Per the language spec clarification in https://golang.org/cl/14727.
Updates #12576
Updates #12621
Change-Id: I1e459c3c11a571bd29582761faacaa9ca3178ba6
Reviewed-on: https://go-review.googlesource.com/91895
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Transform (ADDQconst SP) into (LEA SP), because lea is rematerializeable,
so this avoids register spill. We can't mark ADDQconst as rematerializeable,
because it clobbers flags. This makes go binary ~2kb smaller.
For reference here is generated code for function from bug report.
Before:
CALL "".g(SB)
MOVBLZX (SP), AX
LEAQ 8(SP), DI
TESTB AX, AX
JEQ 15
MOVQ "".p(SP), SI
DUFFCOPY $196
MOVQ $0, (SP)
PCDATA $0, $1
CALL "".h(SB)
RET
MOVQ DI, ""..autotmp_2-8(SP) // extra spill
PCDATA $0, $2
CALL "".g(SB)
MOVQ ""..autotmp_2-8(SP), DI // extra register fill
MOVQ "".p(SP), SI
DUFFCOPY $196
MOVQ $1, (SP)
PCDATA $0, $1
CALL "".h(SB)
JMP 14
END
After:
CALL "".g(SB)
MOVBLZX (SP), AX
TESTB AX, AX
JEQ 15
LEAQ 8(SP), DI
MOVQ "".p(SP), SI
DUFFCOPY $196
MOVQ $0, (SP)
PCDATA $0, $1
CALL "".h(SB)
RET
PCDATA $0, $0 // no spill
CALL "".g(SB)
LEAQ 8(SP), DI // rematerialized instead
MOVQ "".p(SP), SI
DUFFCOPY $196
MOVQ $1, (SP)
PCDATA $0, $1
CALL "".h(SB)
JMP 14
END
Fixes#22947
Change-Id: I8f33b860dc6c8828373477171b172ca2ce30074f
Reviewed-on: https://go-review.googlesource.com/81815
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Popcnt has false dependency on output register and generates
MOVQ $0, reg to break it. But recently we switched MOVQ $0, reg
encoding from xor reg, reg to actual mov $0, reg. This CL updates
code generation for popcnt to use actual XOR.
Change-Id: I4c1fc11e85758b53ba2679165fa55614ec54b27d
Reviewed-on: https://go-review.googlesource.com/82516
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Because getStackOffset is a function pointer, the compiler assumes that
its arguments escape. Pass a value instead to avoid heap allocations.
Change-Id: Ib94e5941847f134cd00e873040a4d7fcf15ced26
Reviewed-on: https://go-review.googlesource.com/92397
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Not a big improvement, but does help edge cases like the SSA package.
Change-Id: I40e531110b97efd5f45955be477fd0f4faa8d545
Reviewed-on: https://go-review.googlesource.com/92396
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Using bits.TrailingZeroes instead of iterating over each bit is a small
but easy win for the common case of only one or two registers being set.
I copied in the implementation for use with pre-1.9 bootstraps.
Change-Id: Ieaa768554d7d5239a5617fbf34f1ee0b32ce1de5
Reviewed-on: https://go-review.googlesource.com/92395
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Completely redesign and reimplement location list generation to be more
efficient, and hopefully not too hard to understand.
RegKills are gone. Instead of using the regalloc's liveness
calculations, redo them using the Ops' clobber information. Besides
saving a lot of Values, this avoids adding RegKills to blocks that would
be empty otherwise, which was messing up optimizations. This does mean
that it's much harder to tell whether the generation process is buggy
(there's nothing to cross-check it with), and there may be disagreements
with GC liveness. But the performance gain is significant, and it's nice
not to be messing with earlier compiler phases.
The intermediate representations are gone. Instead of producing
ssa.BlockDebugs, then dwarf.LocationLists, and then finally real
location lists, go directly from the SSA to a (mostly) real location
list. Because the SSA analysis happens before assembly, it stores
encoded block/value IDs where PCs would normally go. It would be easier
to do the SSA analysis after assembly, but I didn't want to retain the
SSA just for that.
Generation proceeds in two phases: first, it traverses the function in
CFG order, storing the state of the block at the beginning and end. End
states are used to produce the start states of the successor blocks. In
the second phase, it traverses in program text order and produces the
location lists. The processing in the second phase is redundant, but
much cheaper than storing the intermediate representation. It might be
possible to combine the two phases somewhat to take advantage of cases
where the CFG matches the block layout, but I haven't tried.
Location lists are finalized by adding a base address selection entry,
translating each encoded block/value ID to a real PC, and adding the
terminating zero entry. This probably won't work on OSX, where dsymutil
will choke on the base address selection. I tried emitting CU-relative
relocations for each address, and it was *very* bad for performance --
it uses more memory storing all the relocations than it does for the
actual location list bytes. I think I'm going to end up synthesizing the
relocations in the linker only on OSX, but TBD.
TestNexting needs updating: with more optimizations working, the
debugger doesn't stop on the continue (line 88) any more, and the test's
duplicate suppression kicks in. Also, dx and dy live a little longer
now, but they have the correct values.
Change-Id: Ie772dfe23a4e389ca573624fac4d05401ae32307
Reviewed-on: https://go-review.googlesource.com/89356
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
We're trying to enable location lists by default, and it's easier to do
that if we don't have to worry about scope tracking at the same time.
We can evaluate their performance impact separately.
However, that does mean that "err" is ambiguous in the test case, so
rename it to err2 for now.
Change-Id: I24f119016185c52b7d9affc74207f6a5b450fb6f
Reviewed-on: https://go-review.googlesource.com/89355
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
By default futexes are permitted in shared memory regions, which
requires the kernel to translate the memory address. Since our futexes
are never in shared memory, set FUTEX_PRIVATE_FLAG, which makes futex
operations slightly more efficient.
Change-Id: I2a82365ed27d5cd8d53c5382ebaca1a720a80952
Reviewed-on: https://go-review.googlesource.com/80144
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
The current assembler cannot handle PRFM(immediate) instruciton.
The fix creates a prfopfield struct that contains the eight
prefetch operations and the value to use in instruction. And add
the test cases.
Fixes#22932
Change-Id: I621d611bd930ef3c42306a4372447c46d53b2ccf
Reviewed-on: https://go-review.googlesource.com/81675
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Add support of NEG{V,W} pseudo-instructions, which are translated
to a SUB instruction from R0 with proper width.
Also turn illegal instruction to UNDEF, to avoid crashing in
asmout when it tries to read the operands.
Fixes#23548.
Change-Id: I047b27559ccd9594c3dcf62ab039b636098f30a3
Reviewed-on: https://go-review.googlesource.com/89896
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Just like all.bash passes flags to make.bash, I think it makes
sense that naclmake.bash and nacltest.bash do so as well. For
example, on a slow machine I can do "./nacltest.bash -v" to see
the build progress.
Change-Id: Id766dd590e6b83e8b5345822580dc1b05eac8ea3
Reviewed-on: https://go-review.googlesource.com/93117
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
On nacl/arm, R12 is clobbered by the RET instruction in function
that has a frame. runtime.udiv doesn't have a frame, so it does
not clobber R12.
Change-Id: I0de448749f615908f6659e92d201ba3eb2f8266d
Reviewed-on: https://go-review.googlesource.com/93116
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Replace explicit range loop that applies orderexprinplace on a
list of nodes with existing helper function orderexprlistinplace.
Passes toolstash -cmp.
Change-Id: Ic8098ed08cf67f319de3faa83b00a5b73bbde95d
Reviewed-on: https://go-review.googlesource.com/88815
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The pipe2 syscall is part of NetBSD since version 6.0 and thus exists in
all officially supported versions (6.0 through 6.1 and 7.0+).
Follows CL 38426
Change-Id: I7b62b507300c3dfbcc6ae56408a7d7088ddccc77
Reviewed-on: https://go-review.googlesource.com/94035
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The previous fix had "bug" and "build" in the wrong order.
Fixes#23791
Change-Id: I4897428516b159966c13c1054574c4f6fbf0fbac
Reviewed-on: https://go-review.googlesource.com/94017
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Include reader / writer interactions of RWMutex in the mutex profile.
Writer contention is already included in the profile, since a plain Mutex
is used to control exclusion.
Fixes#18496
Change-Id: Ib0dc1ffa0fd5e6d964a6f7764d7f09556eb63f00
Reviewed-on: https://go-review.googlesource.com/87095
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Peter Weinberger <pjw@google.com>
GCM allows using tag sizes smaller than the block size. This adds a
NewGCMWithNonceAndTagSize function which allows specifying the tag
size.
Fixes#19594
Change-Id: Ib2008c6f13ad6d916638b1523c0ded8a80eaf42d
Reviewed-on: https://go-review.googlesource.com/48510
Reviewed-by: Filippo Valsorda <hi@filippo.io>
Run-TryBot: Filippo Valsorda <hi@filippo.io>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The stackguard is set to stackPreempt earlier in reentersyscall, and
as it comes with throwsplit = true there's no way for the stackguard
to be set to anything else by the end of reentersyscall.
Change-Id: I4e942005b22ac784c52398c74093ac887fc8ec24
Reviewed-on: https://go-review.googlesource.com/65673
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Increase the sleep and wait for up to 2 seconds for the dup2.
Apparently it can sometimes take a long time.
Fixes#23784
Change-Id: I929530b057bbcd842b28a7640c39dd68d719ff7d
Reviewed-on: https://go-review.googlesource.com/93895
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Backslashes are ignored in Match and Glob on Windows, since those
collide with the separator character. However, they should still work in
both functions on other operating systems.
hasMeta did not reflect this logic - it always treated a backslash as a
non-special character. Do that only on Windows.
Assuming this is what the TODO was referring to, remove it. There are no
other characters that scanChunk treats especially.
Fixes#23418.
Change-Id: Ie0bd795812e0ed9d8c8c1bbc3137f29d960cba84
Reviewed-on: https://go-review.googlesource.com/87455
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
path.Match works purely with strings, not file paths. That's what sets
it apart from filepath.Match. For example, only filepath.Match will
change its behavior towards backslashes on Windows, to accomodate for
the file path separator on that system.
As such, path.Match should make no mention of file names. Nor should
path.ErrBadPattern mention globbing at all - the package has no notion
of globbing, and the error concerns only patterns.
For a similar reason, remove the mention of globbing from
filepath.ErrBadPattern. The error isn't reserved to just globbing, as it
can be returned from filepath.Match. And, as before, it only concerns
the patterns themselves.
Change-Id: I58a83ffa3e2549625d8e546ef916652525504bd1
Reviewed-on: https://go-review.googlesource.com/87857
Reviewed-by: Rob Pike <r@golang.org>
Fatalf calls in two Float tests use the %s verb with Floats values,
which is not allowed and results in failure messages that look like
this:
float_test.go:1385: i = 0, prec = 1, ToZero:
%!s(*big.Float=1) [0]
/ %!s(*big.Float=1) [0]
= %!s(*big.Float=0.0625)
want %!s(*big.Float=1)
Switch to %v.
Change-Id: Ifdc80bf19c91ca1b190f6551a6d0a51b42ed5919
Reviewed-on: https://go-review.googlesource.com/87199
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
ordercopyexpr is only called with 0 or 1 as value for the clear
argument. The clear variable in ordercopyexpr is only used in the
call to ordertemp which has a clear argument of type bool.
Change the clear argument of ordercopyexpr from int to bool and change
calls to ordercopyexpr to use false instead of 0 and true instead of 1.
Passes toolstash -cmp.
Change-Id: Ic264aafd3b0c8b99f6ef028ffaa2e30f23f9125a
Reviewed-on: https://go-review.googlesource.com/88115
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
This makes the constant names less verbose and aligns them more
with the Linux kernel which uses HWCAP_XXX for the constant names.
Change-Id: Ia7d079b59b57978adc045945951eaa1d99b41fac
Reviewed-on: https://go-review.googlesource.com/91738
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Also order the syscall number list by numerically for mips64x.
Follow-up for CL 92895.
Change-Id: I5f01f8c626132a06160997fce8a2aef0c486bb1c
Reviewed-on: https://go-review.googlesource.com/93616
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This moves the paragraph mentioning the use of _ higher up
to emphasize the warning and thereby reducing chances of getting
stuck.
Fixes#22617
Change-Id: I64352a3e966a22d86fc9d381332bade49d74714a
Reviewed-on: https://go-review.googlesource.com/87375
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The sub-word shifts need to sign-extend before shifting, to avoid
bringing in data from higher in the argument.
Fixes#23812
Change-Id: I0a95a0b49c48f3b40b85765bb4a9bb492be0cd73
Reviewed-on: https://go-review.googlesource.com/93716
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
As usual, adding go1.11 early in the cycle so that we can start
regression testing of the master toolchain.
Change-Id: Ie96eca7223722d60d7acc6b3b996b76740c36419
Reviewed-on: https://go-review.googlesource.com/93775
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
CL 92916 added the GOMAXPROCS test in TestTraceSymbolize.
This test only succeeds when the value of GOMAXPROCS changes.
Since the test calls runtime.GOMAXPROCS(1), it will fails
on machines where GOMAXPROCS=1.
This change fixes the test by calling runtime.GOMAXPROCS(oldGoMaxProcs+1).
Fixes#23816.
Change-Id: I1183dbbd7db6077cbd7fa0754032ff32793b2195
Reviewed-on: https://go-review.googlesource.com/93735
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
It makes no sense to try to get the zero value of a nil type, hence the
panic. When we have a nil type, use reflect.ValueOf(nil) instead.
This was showing itself if one used a missing field on the data between
parentheses, when the data was a nil interface:
t := template.Must(template.New("test").Parse(`{{ (.).foo }}`))
var v interface{}
t.Execute(os.Stdout, v)
Resulting in:
panic: reflect: Zero(nil) [recovered]
panic: reflect: Zero(nil)
Fixes#21171.
Change-Id: Ifcc4a0c67e6df425b65bc9f82fde6fcf03828579
Reviewed-on: https://go-review.googlesource.com/84482
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
Since https://golang.org/cl/38533, this validation is performed in
driverArgs.
Change-Id: I13a3ca46a1aa3197370de1095fb46ab83ea4628c
Reviewed-on: https://go-review.googlesource.com/91115
Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
Run-TryBot: Daniel Theophanes <kardianos@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
When 'convertAssign' gives an error, instead of giving just the index of
the failing column -- which is not always helpful, especially when there
are lots of columns in the query -- utilize 'rs.rowsi.Columns()' to
extract the underlying column name and include that in the error string:
sql: Scan error on column index 0, name "some_column": ...
Fixes#23362
Change-Id: I0fe71ff3c25f4c0dd9fc6aa2c2da2360dd93e3e0
Reviewed-on: https://go-review.googlesource.com/86537
Reviewed-by: Harald Nordgren <haraldnordgren@gmail.com>
Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
Run-TryBot: Daniel Theophanes <kardianos@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The former checks if a type has a method called "Format". The latter
checks if a type satisfies fmt.Formatter.
isFormatter does exactly what we want, so it's both simpler and more
accurate. Remove the only use of hasMethod in its favor.
Change-Id: Idc156a99081c3308f98512b87011a04aa8c6638d
Reviewed-on: https://go-review.googlesource.com/91215
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
All the other tools and commands print the usage text to standard error.
"go tool compile" was the odd one out, so fix it.
While at it, make objabi.Flagprint a bit more Go-like with an io.Writer
instead of a file descriptor, which is likely a leftover from the C
days.
Fixes#23234.
Change-Id: I9abf2e79461e61c8c8bfaee2c6bf8faf26e0e6c3
Reviewed-on: https://go-review.googlesource.com/85418
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Currently, if a sigpanic call is injected into C code, it's possible
for preparePanic to leave the stack in a state where traceback can't
unwind correctly past the sigpanic.
Specifically, shouldPushPanic sniffs the stack to decide where to put
the PC from the signal context. In the cgo case, it will find that
!findfunc(pc).valid() because pc is in C code, and then it will check
if the top of the stack looks like a Go PC. However, this stack slot
is just in a C frame, so it could be uninitialized and contain
anything, including what looks like a valid Go PC. For example, in
https://build.golang.org/log/c601a18e2af24794e6c0899e05dddbb08caefc17,
it sees 1c02c23a <runtime.newproc1+682>. When this condition is met,
it skips putting the signal PC on the stack at all. As a result, when
we later unwind from the sigpanic, we'll "successfully" but
incorrectly unwind to whatever PC was in this uninitialized slot and
go who knows where from there.
Fix this by making shouldPushPanic assume that the signal PC is always
usable if we're running C code, so we always make it appear like
sigpanic's caller.
This lets us be pickier again about unexpected return PCs in
gentraceback.
Updates #23640.
Change-Id: I1e8ade24b031bd905d48e92d5e60c982e8edf160
Reviewed-on: https://go-review.googlesource.com/91137
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This logic is duplicated in all of the preparePanic functions. Pull it
out into one architecture-independent function.
Change-Id: I7ef4e78e3eda0b7be1a480fb5245fc7424fb2b4e
Reviewed-on: https://go-review.googlesource.com/91255
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
While doing that, establish a negative value as signal for unknown
array lengths and adjust various array-length processing code to
handle that case.
Fixes#23712.
Change-Id: Icf488faaf972638b42b22d4b4607d1c512c8fc2c
Reviewed-on: https://go-review.googlesource.com/93438
Reviewed-by: Alan Donovan <adonovan@google.com>
