall, so the receiver argument
will be treated as others.
Fixes#44415
Change-Id: I396182983c85d9c5e4494657da79d25636e8a079
Reviewed-on: https://go-review.googlesource.com/c/go/+/294849
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This CL splits out ORETJMP as a new TailCallStmt node, separate from
the other BranchStmt nodes. In doing so, this allows us to change it
from identifying a function by *types.Sym to identifying one by
directly pointing to the *ir.Func.
While here, also rename the operation to OTAILCALL.
Passes toolstash -cmp.
Change-Id: I273e6ea5d92bf3005ae02fb59b3240a190a6cf1b
Reviewed-on: https://go-review.googlesource.com/c/go/+/284227
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>
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>
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 $pkgs)"' {
var n *Name
var c Class
n.Class() -> n.Class_
n.SetClass(c) -> n.Class_ = c
}
rm Name.Class
rm Name.SetClass
mv Name.Class_ Name.Class
'
Change-Id: Ifb304bf4691a8c455456aabd8aa77178d4a49500
Reviewed-on: https://go-review.googlesource.com/c/go/+/281294
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Just "=". It's cleaner.
Passes toolstash -cmp.
[git-generate]
cd src/cmd/compile/internal/ir
pkgs=$(go list . ../...)
rf '
ex '"$(echo $pkgs)"' {
var l Nodes
var p *Nodes
p.Set(l) -> *p = l
}
ex '"$(echo $pkgs)"' {
var n InitNode
var l Nodes
*n.PtrInit() = l -> n.SetInit(l)
}
rm Nodes.Set
'
Change-Id: Ic97219792243667146a02776553942ae1189ff7d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281002
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>
This CL separates out PtrInit and SetInit into a new InitNode
extension interface, and adds a new TakeInit helper function for
taking and clearing the Init list (if any) from a Node.
This allows removing miniNode.SetInit and miniNode.PtrInit, which in
turn allow getting rid of immutableEmptyNodes, and will allow
simplification of the Nodes API.
It would be nice to get rid of the default Init method too, but
there's way more code that expects to be able to call that at the
moment, so that'll have to wait.
Passes toolstash -cmp.
Change-Id: Ia8c18fab9555b774376f7f43eeecfde4f07b5946
Reviewed-on: https://go-review.googlesource.com/c/go/+/281001
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>
Back to pre Russquake, Node.Nbody of OCALL* node is used to attach
variables which must be kept alive during that call.
Now after Russquake, we have CallExpr to represent a function call,
so use a dedicated field for those variables instead.
Passes toolstash -cmp.
Change-Id: I4f40ebefcc7c41cdcc4e29c7a6d8496a083b68f4
Reviewed-on: https://go-review.googlesource.com/c/go/+/280733
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>
