Age | Commit message (Collapse) | Author |
|
The checkptr instrumentation is currently inserted before slice
operation has validated that n <= Big. So instead of panic, checkptr
have false positive throws.
To fix this, just insert the checkptr instrumentation after the bound
checking during SSA generation.
Fixes #46938
Change-Id: I9dbf84441c711842ccc883f3654ca8766ac696d8
Reviewed-on: https://go-review.googlesource.com/c/go/+/343972
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>
|
|
According to RISCV instruction set manual v2.2 Sec 6.1
MULHU followed by MUL will be fused into one multiply by microarchitecture
name old time/op new time/op delta
MulUintptr/small 11.2ns ±24% 9.2ns ± 0% -17.54% (p=0.000 n=10+9)
MulUintptr/large 15.9ns ± 0% 10.9ns ± 0% -31.55% (p=0.000 n=8+8)
Change-Id: I3d152218f83948cbc5c576bda29dc86e9b4206ee
Reviewed-on: https://go-review.googlesource.com/c/go/+/338753
Trust: Meng Zhuo <mzh@golangcn.org>
Reviewed-by: Joel Sing <joel@sing.id.au>
|
|
According to RISCV instruction set manual v2.2 Sec 6.1
MULHU followed by MUL will be fused into one multiply by microarchitecture
Benchstat on Hifive unmatched:
name old time/op new time/op delta
Hash8Bytes 245ns ± 3% 186ns ± 4% -23.99% (p=0.000 n=10+10)
Hash320Bytes 1.94µs ± 1% 1.31µs ± 1% -32.38% (p=0.000 n=9+10)
Hash1K 5.84µs ± 0% 3.84µs ± 0% -34.20% (p=0.000 n=10+9)
Hash8K 45.3µs ± 0% 29.4µs ± 0% -35.04% (p=0.000 n=10+10)
name old speed new speed delta
Hash8Bytes 32.7MB/s ± 3% 43.0MB/s ± 4% +31.61% (p=0.000 n=10+10)
Hash320Bytes 165MB/s ± 1% 244MB/s ± 1% +47.88% (p=0.000 n=9+10)
Hash1K 175MB/s ± 0% 266MB/s ± 0% +51.98% (p=0.000 n=10+9)
Hash8K 181MB/s ± 0% 279MB/s ± 0% +53.94% (p=0.000 n=10+10)
Change-Id: I3561495d02a4a0ad8578e9b9819bf0a4eaca5d12
Reviewed-on: https://go-review.googlesource.com/c/go/+/329970
Reviewed-by: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Meng Zhuo <mzh@golangcn.org>
|
|
Currently, deferreturn runs deferred functions by backing up its
return PC to the deferreturn call, and then effectively tail-calling
the deferred function (via jmpdefer). The effect of this is that the
deferred function appears to be called directly from the deferee, and
when it returns, the deferee calls deferreturn again so it can run the
next deferred function if necessary.
This unusual flow control leads to a large number of special cases and
complications all over the tool chain.
This used to be necessary because deferreturn copied the deferred
function's argument frame directly into its caller's frame and then
had to invoke that call as if it had been called from its caller's
frame so it could access it arguments. But now that we've simplified
defer processing so the runtime only deals with argument-less
closures, this approach is no longer necessary.
This CL simplifies all of this by making deferreturn simply call
deferred functions in a loop.
This eliminates the need for jmpdefer, so we can delete a bunch of
per-architecture assembly code.
This eliminates several special cases on Wasm, since it couldn't
support these calling shenanigans directly and thus had to simulate
the loop a different way. Now Wasm can largely work the way the other
platforms do.
This eliminates the per-architecture Ginsnopdefer operation. On PPC64,
this was necessary to reload the TOC pointer after the tail call
(since TOC pointers in general make tail calls impossible). The tail
call is gone, and in the case where we do force a jump to the
deferreturn call when recovering from an open-coded defer, we go
through gogo (via runtime.recovery), which handles the TOC. On other
platforms, we needed a NOP so traceback didn't get confused by seeing
the return to the CALL instruction, rather than the usual return to
the instruction following the CALL instruction. Now we don't inject a
return to the CALL instruction at all, so this NOP is also
unnecessary.
The one potential effect of this is that deferreturn could now appear
in stack traces from deferred functions. However, this could already
happen from open-coded defers, so we've long since marked deferreturn
as a "wrapper" so it gets elided not only from printed stack traces,
but from runtime.Callers*.
This is a retry of CL 337652 because we had to back out its parent.
There are no changes in this version.
Change-Id: I3f54b7fec1d7ccac71cc6cf6835c6a46b7e5fb6c
Reviewed-on: https://go-review.googlesource.com/c/go/+/339397
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Previously, softfloat mode does not work with register ABI, mainly
because the compiler doesn't know how to pass floating point
arguments and results. According to the ABI it should be passed in
FP registers, but there isn't any in softfloat mode.
This CL makes it work. When softfloat is used, we define the ABI
as having 0 floating point registers (because there aren't any).
The integer registers are unchanged. So floating point arguments
and results are passed in memory.
Another option is to pass (the bit representation of) floating
point values in integer registers. But this complicates things
because it'd need to reorder integer argument registers.
Change-Id: Ibecbeccb658c10a868fa7f2dcf75138f719cc809
Reviewed-on: https://go-review.googlesource.com/c/go/+/327274
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
|
|
replace jmpdefer with a loop"
This reverts CL 227652.
I'm reverting CL 337651 and this builds on top of it.
Change-Id: I03ce363be44c2a3defff2e43e7b1aad83386820d
Reviewed-on: https://go-review.googlesource.com/c/go/+/338709
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Currently, deferreturn runs deferred functions by backing up its
return PC to the deferreturn call, and then effectively tail-calling
the deferred function (via jmpdefer). The effect of this is that the
deferred function appears to be called directly from the deferee, and
when it returns, the deferee calls deferreturn again so it can run the
next deferred function if necessary.
This unusual flow control leads to a large number of special cases and
complications all over the tool chain.
This used to be necessary because deferreturn copied the deferred
function's argument frame directly into its caller's frame and then
had to invoke that call as if it had been called from its caller's
frame so it could access it arguments. But now that we've simplified
defer processing so the runtime only deals with argument-less
closures, this approach is no longer necessary.
This CL simplifies all of this by making deferreturn simply call
deferred functions in a loop.
This eliminates the need for jmpdefer, so we can delete a bunch of
per-architecture assembly code.
This eliminates several special cases on Wasm, since it couldn't
support these calling shenanigans directly and thus had to simulate
the loop a different way. Now Wasm can largely work the way the other
platforms do.
This eliminates the per-architecture Ginsnopdefer operation. On PPC64,
this was necessary to reload the TOC pointer after the tail call
(since TOC pointers in general make tail calls impossible). The tail
call is gone, and in the case where we do force a jump to the
deferreturn call when recovering from an open-coded defer, we go
through gogo (via runtime.recovery), which handles the TOC. On other
platforms, we needed a NOP so traceback didn't get confused by seeing
the return to the CALL instruction, rather than the usual return to
the instruction following the CALL instruction. Now we don't inject a
return to the CALL instruction at all, so this NOP is also
unnecessary.
The one potential effect of this is that deferreturn could now appear
in stack traces from deferred functions. However, this could already
happen from open-coded defers, so we've long since marked deferreturn
as a "wrapper" so it gets elided not only from printed stack traces,
but from runtime.Callers*.
Change-Id: Ie9f700cd3fb774f498c9edce363772a868407bf7
Reviewed-on: https://go-review.googlesource.com/c/go/+/337652
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Introduce new dynamic dottype operations which take a dynamic
instead of static type to convert to.
Change-Id: I5824a1fea056fe811b1226ce059e1e8da1baa335
Reviewed-on: https://go-review.googlesource.com/c/go/+/337609
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
|
|
Conflicts:
- src/cmd/compile/internal/ssagen/ssa.go
CL 336629 touched code that had already been removed on dev.typeparams.
Merge List:
+ 2021-07-26 ecaa6816bf doc: clarify non-nil zero length slice to array pointer conversion
+ 2021-07-26 1868f8296e crypto/x509: update iOS bundled roots to version 55188.120.1.0.1
+ 2021-07-25 849b791129 spec: use consistent capitalization for rune literal hex constants
+ 2021-07-23 0914646ab9 doc/1.17: fix two dead rfc links
+ 2021-07-22 052da5717e cmd/compile: do not change field offset in ABI analysis
Change-Id: Ie570ec3f6a3241e0495e39e8a73b3a09a9368605
|
|
Currently, the ABI analysis assigns parameter/result offsets
to the fields of function *Type. In some cases, we may have
an ABI0 function reference and an ABIInternal reference share
the same function *Type. For example, for an ABI0 function F,
"f := F" will make f and (ABI0) F having the same *Type. But f,
as a func value, should use ABIInternal. Analyses on F and f will
collide and cause ICE.
Also, changing field offsets in ABI analysis has to be done very
carefully to avoid data races. It has been causing
trickiness/difficulty.
This CL removes the change of field offsets in ABI analysis
altogether. The analysis result is stored in ABIParamAssignment,
which is the only way to access parameter/result stack offset now.
Fixes #47317.
Fixes #47227.
Change-Id: I23a3e081a6cf327ac66855da222daaa636ed1ead
Reviewed-on: https://go-review.googlesource.com/c/go/+/336629
Trust: Cherry Mui <cherryyz@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Conflicts:
- src/runtime/internal/sys/zgoarch_386.go
- src/runtime/internal/sys/zgoarch_amd64.go
- src/runtime/internal/sys/zgoarch_arm.go
- src/runtime/internal/sys/zgoarch_arm64.go
- src/runtime/internal/sys/zgoarch_arm64be.go
- src/runtime/internal/sys/zgoarch_armbe.go
- src/runtime/internal/sys/zgoarch_mips.go
- src/runtime/internal/sys/zgoarch_mips64.go
- src/runtime/internal/sys/zgoarch_mips64le.go
- src/runtime/internal/sys/zgoarch_mips64p32.go
- src/runtime/internal/sys/zgoarch_mips64p32le.go
- src/runtime/internal/sys/zgoarch_mipsle.go
- src/runtime/internal/sys/zgoarch_ppc.go
- src/runtime/internal/sys/zgoarch_ppc64.go
- src/runtime/internal/sys/zgoarch_ppc64le.go
- src/runtime/internal/sys/zgoarch_riscv.go
- src/runtime/internal/sys/zgoarch_riscv64.go
- src/runtime/internal/sys/zgoarch_s390.go
- src/runtime/internal/sys/zgoarch_s390x.go
- src/runtime/internal/sys/zgoarch_sparc.go
- src/runtime/internal/sys/zgoarch_sparc64.go
- src/runtime/internal/sys/zgoarch_wasm.go
On dev.typeparams, CL 328336 moved these files to internal/goarch;
whereas on master, CL 333909 reserved GOARCH=loong64. For this CL,
I resolved the conflict by simply running "go generate internal/goarch".
Merge List:
+ 2021-07-19 c8f4e6152d spec: correct example comment in Conversions from slice to array
+ 2021-07-19 1d91551b73 time: correct typo in documentation for UnixMicro
+ 2021-07-19 404127c30f cmd/compile: fix off-by-one error in traceback argument counting
+ 2021-07-19 6298cfe672 cmd/compile: fix typo in fatal message of builtinCall
+ 2021-07-19 49402bee36 cmd/{compile,link}: fix bug in map.zero handling
+ 2021-07-18 a66190ecee test/bench/go1: fix size for RegexpMatchMedium_32
+ 2021-07-18 650fc2117a text/scanner: use Go convention in Position doc comment
+ 2021-07-16 aa4e0f528e net/http: correct capitalization in cancelTimeBody comment
+ 2021-07-15 0941dbca6a testing: clarify in docs that TestMain is advanced
+ 2021-07-15 69728ead87 cmd/go: update error messages in tests to match CL 332573
+ 2021-07-15 c1cc9f9c3d cmd/compile: fix lookup package of redeclared dot import symbol
+ 2021-07-15 21a04e3335 doc/go1.17: mention GOARCH=loong64
+ 2021-07-14 2b00a54baf go/build, runtime/internal/sys: reserve GOARCH=loong64
+ 2021-07-14 60ddf42b46 cmd/go: change link in error message from /wiki to /doc.
+ 2021-07-13 d8f348a589 cmd/go: remove a duplicated word from 'go help mod graph'
+ 2021-07-12 a98589711d crypto/tls: test key type when casting
+ 2021-07-12 cfbd73ba33 doc/go1.17: editing pass over the "Compiler" section
+ 2021-07-09 ab4085ce84 runtime/pprof: call runtime.GC twice in memory profile test
Change-Id: I1490a4c7e4c560659c21a4eb67d243f35d1f908e
|
|
For traceback argument printing, we want to print at most 10
words, then print "..." if there are still more args and/or
fields. The current code has off-by-one error that for 11
non-aggregate typed args, it prints the first 10 but without the
"...". Also, for aggregate-typed args, in some cases it may print
an extra "..." when there is actually no more fields.
The problem for this is that visitType return false (meaning not
to continue visiting) if it reaches the limit anywhere during the
recursive visit. It doesn't distinguish whether it has printed
anything for the current arg. If it reaches the limit before it
prints anything, it means that we're visiting the extra arg/field,
so the caller should print "..." and stop. If it prints
something then reaches the limit, however, the caller should keep
going, and only print "..." at the next iteration when there is
actually an extra arg/field. This CL does so.
Fixes #47159.
Change-Id: I93fc25b73ada2b5a98df780c45e5b0c9565dc2fc
Reviewed-on: https://go-review.googlesource.com/c/go/+/334710
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
This CL reuses the symabi wrapper information to warn when a function
is defined both in Go (i.e., has a function declaration with a body)
and in assembly (i.e., has a TEXT instruction).
This will eventually produce a linker error anyway, but we can provide
a slightly nicer error message earlier.
Change-Id: Ia107f813343c0b10f4cd1013e7c72e67149ee52e
Reviewed-on: https://go-review.googlesource.com/c/go/+/333454
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
CL 330831 moved rewrite method calls to typecheck pass, then add Fatalf
check for mis-used of OCALLMETH in all frontend passes. The check in SSA
generation pass is now redundant.
Change-Id: If959b774ab3c3aeefabf32333f77b9a93c156ce3
Reviewed-on: https://go-review.googlesource.com/c/go/+/330834
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>
|
|
Conflicts:
- src/go/types/check_test.go
CL 330629 fixed a bug in package qualification logic
- src/internal/buildcfg/exp.go
CL 329930 make parseExperiments get go arch string as input param
Merge List:
+ 2021-06-25 37f9a8f69d go/types: fix a bug in package qualification logic
+ 2021-06-24 c309c89db5 reflect: document that InterfaceData is a low-entropy RNG
+ 2021-06-24 cce621431a cmd/compile: fix wrong type in SSA generation for OSLICE2ARRPTR
+ 2021-06-24 600a2a4ffb cmd/go: don't try to add replaced versions that won't be selected
+ 2021-06-24 a9bb38222a net: remove hard-coded timeout in dialClosedPort test helper
+ 2021-06-24 86d72fa2cb time: handle invalid UTF-8 byte sequences in quote to prevent panic
+ 2021-06-24 44a12e5f33 cmd/go: search breadth-first instead of depth-first for test dependency cycles
+ 2021-06-24 73496e0df0 net: use absDomainName in the Windows lookupPTR test helper
+ 2021-06-24 222ed1b38a os: enable TestFifoEOF on openbsd
+ 2021-06-22 0ebd5a8de0 cmd/go: update ToolTags based on GOARCH value
+ 2021-06-22 5bd09e5efc spec: unsafe.Add/Slice are not permitted in statement context
+ 2021-06-22 666315b4d3 runtime/internal/atomic: remove incorrect pointer indirection in comment
+ 2021-06-22 63daa774b5 go/types: guard against checking instantiation when generics is disabled
+ 2021-06-22 197a5ee2ab cmd/gofmt: remove stale documentation for the -G flag
+ 2021-06-22 9afd158eb2 go/parser: parse an ast.IndexExpr for a[]
+ 2021-06-21 1bd5a20e3c cmd/go: add a -go flag to 'go mod graph'
+ 2021-06-21 761edf71f6 cmd/internal/moddeps: use a temporary directory for GOMODCACHE if needed
+ 2021-06-21 a0400420ad cmd/internal/moddeps: use -mod=readonly instead of -mod=mod
+ 2021-06-21 3f9ec83b10 cmd/go: document GOPPC64 environment variable
+ 2021-06-21 20bdfba325 go/scanner: fall back to next() when encountering 0 bytes in parseIdentifier
+ 2021-06-21 44f9a3566c database/sql: fix deadlock test in prepare statement
Change-Id: I16490e8ea70ee65081f467223857033842da513a
|
|
OCALLMETH is rewritten by walkCall to OCALLFUNC, and other places in
backend have already caught it. So do the same thing in state.expr for
consistency and prevent mis-use in frontend side.
While at it, also remove un-used function getParam.
Change-Id: I03e1ea907e0bcb05fa35fa81804c33b5c9a4d77e
Reviewed-on: https://go-review.googlesource.com/c/go/+/330669
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>
|
|
Fixes #46907
Change-Id: I6a2728d2f2159df583b32f40f6100d3e90c34dd7
Reviewed-on: https://go-review.googlesource.com/c/go/+/330672
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>
|
|
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>
|
|
The current go/defer wrapping code goes to some length to clear
ClosureCalled when a function call will end up not being called
directly, and so it will need to use the context register.
But we already have a flag to indicate we need to use the context
register: Needctxt. The real issue here is just that buildssa was
using fn.ClosureCalled instead of fn.Needctxt.
Change-Id: Ic9f5f23b66eb467fc61fa84eacb45d46c54133d2
Reviewed-on: https://go-review.googlesource.com/c/go/+/330329
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
|
|
This CL implements a few improvements to SSA devirtualization to make
it simpler and more general:
1. Change reflectdata.ITabAddr to now immediately generate the wrapper
functions and write out the itab symbol data. Previously, these were
each handled by separate phases later on.
2. Removes the hack in typecheck where we marked itabs that we
expected to need later. Instead, the calls to ITabAddr in walk now
handle generating the wrappers.
3. Changes the SSA interface call devirtualization algorithm to just
use the itab symbol data (namely, its relocations) to figure out what
pointer is available in memory at the given offset. This decouples it
somewhat from reflectdata.
Change-Id: I8fe06922af8f8a1e7c93f5aff2b60ff59b8e7114
Reviewed-on: https://go-review.googlesource.com/c/go/+/327871
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Now that deferred functions are always argumentless, we don't
need the metadata for the frame size, number of arguments, and
the information about each argument.
Change-Id: I99e75248a22bda6efbdf2012a2f35beca4c18fd7
Reviewed-on: https://go-review.googlesource.com/c/go/+/326061
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Now it is only used to save the deferred the function (closure),
which must be a function type. Simplify the code.
Change-Id: Id4b8f2760fbf39a95883df2327f97378e7edab88
Reviewed-on: https://go-review.googlesource.com/c/go/+/326060
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
As deferred function now always has zero arguments, _defer.siz is
always 0 and can be removed.
Change-Id: Ibb89f65b2f9d2ba4aeabe50438cc3d4b6a88320b
Reviewed-on: https://go-review.googlesource.com/c/go/+/325921
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
newproc/deferproc
newproc/deferproc takes a siz argument for the go'd/deferred
function's argument size. Now it is always zero. Remove the
argument.
Change-Id: If1bb8d427e34015ccec0ba10dbccaae96757fa8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/325917
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
Now that defer wrapping is used, deferred function is always
argumentless. Remove the code handling arguments.
This CL is mostly removing the fallback code path. There are more
cleanups to be done, in later CLs.
Change-Id: If6c729d3055c7a507cb1f1a000f5bbd3ad7ff235
Reviewed-on: https://go-review.googlesource.com/c/go/+/325914
Trust: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Hardwire regabidefers to true. Remove it from GOEXPERIMENTs.
Fallback paths are not cleaned up in this CL. That will be done
in later CLs.
Change-Id: Iec1112a1e55d5f6ef70232a5ff6e702f649071c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/325913
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Add code that loads results into registers (used in defer return
code path) and spills argument registers (used for partially lived
in-register args).
Move some code from the amd64 package to a common place.
Change-Id: I8d59b68693048fdba86e10769c4ac58de5fcfb64
Reviewed-on: https://go-review.googlesource.com/c/go/+/322851
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
This CL makes to minor changes motivated by making it easier to make
large-scale changes to the inliner while satisfying toolstash -cmp:
1. When creating inlining variables, make sure to preserve the
AutoTemp flag. This is necessary so that temporary variables
introduced by rewriting f(g()) calls during typecheck stay autotemp
after inlining and are (correctly) omitted from DWARF debugging
information.
2. When sorting variables for stack frame layout, use a stable
sort. This ensures that layout is insensitive to whether deadcode
elimination happens before or after inlining.
Change-Id: I672e752a873c7e16749b9873fd6573607e074309
Reviewed-on: https://go-review.googlesource.com/c/go/+/323011
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
|
|
Spill arg registers before calling morestack, and reload after.
Change-Id: I09404def321b8f935d5e8836a46ccae8256d0d55
Reviewed-on: https://go-review.googlesource.com/c/go/+/322853
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
|
|
steals idea from CL 312093
further investigation revealed additional duplicate
slots (equivalent, but not equal), so delete those too.
Rearranged Func.Names to be addresses of slots,
create canonical addresses so that split slots
(which use those addresses to refer to their parent,
and split slots can be further split)
will preserve "equivalent slots are equal".
Removes duplicates, improves metrics for "args at entry".
Change-Id: I5bbdcb50bd33655abcab3d27ad8cdce25499faaf
Reviewed-on: https://go-review.googlesource.com/c/go/+/312292
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
I found a performance regression between the runtime.Hash
benchmarks when comparing Go 1.16 to latest on ppc64le. This
was due to the addition of Mul64 to runtime/internal/math
with the comments that this should be treated as an intrinsic
on platforms where available. However this is was not being
intrinsified on ppc64le because the code in ssagen/ssa.go didn't
correctly specify ppc64le. It had the argument for ArchPPC64
but should have also included ArchPPC64LE.
Treating Mul64 as an intrinsic shows this improvement, and these
results are better than Go 1.16:
Hash5 27.0ns ± 0% 14.0ns ± 0% -48.1
Hash16 26.6ns ± 0% 14.0ns ± 0% -47.3
Hash64 50.7ns ± 0% 17.9ns ± 0% -64.6
Hash1024 500ns ± 0% 99ns ± 0% -80.1
Hash65536 30.6µs ± 0% 4.0µs ± 0% -86
Many of the Map related benchmarks in the runtime package also showed
improvement once Mul64 is intrinsified.
Change-Id: I3b5ce637b1620348d81a30cfc359c97ab63aae0f
Reviewed-on: https://go-review.googlesource.com/c/go/+/317303
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Paul Murphy <murp@ibm.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
This disables the "testing names" for method names and
trailing input types passed to closure/interface/other calls.
The logic using the names remains, so that editing the change
to enable local testing is not too hard.
Also fixes broken build tag in reflect/abi_test.go
Updates #44816.
Change-Id: I3d222d2473c98d04ab6f1122ede9fea70c994af1
Reviewed-on: https://go-review.googlesource.com/c/go/+/300150
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Lots of constant SSA values we put in the entry block so that
CSE can easily combine them. With -N, however, we don't run CSE, so
putting values in the entry block only serves to extend their lifetime
for no benefit.
Fixes #45897. The number of live SSA values per block goes from >5K to 22.
Memory use goes from ~3GB to ~400MB.
Change-Id: I620b423611790a900e0d4cd270eac5dbdddf2a2b
Reviewed-on: https://go-review.googlesource.com/c/go/+/316369
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
|
|
Fix up a slightly stale comment in the part of ssa generation that
zeros ambiguously live variables: with the advent of the register ABI,
the ir.Func "Dcl" slice is no longer entirely sorted by frame offset,
although this is still the case for the local vars in Dcl.
Change-Id: I633f43d16f0d4e0b444193a6edb6b2aa1154eea7
Reviewed-on: https://go-review.googlesource.com/c/go/+/316309
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>
|
|
For ABI0 assembly functions that have Go declarations, generate
traceback argument info and attach it to the assembly functions.
So we can print argument in tracebacks if e.g. assembly function
panics.
Only do this for ABI0 functions, as for ABIInternal assembly
functions it is likely that they never spill arguments to memory.
Change-Id: I7e601ccd9aded5e6af2f02be975bf81ff9948f4d
Reviewed-on: https://go-review.googlesource.com/c/go/+/315870
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>
|
|
Updates #19367.
Updates #40481.
Change-Id: Iabd2afdd0d520e5d68fd9e6dedd013335a4b3886
Reviewed-on: https://go-review.googlesource.com/c/go/+/312214
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
Revise the code that generates DWARF location expressions for input
parameters to get it to work properly with the new register ABI when
optimization is turned off.
The previously implementation assumed stack locations for all
input+output parameters when -N (disable optimization) was in effect.
In the new implementation, a register-resident input parameter is
given a 2-element location list, the first list element pointing to
the ABI register(s) containing the param, and the second element
pointing to the stack home once it has been spilled.
NB, this change fixes a bunch of the Delve pkg/proc unit tests (maybe
about half of the outstanding failures). Still a good number that need
to be investigated, however.
Updates #40724.
Updates #45720.
Change-Id: I743bbb9af187bcdebeb8e690fdd6db58094ca415
Reviewed-on: https://go-review.googlesource.com/c/go/+/314431
Trust: Than McIntosh <thanm@google.com>
Trust: Cherry Zhang <cherryyz@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 SSA code for debug variable location analysis (for DWARF) has two
special 'sentinel' values that it uses to handshake with the
debugInfo.GetPC callback when capturing the PC values of debug
variable ranges after prog generatoin: "BlockStart" and "BlockEnd".
"BlockStart" has the expected semantics: it means "the PC value of the
first instruction of block B", but "BlockEnd" does not mean "PC value
of the last instruction of block B", but rather it is implemented as
"the PC value of the last instruction of the function". This causes
confusion when reading the code, and seems to to result in implementation
flaws in the past, leading to incorrect ranges in some cases.
To help with this, add a new sentinel "FuncEnd" (which has the "last
inst in the function" semantics) and change the implementation of
"BlockEnd" to actually mean what its name implies (last inst in
block).
Updates #45720.
Change-Id: Ic3497fb60413e898d2bfe27805c3db56483d12a2
Reviewed-on: https://go-review.googlesource.com/c/go/+/314930
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>
|
|
Remove unused function AddrForParamSlot.
Change-Id: I8e3ed8cc6607d30ad6da7bc6ccbaa87b7e001e79
Reviewed-on: https://go-review.googlesource.com/c/go/+/314909
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>
|
|
With the new register ABI, the compiler sometimes introduces spills of
argument registers in function prologs; depending on the positions
assigned to these spills and whether they have the IsStmt flag set,
this can degrade the debugging experience. For example, in this
function from one of the Delve regression tests:
L13: func foo((eface interface{}) {
L14: if eface != nil {
L15: n++
L16: }
L17 }
we wind up with a prolog containing two spill instructions, the first
with line 14, the second with line 13. The end result for the user
is that if you set a breakpoint in foo and run to it, then do "step",
execution will initially stop at L14, then jump "backwards" to L13.
The root of the problem in this case is that an ArgIntReg pseudo-op is
introduced during expand calls, then promoted (due to lowering) to a
first-class statement (IsStmt flag set), which in turn causes
downstream handling to propagate its position to the first of the register
spills in the prolog.
To help improve things, this patch changes the rewriter to avoid
moving an "IsStmt" flag from a deleted/replaced instruction to an
Arg{Int,Float}Reg value, and adds Arg{Int,Float}Reg to the list of
opcodes not suitable for selection as statement boundaries, and
suppresses generation of additional register spills in defframe() when
optimization is disabled (since in that case things will get spilled
in any case).
This is not a comprehensive/complete fix; there are still cases where
we get less-than-ideal source position markers (ex: issue 45680).
Updates #40724.
Change-Id: Ica8bba4940b2291bef6b5d95ff0cfd84412a2d40
Reviewed-on: https://go-review.googlesource.com/c/go/+/312989
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
e.g., LeadingZeros, bits.LeadingZeros, math/bits.LeadingZeros
but not its.LeadingZeros
Change-Id: Ib9a57e4db0af03c55bf5b5027aa6f8a5a84f2134
Reviewed-on: https://go-review.googlesource.com/c/go/+/312291
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Currently, when the runtime printing a stack track (at panic, or
when runtime.Stack is called), it prints the function arguments
as words in memory. With a register-based calling convention,
the layout of argument area of the memory changes, so the
printing also needs to change. In particular, the memory order
and the syntax order of the arguments may differ. To address
that, this CL lets the compiler to emit some metadata about the
memory layout of the arguments, and the runtime will use this
information to print arguments in syntax order.
Previously we print the memory contents of the results along with
the arguments. The results are likely uninitialized when the
traceback is taken, so that information is rarely useful. Also,
with a register-based calling convention the results may not
have corresponding locations in memory. This CL changes it to not
print results.
Previously the runtime simply prints the memory contents as
pointer-sized words. With a register-based calling convention,
as the layout changes, arguments that were packed in one word
may no longer be in one word. Also, as the spill slots are not
always initialized, it is possible that some part of a word
contains useful informationwhile the rest contains garbage.
Instead of letting the runtime recreating the ABI0 layout and
print them as words, we now print each component separately.
Aggregate-typed argument/component is surrounded by "{}".
For example, for a function
F(int, [3]byte, byte) int
when called as F(1, [3]byte{2, 3, 4}, 5), it used to print
F(0x1, 0x5040302, 0xXXXXXXXX) // assuming little endian, 0xXXXXXXXX is uninitilized result
Now prints
F(0x1, {0x2, 0x3, 0x4}, 0x5).
Note: the liveness tracking of the spill splots has not been
implemented in this CL. Currently the runtime just assumes all
the slots are live and print them all.
Increase binary sizes by ~1.5%.
old new
hello (println) 1171328 1187712 (+1.4%)
hello (fmt) 1877024 1901600 (+1.3%)
cmd/compile 22326928 22662800 (+1.5%)
cmd/go 13505024 13726208 (+1.6%)
Updates #40724.
Change-Id: I351e0bf497f99bdbb3f91df2fb17e3c2c5c316dc
Reviewed-on: https://go-review.googlesource.com/c/go/+/304470
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
Panic if the slice is too short.
Updates #395
Change-Id: I90f4bff2da5d8f3148ba06d2482084f32b25c29a
Reviewed-on: https://go-review.googlesource.com/c/go/+/301650
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
Apparently, f.Nname.Ntype.Type() doesn't work with types2, as it
doesn't set Ntype, unlike the old type checker. f.Nname.Type()
works for both.
Change-Id: I6fa8a81c9fc7b65e008d9f158b88f0d56d84c3ff
Reviewed-on: https://go-review.googlesource.com/c/go/+/312089
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
The code that does partially live in-register arg spilling is
currently guarded with GOEXPERIMENT=regabiargs. But on platforms
where GOEXPERIMENT=regabiargs is not enabled there are still tests
that use register args. Guard it with actual number of registers
used, so it covers both.
Should fix the freeBSD builder.
Change-Id: I0d3c49d7a2389096cb6b17ca35b9b4ce567bc91e
Reviewed-on: https://go-review.googlesource.com/c/go/+/311830
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>
|
|
former code only spilled those parameters mentioned in code
AT THE REGISTER LEVEL, this caused problems with liveness
sometimes (which worked on whole variables including
aggregates).
Updates #40724.
Change-Id: Ib9fdc50d95d1d2b1f1e405dd370540e88582ac71
Reviewed-on: https://go-review.googlesource.com/c/go/+/310690
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
The go/build package needs access to this configuration,
so move it into a new package available to the standard library.
Change-Id: I868a94148b52350c76116451f4ad9191246adcff
Reviewed-on: https://go-review.googlesource.com/c/go/+/310731
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
|
|
Currently, when we about to emit code that sets the function
results and returns, it emits a VarDef. But in some cases, the
result node is actually live and holding useful data. VarDef
means that we are about to (re)initialize it so all previous
data are dead, but that is not true. Don't insert that.
Also don't add VarDef for register results. We are not going to
store anything (currently it doesn't cause problem, just
unnecessary).
Change-Id: I9dd3b70b4a3f5035af028b143fde8fafa2f11fa0
Reviewed-on: https://go-review.googlesource.com/c/go/+/310589
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>
|
|
In CL 307909 we generate code that spills pointer-typed argument
registers if it is part of an SSA-able aggregate. The current
code spill the register unconditionally. Sometimes it is
unnecessary, because it is already spilled, or it is never live.
This CL reworks the spill generation. We move it to the end of
compilation, after liveness analysis, so we have information about
if a spill is necessary, and only generate spills for the
necessary ones.
Change-Id: I8d60be9b2c47651aeda14f5e2d1bbd207c134b26
Reviewed-on: https://go-review.googlesource.com/c/go/+/309331
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
|
|
Fixes this failure:
go test cmd/compile/internal/ssa -run TestStmtLines -v
=== RUN TestStmtLines
stmtlines_test.go:115: Saw too many (amd64, > 1%) lines without
statement marks, total=88263, nostmt=1930
('-run TestStmtLines -v' lists failing lines)
The failure has two causes.
One is that the first-line adjuster in code generation was relocating
"first lines" to instructions that would either not have any code generated,
or would have the statment marker removed by a different believed-good heuristic.
The other was that statement boundaries were getting attached to register
values (that with the old ABI were loads from the stack, hence real instructions).
The register values disappear at code generation.
The fixes are to (1) note that certain instructions are not good choices for
"first value" and skip them, and (2) in an expandCalls post-pass, look for
register valued instructions and under appropriate conditions move their
statement marker to a compatible use.
Also updates TestStmtLines to always log the score, for easier comparison of
minor compiler changes.
Updates #40724.
Change-Id: I485573ce900e292d7c44574adb7629cdb4695c3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/309649
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|