When heapBitsSetType repeats a source bitmap with a scalar tail
(typ.ptrdata < typ.size), it lays out the tail upon reaching the end
of the source bitmap by simply increasing the number of bits claimed
to be in the incoming bit buffer. This causes later iterations to read
the appropriate number of zeros out of the bit buffer before starting
on the next repeat of the source bitmap.
Currently, however, later iterations of the loop continue to read bits
from the source bitmap *regardless of the number of bits currently in
the bit buffer*. The bit buffer can only hold 32 or 64 bits, so if the
scalar tail is large and the padding bits exceed the size of the bit
buffer, the read from the source bitmap on the next iteration will
shift the incoming bits into oblivion when it attempts to put them in
the bit buffer. When the buffer does eventually shift down to where
these bits were supposed to be, it will contain zeros. As a result,
words that should be marked as pointers on later repetitions are
marked as scalars, so the garbage collector does not trace them. If
this is the only reference to an object, it will be incorrectly freed.
Fix this by adding logic to drain the bit buffer down if it is large
instead of reading more bits from the source bitmap.
Fixes#11286.
Change-Id: I964432c4b9f1cec334fc8c3da0ff16460203feb6
Reviewed-on: https://go-review.googlesource.com/11360
Reviewed-by: Russ Cox <rsc@golang.org>
When a method is called using the Type.Method(receiver, args...) syntax
without the receiver, or enough arguments, provide the more helpful
error message "not enough arguments in call to method expression
Type.Method" instead of the old message "not enough arguments in call
to Type.Method".
Fixes#8385
Change-Id: Id5037eb1ee5fa93687d4a6557b4a8233b29e9df2
Reviewed-on: https://go-review.googlesource.com/2193
Reviewed-by: Russ Cox <rsc@golang.org>
Also modified test/run.go to ignore messages prefixed <autogenerated>
because those cannot be described with "// ERROR ...", and backed out
patch from issue #9537 because it is no longer necessary. The reasons
described in the 9537 discussion for why escape analysis cannot run
late no longer hold, happily.
Fixes#11053.
Change-Id: Icb14eccdf2e8cde3d0f8fb8a216b765400a96385
Reviewed-on: https://go-review.googlesource.com/11088
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: David Chase <drchase@google.com>
Bool codegen was generating a temp for function calls
and other complex expressions, but was not using it.
This was a refactoring bug introduced by CL 7853.
The cmp code used to do (in short):
l, r := &n1, &n2
It was changed to:
l, r := nl, nr
But the requisite assignments:
nl, nr = &n1, &n2
were only introduced on one of two code paths.
Fixes#10654.
Change-Id: Ie8de0b3a333842a048d4308e02911bb10c6915ce
Reviewed-on: https://go-review.googlesource.com/10844
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
People invoking the linker directly already have to change their scripts
to use the new "go tool link", so this is a good time to make the -X flag
behave like all other Go flags and take just a single argument.
The old syntax will continue to be accepted (it is rewritten into the new
syntax before flag parsing). Maybe some day we will be able to retire it.
Even if we never retire the old syntax, having the new syntax at least
makes the rewriting much less of a kludge.
Change-Id: I91e8df94f4c22b2186e81d7f1016b8767d777eac
Reviewed-on: https://go-review.googlesource.com/10310
Reviewed-by: Rob Pike <r@golang.org>
Also modifies 'dist test' to use that sharding, and removes some old
temporary stuff from dist test which are no longer required.
'dist test' now also supports running a list of tests given in
arguments, mutually exclusive with the existing -run=REGEXP flag. The
hacky fast paths for avoiding the 1 second "go list" latency are now
removed and only apply to the case where partial tests are run via
args, instead of regex. The build coordinator will use both styles
for awhile. (the statically-sharded ARM builders on scaleway will
continue to use regexps, but the dynamically-shared builders on GCE
will use the list of tests)
Updates #10029
Change-Id: I557800a54dfa6f3b5100ef4c26fe397ba5189813
Reviewed-on: https://go-review.googlesource.com/10688
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Memory usage has been reduced.
The tests are still slow,
but that is issue #10571.
/usr/bin/time shows significant variation
in the peak memory usage compiling with tip.
This is unsurprising, given GC.
Using Go 1.4.2, memory is stable at 410mb.
Using tip at d2ee09298,
memory ranges from 470mb (+15%) to 534mb (+30%),
with a mean of 504mb (+23%), with n=50.
Fixes#9933.
Change-Id: Id31f3ae086ec324abf70e8f1a8044c4a0c27e274
Reviewed-on: https://go-review.googlesource.com/10211
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Use pkgimport == nil (or not) to distinguish between
parsing .go source files where "p" exponent specifier
is not allowed and parsing .a or .o export data where
it is. Use that to control error when p-exponent is
seen.
Fixes#9036
Change-Id: I8924f09c91d4945ef3f20e80a6e544008a94a7e4
Reviewed-on: https://go-review.googlesource.com/10450
Reviewed-by: Russ Cox <rsc@golang.org>
This is dead code that was missed
during the 'go tool compile' migration.
Change-Id: Ice2af8a9ef72f8fd5f82225ee261854d93b659f1
Reviewed-on: https://go-review.googlesource.com/10430
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Added a lineno parameter to treecopy and listtreecopy
(ignored if = 0). When nodes are copied the copy is
assigned the non-zero lineno (normally this would be
the destination).
Fixes#8183
Change-Id: Iffb767a745093fb89aa08bf8a7692c2f0122be98
Reviewed-on: https://go-review.googlesource.com/10334
Reviewed-by: Russ Cox <rsc@golang.org>
Before this change, the check for too-large arrays (and other large
types) occurred after escape analysis. If the data moved off stack
and onto the heap contained any pointers, it would therefore escape,
but because the too-large check occurred after escape analysis this
would not be recorded and a stack pointer would leak to the heap
(see the modified escape_array.go for an example).
Some of these appear to remain, in calls to typecheck from within walk.
Also corrected a few comments in escape_array.go about "BAD"
analysis that is now done correctly.
Enhanced to move aditional EscNone-but-large-so-heap checks into esc.c.
Change-Id: I770c111baff28a9ed5f8beb601cf09dacc561b83
Reviewed-on: https://go-review.googlesource.com/10268
Reviewed-by: Russ Cox <rsc@golang.org>
Indirect function and method calls should leak everything,
but they didn't.
This fix had no particular effect on the cost of running the
compiler on html/template/*.go and added a single new "escape"
to the standard library:
syscall/syscall_unix.go:85: &b[0] escapes to heap
in
if errno := m.munmap(uintptr(unsafe.Pointer(&b[0])),
uintptr(len(b))); errno != nil {
Added specific escape testing to escape_calls.go
(and verified that it fails without this patch)
I also did a little code cleanup around the changes in esc.c.
Fixes#10925
Change-Id: I9984b701621ad4c49caed35b01e359295c210033
Reviewed-on: https://go-review.googlesource.com/10295
Reviewed-by: Russ Cox <rsc@golang.org>
This CL removes the remaining visible uses of the "architecture letter" concept.
(They are no longer in tool names nor in source directory names.)
Because the architecture letter concept is now gone, delete GOCHAR
from "go env" output, and change go/build.ArchChar to return an
error always.
The architecture letter is still used in the compiler and linker sources
as a clumsy architecture enumeration, but that use is not visible to
Go users and can be cleaned up separately.
Change-Id: I4d97a38f372003fb610c9c5241bea440d9dbeb8d
Reviewed-on: https://go-review.googlesource.com/10289
Reviewed-by: Rob Pike <r@golang.org>
This CL fixes the build to use the newly created go tool compile
and go tool link in place of go tool 5g, go tool 5l, and so on.
See golang-dev thread titled "go tool compile, etc" for background.
Although it was not a primary motivation, this conversion does
reduce the wall clock time and cpu time required for make.bash
by about 10%.
Change-Id: I79cbbdb676cab029db8aeefb99a53178ff55f98d
Reviewed-on: https://go-review.googlesource.com/10288
Reviewed-by: Rob Pike <r@golang.org>
Modified esc.go to allow slice literals (before append)
to be non-escaping. Modified tests to account for changes
in escape behavior and to also test the two cases that
were previously not tested.
Also minor cleanups to debug-printing within esc.go
Allocation stats for running compiler
( cd src/html/template;
for i in {1..5} ; do
go tool 6g -memprofile=testzz.${i}.prof -memprofilerate=1 *.go ;
go tool pprof -alloc_objects -text testzz.${i}.prof ;
done ; )
before about 86k allocations
after about 83k allocations
Fixes#8972
Change-Id: Ib61dd70dc74adb40d6f6fdda6eaa4bf7d83481de
Reviewed-on: https://go-review.googlesource.com/10118
Reviewed-by: Russ Cox <rsc@golang.org>
This reverts commit 5726af54eb.
It broke all the builds.
Change-Id: I4b1dde86f9433717d303c1dabd6aa1a2bf97fab2
Reviewed-on: https://go-review.googlesource.com/10143
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Set overflowing integer constants to 1 rather than 0 to avoid
spurious div-zero errors in subsequent constant expressions.
Also: Exclude new test case from go/types test since it's
running too long (go/types doesn't have an upper constant
size limit at the moment).
Fixes#7746.
Change-Id: I3768488ad9909a3cf995247b81ee78a8eb5a1e41
Reviewed-on: https://go-review.googlesource.com/9165
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
The code generated for x = append(x, v) is roughly:
t := x
if len(t)+1 > cap(t) {
t = grow(t)
}
t[len(t)] = v
len(t)++
x = t
We used to generate this code as Go pseudocode during walk.
Generate it instead as actual instructions during gen.
Doing so lets us apply a few optimizations. The most important
is that when, as in the above example, the source slice and the
destination slice are the same, the code can instead do:
t := x
if len(t)+1 > cap(t) {
t = grow(t)
x = {base(t), len(t)+1, cap(t)}
} else {
len(x)++
}
t[len(t)] = v
That is, in the fast path that does not reallocate the array,
only the updated length needs to be written back to x,
not the array pointer and not the capacity. This is more like
what you'd write by hand in C. It's faster in general, since
the fast path elides two of the three stores, but it's especially
faster when the form of x is such that the base pointer write
would turn into a write barrier. No write, no barrier.
name old mean new mean delta
BinaryTree17 5.68s × (0.97,1.04) 5.81s × (0.98,1.03) +2.35% (p=0.023)
Fannkuch11 4.41s × (0.98,1.03) 4.35s × (1.00,1.00) ~ (p=0.090)
FmtFprintfEmpty 92.7ns × (0.91,1.16) 86.0ns × (0.94,1.11) -7.31% (p=0.038)
FmtFprintfString 281ns × (0.96,1.08) 276ns × (0.98,1.04) ~ (p=0.219)
FmtFprintfInt 288ns × (0.97,1.06) 274ns × (0.98,1.06) -4.94% (p=0.002)
FmtFprintfIntInt 493ns × (0.97,1.04) 506ns × (0.99,1.01) +2.65% (p=0.009)
FmtFprintfPrefixedInt 423ns × (0.97,1.04) 391ns × (0.99,1.01) -7.52% (p=0.000)
FmtFprintfFloat 598ns × (0.99,1.01) 566ns × (0.99,1.01) -5.27% (p=0.000)
FmtManyArgs 1.89µs × (0.98,1.05) 1.91µs × (0.99,1.01) ~ (p=0.231)
GobDecode 14.8ms × (0.98,1.03) 15.3ms × (0.99,1.02) +3.01% (p=0.000)
GobEncode 12.3ms × (0.98,1.01) 11.5ms × (0.97,1.03) -5.93% (p=0.000)
Gzip 656ms × (0.99,1.05) 645ms × (0.99,1.01) ~ (p=0.055)
Gunzip 142ms × (1.00,1.00) 142ms × (1.00,1.00) -0.32% (p=0.034)
HTTPClientServer 91.2µs × (0.97,1.04) 90.5µs × (0.97,1.04) ~ (p=0.468)
JSONEncode 32.6ms × (0.97,1.08) 32.0ms × (0.98,1.03) ~ (p=0.190)
JSONDecode 114ms × (0.97,1.05) 114ms × (0.99,1.01) ~ (p=0.887)
Mandelbrot200 6.11ms × (0.98,1.04) 6.04ms × (1.00,1.01) ~ (p=0.167)
GoParse 6.66ms × (0.97,1.04) 6.47ms × (0.97,1.05) -2.81% (p=0.014)
RegexpMatchEasy0_32 159ns × (0.99,1.00) 171ns × (0.93,1.07) +7.19% (p=0.002)
RegexpMatchEasy0_1K 538ns × (1.00,1.01) 550ns × (0.98,1.01) +2.30% (p=0.000)
RegexpMatchEasy1_32 138ns × (1.00,1.00) 135ns × (0.99,1.02) -1.60% (p=0.000)
RegexpMatchEasy1_1K 869ns × (0.99,1.01) 879ns × (1.00,1.01) +1.08% (p=0.000)
RegexpMatchMedium_32 252ns × (0.99,1.01) 243ns × (1.00,1.00) -3.71% (p=0.000)
RegexpMatchMedium_1K 72.7µs × (1.00,1.00) 70.3µs × (1.00,1.00) -3.34% (p=0.000)
RegexpMatchHard_32 3.85µs × (1.00,1.00) 3.82µs × (1.00,1.01) -0.81% (p=0.000)
RegexpMatchHard_1K 118µs × (1.00,1.00) 117µs × (1.00,1.00) -0.56% (p=0.000)
Revcomp 920ms × (0.97,1.07) 917ms × (0.97,1.04) ~ (p=0.808)
Template 129ms × (0.98,1.03) 114ms × (0.99,1.01) -12.06% (p=0.000)
TimeParse 619ns × (0.99,1.01) 622ns × (0.99,1.01) ~ (p=0.062)
TimeFormat 661ns × (0.98,1.04) 665ns × (0.99,1.01) ~ (p=0.524)
See next CL for combination with a similar optimization for slice.
The benchmarks that are slower in this CL are still faster overall
with the combination of the two.
Change-Id: I2a7421658091b2488c64741b4db15ab6c3b4cb7e
Reviewed-on: https://go-review.googlesource.com/9812
Reviewed-by: David Chase <drchase@google.com>
Today's earlier fix can stay, but it's a band-aid over the real problem,
which is that bad code was slipping through the type checker
into the back end (and luckily causing a type error there).
I discovered this because my new append does not use the same
temporaries and failed the test as written.
Fixes#9521.
Change-Id: I7e33e2ea15743406e15c6f3fdf73e1edecda69bd
Reviewed-on: https://go-review.googlesource.com/9921
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Instead of errors like:
./blank2.go:15: cannot use ~b1 (type []int) as type int in assignment
we now have:
./blank2.go:15: cannot use _ (type []int) as type int in assignment
Less confusing for users.
Fixes#9521
Change-Id: Ieab9859040e8e0df95deeaee7eeb408d3be61c0f
Reviewed-on: https://go-review.googlesource.com/9902
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Try to provide hints for common areas, either *interface
were interface would have been better, and note incorrect
capitalization (but don't be more ambitious than that, at
least not today).
Added code and test for cases
ptrInterface.ExistingMethod
ptrInterface.unexportedMethod
ptrInterface.MissingMethod
ptrInterface.withwRongcASEdMethod
interface.withwRongcASEdMethod
ptrStruct.withwRongcASEdMethod
struct.withwRongcASEdMethod
also included tests for related errors to check for
unintentional changes and consistent wording.
Somewhat simplified from previous versions to avoid second-
guessing user errors, yet also biased to point out most-likely
root cause.
Fixes#10700
Change-Id: I16693e93cc8d8ca195e7742a222d640c262105b4
Reviewed-on: https://go-review.googlesource.com/9731
Reviewed-by: Russ Cox <rsc@golang.org>
Noopt builds get a larger stack guard. This test must take that into account.
Change-Id: I1b5cbafdbbfee8c369ae1bebd0b900524ebf0d7d
Reviewed-on: https://go-review.googlesource.com/9610
Reviewed-by: Russ Cox <rsc@golang.org>
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720Fixes#10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Clean up after CL 5310.
Change-Id: Ib870e7b9d26eb118eefdaa3e76dcec4a4d459584
Reviewed-on: https://go-review.googlesource.com/9398
Reviewed-by: Ian Lance Taylor <iant@golang.org>
With this fix,
GOMAXPROCS=8 ./all.bash
passes, at least on my machine.
Fixes#10216.
Change-Id: Ib5991950892a1399ec81aced0a52b435e6f83fdf
Reviewed-on: https://go-review.googlesource.com/9392
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
These were fixed a little while ago, but overlooked when reenabling
disabled tests.
Update #9968.
Change-Id: I301ef587e580c517a170ad08ff897118b58cedec
Reviewed-on: https://go-review.googlesource.com/9347
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
We can expand the test cases as we discover problems.
This is some basic tests plus all the things I got wrong
in some recent work.
Change-Id: Id875fcfaf74eb087ae42b441fe47a34c5b8ccb39
Reviewed-on: https://go-review.googlesource.com/9158
Reviewed-by: Rick Hudson <rlh@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
In https://golang.org/cl/7797 I attempted to use myimportpath to set the value
of the go.importpath.$foo. symbol for the module being compiled, but I messed
it up and only set the name (which the linker rewrites anyway). This lead to
the importpath for the module being compiled being "". This was hard to notice,
because all modules that import another define the importpath for their
imported modules correctly -- but main is not imported, and this meant that the
reflect module saw all fields of all types defined in the main module as
exported.
The fix is to do what I meant to do the first time, add a test and change the
go tool to compile main packages with -p main and not -p
command-line-arguments.
Fixes#10332
Change-Id: I5fc6e9b1dc2b26f058641e382f9a56a526eca291
Reviewed-on: https://go-review.googlesource.com/8481
Reviewed-by: Russ Cox <rsc@golang.org>
The flag updates error annotations in test files from actual compiler output.
This is useful when doing compiler changes that add/remove/change lots of errors,
or when adding lots of new tests.
Also I noticed at least 2 cases where annotation were sub-optimal:
1. The annotation was "leaking param p" when the actual error is
"leaking param p to result ~r1".
2. The annotation was "leaking param m" when the actual errors
are "leaking param m" and "leaking param mv1".
For now it works only for errorcheck mode.
Also, apply the update to escape and liveness tests.
Some files have gccgo-specific errors of the form "gc error|gccgo error",
so it is risky to run update on all files. Gccgo-specific error
does not necessary contain '|', it can be just truncated.
Change-Id: Iaaae767f859dcb8321a8cb4970b2b70969e8a345
Reviewed-on: https://go-review.googlesource.com/5310
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This CL extends cmd/yacc to expose a yyErrorVerbose variable that
changes the error messages from just "syntax error" to "syntax error:
unexpected ${tokname}".
It also moves the yyToknames table generation to after rules have been
processed so that entries can be generated for tokens that aren't
mentioned in the preamble (e.g., '.' in the case of go.y).
Lastly, it restores gc's old code for applying yytfix to yyToknames,
except that substituting "LLITERAL" with litbuf happens in Yyerror.
Fixes#9968.
Change-Id: Icec188d11fdabc1dae31b8a471c35b5c7f6deec7
Reviewed-on: https://go-review.googlesource.com/8432
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
These registers are not available for programs to use. Prior to this
change, the compiler would crash attempting to use ZR as a general
purpose register. Other programs would compile but on execution would
overwrite the G register and cause havoc.
Fixes linux/arm64 build.
Fixes#10304Fixes#10320
Change-Id: I5cf51d3b77cfe3db7dd6377324950cafb02f8d8b
Reviewed-on: https://go-review.googlesource.com/8456
Reviewed-by: Minux Ma <minux@golang.org>
The original implementation used 16 int "words" but only 29 bits per word
for a total of 16*29 = 464 bits, with a space consumption of 16*64 = 1024
bits on a 64 bit machine. Switching to 512 bits increases precision while
still using (in the worst case) half the amount of memory per mp value on
a 64 bit machine.
Also: Decreased permitted number of least-significant mantissa bits which
may be incorrect when considering if a precise floating-point constant is
an integer from 29 to 16 bits.
Change-Id: Iee9287056f0e9aa4f06ceac0724ff4674f710c53
Reviewed-on: https://go-review.googlesource.com/8429
Reviewed-by: Russ Cox <rsc@golang.org>
All multi-precision arithmetic is now based on math/big.
- passes all.bash
- added test cases for fixed bugs
Fixes#7740.
Fixes#6866.
Change-Id: I67268b91766970ced3b928260053ccdce8753d58
Reviewed-on: https://go-review.googlesource.com/7912
Reviewed-by: Russ Cox <rsc@golang.org>
This restores go.errors from before 3af0d79 along with a fixed up
version of the bisonerrors AWK script, translated to Go.
However, this means Yyerror needs access to the yacc parser's state,
which is currently private. To workaround that, add a "state"
accessor method like the Lookahead method added in c7fa3c6.
Update issue #9968.
Change-Id: Ib868789e92fdb7d135442120a392457923e50121
Reviewed-on: https://go-review.googlesource.com/7270
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
On arm64, CMP $foo, R is encoded as from=$foo, reg=R, not as from=$foo,
to=R. The progtable entry for ACMP incorrectly described the latter
form. Because of this, the registerizer was not accounting the registers
used in CMP instructions and was incorrectly re-assigning those registers.
This was an old problem, but it only became apparent after b115c35
(cmd/internal/gc: move cgen, regalloc, et al to portable code). Previous
to this commit, the compiler used a slightly larger register set for the
temps than it used for register variables. Since it had plenty registers
dedicated to temps, the registers used in CMP instruction never clashed
with registers assigned to register variables.
Fixes#10253
Change-Id: Iedf4bd882bd59440dff310ac0f81e0f53d80d7ed
Reviewed-on: https://go-review.googlesource.com/8387
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
