This reverts commit 080187f4f7.
It broke build of golang.org/x/exp/shiny/iconvg
See issue 24395 for details
Change-Id: Ifd6134f6214e6cee40bd3c63c32941d5fc96ae8b
Reviewed-on: https://go-review.googlesource.com/100755
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This change ports all the remaining tests checking that small memmoves
are replaced with MOVs to the new codegen test harness, and deletes
them from the asm_test file.
Change-Id: I01c94b441e27a5d61518035af62d62779dafeb56
Reviewed-on: https://go-review.googlesource.com/100476
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Add tests for the "negative size argument in make.*" and "size argument
too large in make.*" error messages to appear at call sites in case the
size is a const defined on another line.
As suggested by Matthew in a comment on CL 69910.
Change-Id: I5c33d4bec4e3d20bb21fe8019df27999997ddff3
Reviewed-on: https://go-review.googlesource.com/100395
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
As a side effect of working on mid-stack inlining, we've fixed support
for inlining variadic functions. Might as well enable it.
Change-Id: I7f555f8b941969791db7eb598c0b49f6dc0820aa
Reviewed-on: https://go-review.googlesource.com/100456
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Currently, order desugars map assignment operations like
m[k] op= r
into
m[k] = m[k] op r
which in turn is transformed during walk into:
tmp := *mapaccess(m, k)
tmp = tmp op r
*mapassign(m, k) = tmp
However, this is suboptimal, as we could instead produce just:
*mapassign(m, k) op= r
One complication though is if "r == 0", then "m[k] /= r" and "m[k] %=
r" will panic, and they need to do so *before* calling mapassign,
otherwise we may insert a new zero-value element into the map.
It would be spec compliant to just emit the "r != 0" check before
calling mapassign (see #23735), but currently these checks aren't
generated until SSA construction. For now, it's simpler to continue
desugaring /= and %= into two map indexing operations.
Fixes#23661.
Change-Id: I46e3739d9adef10e92b46fdd78b88d5aabe68952
Reviewed-on: https://go-review.googlesource.com/91557
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
cmd/asm now supports three-operand form of IMUL,
so instead of using IMUL with resultInArg0, emit IMUL3 instruction.
This results in less redundant MOVs where SSA assigns
different registers to input[0] and dst arguments.
Note: these have exactly the same encoding when reg0=reg1:
IMUL3x $const, reg0, reg1
IMULx $const, reg
Two-operand IMULx is like a crippled IMUL3x, with dst fixed to input[0].
This is why we don't bother to generate IMULx for the case where
dst is the same as input[0].
Change-Id: I4becda475b3dffdd07b6fdf1c75bacc82af654e4
Reviewed-on: https://go-review.googlesource.com/99656
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Keith Randall <khr@golang.org>
Remove old tests from asm_test.
Change-Id: Ib408ec7faa60068bddecf709b93ce308e0ef665a
Reviewed-on: https://go-review.googlesource.com/100075
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
This change adds a README file inside the test/codegen directory,
explaining how to run the codegen tests and the syntax of the regexps
comments used to match assembly instructions.
Change-Id: Ica4eb3ffa9c6975371538cc8ae0ac3c1a3a03baf
Reviewed-on: https://go-review.googlesource.com/99156
Reviewed-by: Keith Randall <khr@golang.org>
And delete them from asm_go.
Change-Id: I0057cbd90ca55fa51c596e32406e190f3866f93e
Reviewed-on: https://go-review.googlesource.com/99815
Reviewed-by: Keith Randall <khr@golang.org>
Only RotateLeft{64,32} were tested, and just for ppc64. This CL adds
tests for RotateLeft{64,32,16,8} on arm64 and amd64/386, for the cases
where the calls are actually instrinsified.
RotateLeft tests (the last ones for math/bits functions) are deleted
from asm_test.
This CL also adds a space between the "//" and the arch name in the
comments, to uniform this file to the style used in all the other
files.
Change-Id: Ifc2a27261d70bcc294b4ec64490d8367f62d2b89
Reviewed-on: https://go-review.googlesource.com/99596
Reviewed-by: Giovanni Bajo <rasky@develer.com>
This adds four new deductions to the prove pass, all related to adding
or subtracting one from a value. This is the first hint of actual
arithmetic relations in the prove pass.
The most effective of these is
x-1 >= w && x > min ⇒ x > w
This helps eliminate bounds checks in code like
if x > 0 {
// do something with s[x-1]
}
Altogether, these deductions prove an additional 260 branches in std
and cmd. Furthermore, they will let us eliminate some tricky
compiler-inserted panics in the runtime that are interfering with
static analysis.
Fixes#23354.
Change-Id: I7088223e0e0cd6ff062a75c127eb4bb60e6dce02
Reviewed-on: https://go-review.googlesource.com/87480
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro>
This adds a few simple deductions to the prove pass' fact table to
derive unsigned concrete limits from signed concrete limits where
possible.
This tweak lets the pass prove 70 additional branch conditions in std
and cmd.
This is based on a comment from the recently-deleted factsTable.get:
"// TODO: also use signed data if lim.min >= 0".
Change-Id: Ib4340249e7733070f004a0aa31254adf5df8a392
Reviewed-on: https://go-review.googlesource.com/87479
Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro>
Reviewed-by: Keith Randall <khr@golang.org>
Currently the prove pass uses implication queries. For each block, it
collects the set of branch conditions leading to that block, and
queries this fact table for whether any of these facts imply the
block's own branch condition (or its inverse). This works remarkably
well considering it doesn't do any deduction on these facts, but it
has various downsides:
1. It requires an implementation both of adding facts to the table and
determining implications. These are very nearly duals of each
other, but require separate implementations. Likewise, the process
of asserting facts of dominating branch conditions is very nearly
the dual of the process of querying implied branch conditions.
2. It leads to less effective use of derived facts. For example, the
prove pass currently derives facts about the relations between len
and cap, but can't make use of these unless a branch condition is
in the exact form of a derived fact. If one of these derived facts
contradicts another fact, it won't notice or make use of this.
This CL changes the approach of the prove pass to instead use
*contradiction* instead of implication. Rather than ever querying a
branch condition, it simply adds branch conditions to the fact table.
If this leads to a contradiction (specifically, it makes the fact set
unsatisfiable), that branch is impossible and can be cut. As a result,
1. We can eliminate the code for determining implications
(factsTable.get disappears entirely). Also, there is now a single
implementation of visiting and asserting branch conditions, since
we don't have to flip them around to treat them as facts in one
place and queries in another.
2. Derived facts can be used effectively. It doesn't matter *why* the
fact table is unsatisfiable; a contradiction in any of the facts is
enough.
3. As an added benefit, it's now quite easy to avoid traversing beyond
provably-unreachable blocks. In contrast, the current
implementation always visits all blocks.
The prove pass already has nearly all of the mechanism necessary to
compute unsatisfiability, which means this both simplifies the code
and makes it more powerful.
The only complication is that the current implication procedure has a
hack for dealing with the 0 <= Args[0] condition of OpIsInBounds and
OpIsSliceInBounds. We replace this with asserting the appropriate fact
when we process one of these conditions. This seems much cleaner
anyway, and works because we can now take advantage of derived facts.
This has no measurable effect on compiler performance.
Effectiveness:
There is exactly one condition in all of std and cmd that this fails
to prove that the old implementation could: (int64(^uint(0)>>1) < x)
in encoding/gob. This can never be true because x is an int, and it's
basically coincidence that the old code gets this. (For example, it
fails to prove the similar (x < ^int64(^uint(0)>>1)) condition that
immediately precedes it, and even though the conditions are logically
unrelated, it wouldn't get the second one if it hadn't first processed
the first!)
It does, however, prove a few dozen additional branches. These come
from facts that are added to the fact table about the relations
between len and cap. These were almost never queried directly before,
but could lead to contradictions, which the unsat-based approach is
able to use.
There are exactly two branches in std and cmd that this implementation
proves in the *other* direction. This sounds scary, but is okay
because both occur in already-unreachable blocks, so it doesn't matter
what we chose. Because the fact table logic is sound but incomplete,
it fails to prove that the block isn't reachable, even though it is
able to prove that both outgoing branches are impossible. We could
turn these blocks into BlockExit blocks, but it doesn't seem worth the
trouble of the extra proof effort for something that happens twice in
all of std and cmd.
Tests:
This CL updates test/prove.go to change the expected messages because
it can no longer give a "reason" why it proved or disproved a
condition. It also adds a new test of a branch it couldn't prove
before.
It mostly guts test/sliceopt.go, removing everything related to slice
bounds optimizations and moving a few relevant tests to test/prove.go.
Much of this test is actually unreachable. The new prove pass figures
this out and doesn't try to prove anything about the unreachable
parts. The output on the unreachable parts is already suspect because
anything can be proved at that point, so it's really just a regression
test for an algorithm the compiler no longer uses.
This is a step toward fixing #23354. That issue is quite easy to fix
once we can use derived facts effectively.
Change-Id: Ia48a1b9ee081310579fe474e4a61857424ff8ce8
Reviewed-on: https://go-review.googlesource.com/87478
Reviewed-by: Keith Randall <khr@golang.org>
And delete them from the asm_test.go file.
Change-Id: I124c8c352299646ec7db0968cdb0fe59a3b5d83d
Reviewed-on: https://go-review.googlesource.com/99475
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
This was already done for normal parameters, and the same logic
applies for receiver parameters too.
Updates #24305.
Change-Id: Ia2a46f68d14e8fb62004ff0da1db0f065a95a1b7
Reviewed-on: https://go-review.googlesource.com/99335
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This recently added arm64 memmove codegen check:
func movesmall() {
// arm64:-"memmove"
x := [...]byte{1, 2, 3, 4, 5, 6, 7}
copy(x[1:], x[:])
}
is not correct, for two reasons:
1. regexps are matched from the start of the disasm line (excluding
line information). This mean that a negative -"memmove" check will
pass against a 'CALL runtime.memmove' line because the line does
not start with 'memmove' (its starts with CALL...).
The way to specify no 'memmove' match whatsoever on the line is
-".*memmove"
2. AFAIK comments on their own line are matched against the first
subsequent non-comment line. So the code above only verifies that
the x := ... line does not generate a memmove. The comment should
be moved near the copy() line, if it's that one we want to not
generate a memmove call.
The fact that the test above is not effective can be checked by
running `go run run.go -v codegen` in the toplevel test directory with
a go1.10 toolchain (that does not have the memmove-elision
optimization). The test will still pass (it shouldn't).
This change changes the regexp to -".*memmove" and moves it near the
line it needs to (not)match.
Change-Id: Ie01ef4d775e77d92dc8d8b7856b89b200f5e5ef2
Reviewed-on: https://go-review.googlesource.com/98977
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
And remove them from ssa_test.
Change-Id: If767af662801219774d1bdb787c77edfa6067770
Reviewed-on: https://go-review.googlesource.com/98976
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Current implementation doesn't consider MOVDreg type operand and fail to combine
it into larger store. This patch fixes the issue.
Fixes#24242
Change-Id: I7d68697f80e76f48c3528ece01a602bf513248ec
Reviewed-on: https://go-review.googlesource.com/98397
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
And remove them from ssa_test.
Change-Id: I3efac5fea529bb0efa2dae32124530482ba5058e
Reviewed-on: https://go-review.googlesource.com/98815
Reviewed-by: Keith Randall <khr@golang.org>
And remove them from ssa_test.
Change-Id: Ib5de5c0d908f23915e0847eca338cacf2fa5325b
Reviewed-on: https://go-review.googlesource.com/98795
Reviewed-by: Giovanni Bajo <rasky@develer.com>
This change move bits.Len* intrinsification tests to the new codegen
test harness, removing them from the old ssa_test file. Five different
test functions (one for each bit.Len function tested) was used, to
avoid possible unwanted interactions between multiple calls inside one
function.
Change-Id: Iffd5be55b58e88597fa30a562a28dacb01236d8b
Reviewed-on: https://go-review.googlesource.com/98156
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
This CL moves the load/store combining tests into asmcheck.
In addition at being more compact, it's also now easier to
spot what it is missing in each architecture.
While doing so, I think I uncovered a bug in ppc64le and arm64
rules, because they fail to load/store combine in non-trivial
functions. Not sure why, I'll open an issue.
Change-Id: Ia1572d53c0553d9104f3e52b95e4d1768a8440a3
Reviewed-on: https://go-review.googlesource.com/98441
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Before this change, in case of any failure, asmcheck was
dumping to stderr the whole output of compile -S, which
can be very long if it contains multiple functions.
Make it so it filters the output to only display the
assembly output of functions for which at least one opcode
check failed. This greatly simplifies debugging.
Change-Id: I1bbf54473b8252a3384e2c1dade82d926afc119d
Reviewed-on: https://go-review.googlesource.com/98444
Run-TryBot: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Keith Randall <khr@golang.org>
Currently, the top-level testsuite always uses whatever version
of Go is found in the PATH to execute all the tests. This
forces the developers to tweak the PATH to run the testsuite.
Change it to use the same version of Go used to run run.go.
This allows developers to run the testsuite using the tip
compiler by simply saying "../bin/go run run.go".
I think this is a better solution compared to always forcing
"../bin/go", because it allows developers to run the testsuite
using different Go versions, for instance to check if a new
test is fixed in tip compared to the installed compiler.
Fixes#24217
Change-Id: I41b299c753b6e77c41e28be9091b2b630efea9d2
Reviewed-on: https://go-review.googlesource.com/98439
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
In RET instruction, the operand is the return jump's target,
which should be put in Prog.To.
Add an action "buildrundir" to the test driver, which builds
(compile+assemble+link) the code in a directory and runs the
resulting binary.
Fixes#23838.
Change-Id: I7ebe7eda49024b40a69a24857322c5ca9c67babb
Reviewed-on: https://go-review.googlesource.com/94175
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
This avoid simple bugs like "ADD" matching "FADD". Obviously
"ADD" will still match "ADDQ" so some care is still required
in this regard, but at least a first class of possible errors
is taken care of.
Change-Id: I7deb04c31de30bedac9c026d9889ace4a1d2adcb
Reviewed-on: https://go-review.googlesource.com/97817
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Keith Randall <khr@golang.org>
asmcheck comments now support a compact form of specifying
multiple checks for each platform, using the following syntax:
amd64:"SHL\t[$]4","SHR\t[$]4"
Negative checks are also parsed using the following syntax:
amd64:-"ROR"
though they are still not working.
Moreover, out-of-line comments have been implemented. This
allows to specify asmchecks on comment-only lines, that will
be matched on the first subsequent non-comment non-empty line.
// amd64:"XOR"
// arm:"EOR"
x ^= 1
Change-Id: I110c7462fc6a5c70fd4af0d42f516016ae7f2760
Reviewed-on: https://go-review.googlesource.com/97816
Reviewed-by: Keith Randall <khr@golang.org>
When the slice/string length is very large,
probably artifically large as in CL 97523,
adding BX (length) to R11 (pointer) overflows.
As a result, checking DI < R11 yields the wrong result.
Since they will be equal when the loop is done,
just check DI != R11 instead.
Yes, the pointer itself could overflow, but if that happens,
something else has gone pretty wrong; not our concern here.
Fixes#24187
Change-Id: I2f60fc6ccae739345d01bc80528560726ad4f8c6
Reviewed-on: https://go-review.googlesource.com/97802
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
The top-level test harness is modified to support a new kind
of test: "asmcheck". This is meant to replace asm_test.go
as an easier and more readable way to test code generation.
I've added a couple of codegen tests to get initial feedback
on the syntax. I've created them under a common "codegen"
subdirectory, so that it's easier to run them all with
"go run run.go -v codegen".
The asmcheck syntax allows to insert line comments that
can specify a regular expression to match in the assembly code,
for multiple architectures (the testsuite will automatically
build each testfile multiple times, one per mentioned architecture).
Negative matches are unsupported for now, so this cannot fully
replace asm_test yet.
Change-Id: Ifdbba389f01d55e63e73c99e5f5449e642101d55
Reviewed-on: https://go-review.googlesource.com/97355
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
OCOMPLIT stores the pre-typechecked type in n.Right, and then moves it
to n.Type. However, it wasn't clearing n.Right, so n.Right continued
to point to the OTYPE node. (Exception: slice literals reused n.Right
to store the array length.)
When exporting inline function bodies, we don't expect to need to save
any type aliases. Doing so wouldn't be wrong per se, but it's
completely unnecessary and would just bloat the export data.
However, reexportdep (whose role is to identify types needed by inline
function bodies) uses a generic tree traversal mechanism, which visits
n.Right even for O{ARRAY,MAP,STRUCT}LIT nodes. This means it finds the
OTYPE node, and mistakenly interpreted that the type alias needs to be
exported.
The straight forward fix is to just clear n.Right when typechecking
composite literals.
Fixes#24173.
Change-Id: Ia2d556bfdd806c83695b08e18b6cd71eff0772fc
Reviewed-on: https://go-review.googlesource.com/97719
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Otherwise, the error can be confusing if one forgets or doesn't know
that the builtin is being shadowed, which is not common practice.
Fixes#22822.
Change-Id: I735393b5ce28cb83815a1c3f7cd2e7bb5080a32d
Reviewed-on: https://go-review.googlesource.com/97455
Reviewed-by: Robert Griesemer <gri@golang.org>
This change enables printing of relative column information if a
prior line directive specified a valid column. If there was no
line directive, or the line directive didn't specify a column
(or the -C flag is specified), no column information is shown in
file positions.
Implementation: Column values (and line values, for that matter)
that are zero are interpreted as "unknown". A line directive that
doesn't specify a column records that as a zero column in the
respective PosBase data structure. When computing relative columns,
a relative value is zero of the base's column value is zero.
When formatting a position, a zero column value is not printed.
To make this work without special cases, the PosBase for a file
is given a concrete (non-0:0) position 1:1 with the PosBase's
line and column also being 1:1. In other words, at the position
1:1 of a file, it's relative positions are starting with 1:1 as
one would expect.
In the package syntax, this requires self-recursive PosBases for
file bases, matching what cmd/internal/src.PosBase was already
doing. In src.PosBase, file and inlining bases also need to be
based at 1:1 to indicate "known" positions.
This change completes the cmd/compiler part of the issue below.
Fixes#22662.
Change-Id: I6c3d2dee26709581fba0d0261b1d12e93f1cba1a
Reviewed-on: https://go-review.googlesource.com/97375
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
The first word of an interface is a pointer, but for the purposes
of GC we don't need to treat it as such.
1. If it is a non-empty interface, the pointer points to an itab
which is always in persistentalloc space.
2. If it is an empty interface, the pointer points to a _type.
a. If it is a compile-time-allocated type, it points into
the read-only data section.
b. If it is a reflect-allocated type, it points into the Go heap.
Reflect is responsible for keeping a reference to
the underlying type so it won't be GCd.
If we ever have a moving GC, we need to change this for 2b (as
well as scan itabs to update their itab._type fields).
Write barriers on the first word of interfaces have already been removed.
Change-Id: I643e91d7ac4de980ac2717436eff94097c65d959
Reviewed-on: https://go-review.googlesource.com/97518
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
We accidentally overlooked needing to still visit Ninit for OIF
statements with constant conditions in golang.org/cl/96778.
Fixes#24120.
Change-Id: I5b341913065ff90e1163fb872b9e8d47e2a789d2
Reviewed-on: https://go-review.googlesource.com/97475
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Extend cmd/internal/src.PosBase to track column information,
and adjust the meaning of the PosBase position to mean the
position at which the PosBase's relative (line, col) position
starts (rather than indicating the position of the //line
directive). Because this semantic change is made in the
compiler's noder, it doesn't affect the logic of src.PosBase,
only its test setup (where PosBases are constructed with
corrected incomming positions). In short, src.PosBase now
matches syntax.PosBase with respect to the semantics of
src.PosBase.pos.
For #22662.
Change-Id: I5b1451cb88fff3f149920c2eec08b6167955ce27
Reviewed-on: https://go-review.googlesource.com/96535
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
When we go from a branch block to a plain block, reset the
branch prediction bit. Downstream passes asssume that if the
branch prediction is set, then the block has 2 successors.
Fixes#23504
Change-Id: I2898ec002228b2e34fe80ce420c6939201c0a5aa
Reviewed-on: https://go-review.googlesource.com/88955
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
This replaces the contiguous heap arena mapping with a potentially
sparse mapping that can support heap mappings anywhere in the address
space.
This has several advantages over the current approach:
* There is no longer any limit on the size of the Go heap. (Currently
it's limited to 512GB.) Hence, this fixes#10460.
* It eliminates many failures modes of heap initialization and
growing. In particular it eliminates any possibility of panicking
with an address space conflict. This can happen for many reasons and
even causes a low but steady rate of TSAN test failures because of
conflicts with the TSAN runtime. See #16936 and #11993.
* It eliminates the notion of "non-reserved" heap, which was added
because creating huge address space reservations (particularly on
64-bit) led to huge process VSIZE. This was at best confusing and at
worst conflicted badly with ulimit -v. However, the non-reserved
heap logic is complicated, can race with other mappings in non-pure
Go binaries (e.g., #18976), and requires that the entire heap be
either reserved or non-reserved. We currently maintain the latter
property, but it's quite difficult to convince yourself of that, and
hence difficult to keep correct. This logic is still present, but
will be removed in the next CL.
* It fixes problems on 32-bit where skipping over parts of the address
space leads to mapping huge (and never-to-be-used) metadata
structures. See #19831.
This also completely rewrites and significantly simplifies
mheap.sysAlloc, which has been a source of many bugs. E.g., #21044,
#20259, #18651, and #13143 (and maybe #23222).
This change also makes it possible to allocate individual objects
larger than 512GB. As a result, a few tests that expected huge
allocations to fail needed to be changed to make even larger
allocations. However, at the moment attempting to allocate a humongous
object may cause the program to freeze for several minutes on Linux as
we fall back to probing every page with addrspace_free. That logic
(and this failure mode) will be removed in the next CL.
Fixes#10460.
Fixes#22204 (since it rewrites the code involved).
This slightly slows down compilebench and the x/benchmarks garbage
benchmark.
name old time/op new time/op delta
Template 184ms ± 1% 185ms ± 1% ~ (p=0.065 n=10+9)
Unicode 86.9ms ± 3% 86.3ms ± 1% ~ (p=0.631 n=10+10)
GoTypes 599ms ± 0% 602ms ± 0% +0.56% (p=0.000 n=10+9)
Compiler 2.87s ± 1% 2.89s ± 1% +0.51% (p=0.002 n=9+10)
SSA 7.29s ± 1% 7.25s ± 1% ~ (p=0.182 n=10+9)
Flate 118ms ± 2% 118ms ± 1% ~ (p=0.113 n=9+9)
GoParser 147ms ± 1% 148ms ± 1% +1.07% (p=0.003 n=9+10)
Reflect 401ms ± 1% 404ms ± 1% +0.71% (p=0.003 n=10+9)
Tar 175ms ± 1% 175ms ± 1% ~ (p=0.604 n=9+10)
XML 209ms ± 1% 210ms ± 1% ~ (p=0.052 n=10+10)
(https://perf.golang.org/search?q=upload:20171231.4)
name old time/op new time/op delta
Garbage/benchmem-MB=64-12 2.23ms ± 1% 2.25ms ± 1% +0.84% (p=0.000 n=19+19)
(https://perf.golang.org/search?q=upload:20171231.3)
Relative to the start of the sparse heap changes (starting at and
including "runtime: fix various contiguous bitmap assumptions"),
overall slowdown is roughly 1% on GC-intensive benchmarks:
name old time/op new time/op delta
Template 183ms ± 1% 185ms ± 1% +1.32% (p=0.000 n=9+9)
Unicode 84.9ms ± 2% 86.3ms ± 1% +1.65% (p=0.000 n=9+10)
GoTypes 595ms ± 1% 602ms ± 0% +1.19% (p=0.000 n=9+9)
Compiler 2.86s ± 0% 2.89s ± 1% +0.91% (p=0.000 n=9+10)
SSA 7.19s ± 0% 7.25s ± 1% +0.75% (p=0.000 n=8+9)
Flate 117ms ± 1% 118ms ± 1% +1.10% (p=0.000 n=10+9)
GoParser 146ms ± 2% 148ms ± 1% +1.48% (p=0.002 n=10+10)
Reflect 398ms ± 1% 404ms ± 1% +1.51% (p=0.000 n=10+9)
Tar 173ms ± 1% 175ms ± 1% +1.17% (p=0.000 n=10+10)
XML 208ms ± 1% 210ms ± 1% +0.62% (p=0.011 n=10+10)
[Geo mean] 369ms 373ms +1.17%
(https://perf.golang.org/search?q=upload:20180101.2)
name old time/op new time/op delta
Garbage/benchmem-MB=64-12 2.22ms ± 1% 2.25ms ± 1% +1.51% (p=0.000 n=20+19)
(https://perf.golang.org/search?q=upload:20180101.3)
Change-Id: I5daf4cfec24b252e5a57001f0a6c03f22479d0f0
Reviewed-on: https://go-review.googlesource.com/85887
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
