There are often many values to clobber.
Allow passing them all in at once.
The goal is increased rule readability.
As a bonus, it shrinks cmd/compile by ~97k, almost half a percent.
Package SSA requires 1.2% less memory to compile.
The single-line changes were make via regex,
and the remaining multi-line clobbers were manually combined.
Passes toolstash-check -all.
Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19
Reviewed-on: https://go-review.googlesource.com/c/go/+/220691
Reviewed-by: Michael Munday <mike.munday@ibm.com>
The generic Greater and Geq ops can always be replaced with the Less and
Leq ops. This CL therefore removes them. This simplifies the compiler since
it reduces the number of operations that need handling in both code and in
rewrite rules. This will be especially true when adding control flow
optimizations such as the integer-in-range optimizations in CL 165998.
Change-Id: If0648b2b19998ac1bddccbf251283f3be4ec3040
Reviewed-on: https://go-review.googlesource.com/c/go/+/220417
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This reverts CL 213477.
Reason for revert: tests are failing on linux-mips*-rtrk builders.
Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0
Reviewed-on: https://go-review.googlesource.com/c/go/+/220897
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
We store 32-bit floating point constants in a 64-bit field, by
converting that 32-bit float to 64-bit float to store it, and convert
it back to use it.
That works for *almost* all floating-point constants. The exception is
signaling NaNs. The round trip described above means we can't represent
a 32-bit signaling NaN, because conversions strip the signaling bit.
To fix this issue, just forbid NaNs as floating-point constants in SSA
form. This shouldn't affect any real-world code, as people seldom
constant-propagate NaNs (except in test code).
Additionally, NaNs are somewhat underspecified (which of the many NaNs
do you get when dividing 0/0?), so when cross-compiling there's a
danger of using the compiler machine's NaN regime for some math, and
the target machine's NaN regime for other math. Better to use the
target machine's NaN regime always.
This has been a bug since 1.10, and there's an easy workaround
(declare a global varaible containing the signaling NaN pattern, and
use that as the argument to math.Float32frombits) so we'll fix it in
1.15.
Fixes#36400
Update #36399
Change-Id: Icf155e743281560eda2eed953d19a829552ccfda
Reviewed-on: https://go-review.googlesource.com/c/go/+/213477
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
We should panic in this situation. Rewriting to a SSA op just leads
to a compiler panic.
Fixes#36259
Change-Id: I6e0bccbed7dd0fdac7ebae76b98a211947947386
Reviewed-on: https://go-review.googlesource.com/c/go/+/212405
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
CL 213703 converted generated rewrite rules for commutative ops
to use loops instead of duplicated code.
However, it loaded args using expressions like
v.Args[i] and v.Args[i^1], which the compiler could
not eliminate bounds for (including with all outstanding
prove CLs).
Also, given a series of separate rewrite rules for the same op,
we generated bounds checks for every rewrite rule, even though
we were repeatedly loading the same set of args.
This change reduces both sets of bounds checks.
Instead of loading v.Args[i] and v.Args[i^1] for commutative loops,
we now preload v.Args[0] and v.Args[1] into local variables,
and then swap them (as needed) in the commutative loop post statement.
And we now load all top level v.Args into local variables
at the beginning of every rewrite rule function.
The second optimization is the more significant,
but the first helps a little, and they play together
nicely from the perspective of generating the code.
This does increase register pressure, but the reduced bounds
checks more than compensate.
Note that the vast majority of rewrite rules evaluated
are not applied, so the prologue is the most important
part of the rewrite rules.
There is one subtle aspect to the new generated code.
Because the top level v.Args are shared across rewrite rules,
and rule evaluation can swap v_0 and v_1, v_0 and v_1
can end up being swapped from one rule to the next.
That is OK, because any time a rule does not get applied,
they will have been swapped exactly twice.
Passes toolstash-check -all.
name old time/op new time/op delta
Template 213ms ± 2% 211ms ± 2% -0.85% (p=0.000 n=92+96)
Unicode 83.5ms ± 2% 83.2ms ± 2% -0.41% (p=0.004 n=95+90)
GoTypes 737ms ± 2% 733ms ± 2% -0.51% (p=0.000 n=91+94)
Compiler 3.45s ± 2% 3.43s ± 2% -0.44% (p=0.000 n=99+100)
SSA 8.54s ± 1% 8.32s ± 2% -2.56% (p=0.000 n=96+99)
Flate 136ms ± 2% 135ms ± 1% -0.47% (p=0.000 n=96+96)
GoParser 169ms ± 1% 168ms ± 1% -0.33% (p=0.000 n=96+93)
Reflect 456ms ± 3% 455ms ± 3% ~ (p=0.261 n=95+94)
Tar 186ms ± 2% 185ms ± 2% -0.48% (p=0.000 n=94+95)
XML 251ms ± 1% 250ms ± 1% -0.51% (p=0.000 n=91+94)
[Geo mean] 424ms 421ms -0.68%
name old user-time/op new user-time/op delta
Template 275ms ± 1% 274ms ± 2% -0.55% (p=0.000 n=95+98)
Unicode 118ms ± 4% 118ms ± 4% ~ (p=0.642 n=98+90)
GoTypes 983ms ± 1% 980ms ± 1% -0.30% (p=0.000 n=93+93)
Compiler 4.56s ± 6% 4.52s ± 6% -0.72% (p=0.003 n=100+100)
SSA 11.4s ± 1% 11.1s ± 1% -2.50% (p=0.000 n=96+97)
Flate 168ms ± 1% 167ms ± 1% -0.49% (p=0.000 n=92+92)
GoParser 204ms ± 1% 204ms ± 2% -0.27% (p=0.003 n=99+96)
Reflect 599ms ± 2% 598ms ± 2% ~ (p=0.116 n=95+92)
Tar 227ms ± 2% 225ms ± 2% -0.57% (p=0.000 n=95+98)
XML 313ms ± 2% 312ms ± 1% -0.37% (p=0.000 n=89+95)
[Geo mean] 547ms 544ms -0.61%
file before after Δ %
compile 21113112 21109016 -4096 -0.019%
total 131704940 131700844 -4096 -0.003%
Change-Id: Id6c39e0367e597c0c75b8a4b1eb14cc3cbd11956
Reviewed-on: https://go-review.googlesource.com/c/go/+/216218
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Prior to this change, we generated additional rules at rulegen time
for all possible combinations of args to commutative ops.
This is simple and works well, but leads to lots of generated rules.
This in turn has increased the size of the compiler,
made it hard to compile package ssa on small machines,
and provided a disincentive to mark some ops as commutative.
This change reworks how we handle commutative ops.
Instead of generating a rule per argument permutation,
we generate a series of nested loops, one for each commutative op.
Each loop tries both possible argument orderings.
I also considered attempting to canonicalize the inputs to the
rewrite rules. However, because either or both arguments might be
nothing more than an identifier, and because there can be arbitrary
conditions to evaluate during matching, I did not see how to proceed.
The duplicate rule detection now sorts arguments to commutative ops,
so that it can detect commutative-only duplicates.
There may be further optimizations to the new generated code.
In particular, we may not be removing as many bounds checks as before;
I have not investigated deeply. If more work here is needed,
we could do it with more hints or with improvements to the prove pass.
This change has almost no impact on the generated code.
It does not pass toolstash-check, however. In a handful of functions,
for reasons I do not understand, there are minor position changes.
For the entire series ending at this change,
there is negligible compiler performance impact.
The compiler binary shrinks by about 15%,
and package ssa shrinks by about 25%.
Package ssa also compiles ~25% faster with ~25% less memory.
Change-Id: Ia2ee9ceae7be08a17342319d4e31b0bb238a2ee4
Reviewed-on: https://go-review.googlesource.com/c/go/+/213703
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
We absorbed Not into most integer comparisons but not into pointer
and floating point equality checks.
The new cases trigger more than 300 times during make.bash.
Change-Id: I77c6b31fcacde10da5470b73fc001a19521ce78d
Reviewed-on: https://go-review.googlesource.com/c/go/+/200618
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
The rules for extracting the interface data word don't leave
the result typed correctly. If I do i.([1]*int)[0], the result
should have type *int, not [1]*int. Using (IData x) for the result
keeps the typing of the original top-level Value.
I don't think this would ever cause a real codegen bug, bug fixing it
at least makes the typing shown in ssa.html more consistent.
Change-Id: I239d821c394e58347639387981b0510d13b2f7b7
Reviewed-on: https://go-review.googlesource.com/c/go/+/204042
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
'go run run.go -all_codegen -v codegen' passes with following adjustments:
ARM64: The BFXIL pattern ((x << lc) >> rc | y & ac) needed adjustment
since ORshiftRL generation fusing '>> rc' and '|' interferes
with matching ((x << lc) >> rc) to generate UBFX. Previously
ORshiftLL was created first using the shifts generated for (y & ac).
S390X: Add rules for abs and copysign to match use of AND instead of SHIFTs.
Updates #33826
Updates #32781
Change-Id: I5a59f6239660d53c029cd22dfb44ddf39f93a56c
Reviewed-on: https://go-review.googlesource.com/c/go/+/196810
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
First, renove unnecessary "// cond:" lines from the generated files.
This shaves off about ~7k lines.
Second, join "if cond { break }" statements via "||", which allows us to
deduplicate a large number of them. This shaves off another ~25k lines.
This change is not for readability or simplicity; but rather, to avoid
unnecessary verbosity that makes the generated files larger. All in all,
git reports that the generated files overall weigh ~200KiB less, or
about 2.7% less.
While at it, add a -trace flag to rulegen.
Updates #33644.
Change-Id: I3fac0290a6066070cc62400bf970a4ae0929470a
Reviewed-on: https://go-review.googlesource.com/c/go/+/196498
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
'go run run.go -all_codegen -v codegen' passes with following adjustments:
ARM64: The BFXIL pattern ((x << lc) >> rc | y & ac) needed adjustment
since ORshiftRL generation fusing '>> rc' and '|' interferes
with matching ((x << lc) >> rc) to generate UBFX. Previously
ORshiftLL was created first using the shifts generated for (y & ac).
S390X: Add rules for abs and copysign to match use of AND instead of SHIFTs.
Updates #33826
Updates #32781
Change-Id: I43227da76b625de03fbc51117162b23b9c678cdb
Reviewed-on: https://go-review.googlesource.com/c/go/+/194297
Run-TryBot: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
Updates #33826
Updates #32781
Change-Id: I862d3587446410c447b9a7265196b57f85358633
Reviewed-on: https://go-review.googlesource.com/c/go/+/191780
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
The existing pointer comparison optimizations
don't include pointer arithmetic. Add them.
These rules trigger a few times in std cmd, while compiling:
time.Duration.String
cmd/go/internal/tlog.NodeHash
crypto/tls.ticketKeyFromBytes (3 times)
crypto/elliptic.(*p256Point).p256ScalarMult (15 times!)
crypto/elliptic.initTable
These weird comparisons occur when using the copy builtin,
which does a pointer comparison between src and dst.
This also happens to fix#32454, by optimizing enough
early on that all values can be eliminated.
Fixes#32454
Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6
Reviewed-on: https://go-review.googlesource.com/c/go/+/180940
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
EqPtr and NeqPtr are marked as commutative,
so the transformations for rules are already
generated by the preceding two lines.
Change-Id: Ibecba5c8e54d9df00c84e1dae7e5d8cb53eeff43
Reviewed-on: https://go-review.googlesource.com/c/go/+/180939
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
The assembly output for x & c == c, where c is power of 2:
MOVQ "".set+8(SP), AX
ANDQ $8, AX
CMPQ AX, $8
SETEQ "".~r2+24(SP)
With optimization using bitset:
MOVQ "".set+8(SP), AX
BTL $3, AX
SETCS "".~r2+24(SP)
output less than 1 instruction.
However, there is no speed improvement:
name old time/op new time/op delta
AllBitSet-8 0.35ns ± 0% 0.35ns ± 0% ~ (all equal)
Fixes#31904
Change-Id: I5dca4e410bf45716ed2145e3473979ec997e35d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/175957
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
First, add cpu and memory profiling flags, as these are useful to see
where rulegen is spending its time. It now takes many seconds to run on
a recent laptop, so we have to keep an eye on what it's doing.
Second, stop writing '_ = var' lines to keep imports and variables used
at all times. Now that rulegen removes all such unused names, they're
unnecessary.
To perform the removal, lean on go/types to first detect what names are
unused. We can configure it to give us all the type-checking errors in a
file, so we can collect all "declared but not used" errors in a single
pass.
We then use astutil.Apply to remove the relevant nodes based on the line
information from each unused error. This allows us to apply the changes
without having to do extra parser+printer roundtrips to plaintext, which
are far too expensive.
We need to do multiple such passes, as removing an unused variable
declaration might then make another declaration unused. Two passes are
enough to clean every file at the moment, so add a limit of three passes
for now to avoid eating cpu uncontrollably by accident.
The resulting performance of the changes above is a ~30% loss across the
table, since go/types is fairly expensive. The numbers were obtained
with 'benchcmd Rulegen go run *.go', which involves compiling rulegen
itself, but that seems reflective of how the program is used.
name old time/op new time/op delta
Rulegen 5.61s ± 0% 7.36s ± 0% +31.17% (p=0.016 n=5+4)
name old user-time/op new user-time/op delta
Rulegen 7.20s ± 1% 9.92s ± 1% +37.76% (p=0.016 n=5+4)
name old sys-time/op new sys-time/op delta
Rulegen 135ms ±19% 169ms ±17% +25.66% (p=0.032 n=5+5)
name old peak-RSS-bytes new peak-RSS-bytes delta
Rulegen 71.0MB ± 2% 85.6MB ± 2% +20.56% (p=0.008 n=5+5)
We can live with a bit more resource usage, but the time/op getting
close to 10s isn't good. To win that back, introduce concurrency in
main.go. This further increases resource usage a bit, but the real time
on this quad-core laptop is greatly reduced. The final benchstat is as
follows:
name old time/op new time/op delta
Rulegen 5.61s ± 0% 3.97s ± 1% -29.26% (p=0.008 n=5+5)
name old user-time/op new user-time/op delta
Rulegen 7.20s ± 1% 13.91s ± 1% +93.09% (p=0.008 n=5+5)
name old sys-time/op new sys-time/op delta
Rulegen 135ms ±19% 269ms ± 9% +99.17% (p=0.008 n=5+5)
name old peak-RSS-bytes new peak-RSS-bytes delta
Rulegen 71.0MB ± 2% 226.3MB ± 1% +218.72% (p=0.008 n=5+5)
It might be possible to reduce the cpu or memory usage in the future,
such as configuring go/types to do less work, or taking shortcuts to
avoid having to run it many times. For now, ~2x cpu and ~4x memory usage
seems like a fair trade for a faster and better rulegen.
Finally, we can remove the old code that tried to remove some unused
variables in a hacky and unmaintainable way.
Change-Id: Iff9e83e3f253babf5a1bd48cc993033b8550cee6
Reviewed-on: https://go-review.googlesource.com/c/go/+/189798
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
"Division by invariant integers using multiplication" paper
by Granlund and Montgomery contains a method for directly computing
divisibility (x%c == 0 for c constant) by means of the modular inverse.
The method is further elaborated in "Hacker's Delight" by Warren Section 10-17
This general rule can compute divisibilty by one multiplication, and add
and a compare for odd divisors and an additional rotate for even divisors.
To apply the divisibility rule, we must take into account
the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first
optimization pass "opt". This complicates the matching as we want to match
only in the cases where the result of (x/c) is not also needed.
So, we must match on the expanded form of (x/c) in the expression x == c*(x/c)
in the "late opt" pass after common subexpresion elimination.
Note, that if there is an intermediate opt pass introduced in the future we
could simplify these rules by delaying the magic division rewrite to "late opt"
and matching directly on (x/c) in the intermediate opt pass.
On amd64, the divisibility check is 30-45% faster.
name old time/op new time/op delta`
DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4)
DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4)
DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5)
DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4)
DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4)
DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5)
DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5)
DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5)
DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4)
DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5)
DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5)
DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5)
Fixes#30282Fixes#15806
Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0
Reviewed-on: https://go-review.googlesource.com/c/go/+/173998
Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
"Division by invariant integers using multiplication" paper
by Granlund and Montgomery contains a method for directly computing
divisibility (x%c == 0 for c constant) by means of the modular inverse.
The method is further elaborated in "Hacker's Delight" by Warren Section 10-17
This general rule can compute divisibilty by one multiplication and a compare
for odd divisors and an additional rotate for even divisors.
To apply the divisibility rule, we must take into account
the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first
optimization pass "opt". This complicates the matching as we want to match
only in the cases where the result of (x/c) is not also available.
So, we must match on the expanded form of (x/c) in the expression x == c*(x/c)
in the "late opt" pass after common subexpresion elimination.
Note, that if there is an intermediate opt pass introduced in the future we
could simplify these rules by delaying the magic division rewrite to "late opt"
and matching directly on (x/c) in the intermediate opt pass.
Additional rules to lower the generic RotateLeft* ops were also applied.
On amd64, the divisibility check is 25-50% faster.
name old time/op new time/op delta
DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5)
DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5)
DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5)
DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5)
DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5)
DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5)
DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5)
DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5)
DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5)
DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5)
DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4)
DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5)
DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5)
DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5)
DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5)
DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5)
DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5)
DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5)
DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5)
DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4)
DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5)
DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5)
DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5)
DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5)
A follow-up change will address the signed division case.
Updates #30282
Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403
Reviewed-on: https://go-review.googlesource.com/c/go/+/168037
Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
For powers of two (c=1<<k), the divisibility check x%c == 0 can be made
just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed
integers. This avoids division fix-ups when just divisibility check is needed.
To apply this rule, we match on the fixed-up version of the division. This is
neccessary because the mod and division rewrite rules are already applied
during the initial opt pass.
The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%:
name old time/op new time/op delta
DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5)
DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5)
DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5)
DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5)
DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5)
DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5)
DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4)
DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5)
DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5)
DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5)
DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5)
DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5)
A follow-up CL will address the general case of x%c == 0 for signed integers.
Updates #15806
Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a
Reviewed-on: https://go-review.googlesource.com/c/go/+/173557
Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This reverts CL 168038 (git 68819fb6d2)
Reason for revert: Doesn't work on 32 bit archs.
Change-Id: Idec9098060dc65bc2f774c5383f0477f8eb63a3d
Reviewed-on: https://go-review.googlesource.com/c/go/+/173442
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
For powers of two (c=1<<k), the divisibility check x%c == 0 can be made
just by checking the trailing zeroes via a mask x&(c-1)==0 even for signed
integers. This avoids division fixups when just divisibility check is needed.
To apply this rule the generic divisibility rule for A%B = A-(A/B*B) is disabled
on the "opt" pass, but this does not affect generated code as this rule is applied
later.
The speed up on amd64 due to elimination of unneccessary fixup code is ~55%:
name old time/op new time/op delta
DivconstI64-4 2.08ns ± 0% 2.07ns ± 0% ~ (p=0.079 n=5+5)
DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.55% (p=0.008 n=5+5)
DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=1.000 n=5+5)
DivconstI32-4 1.53ns ± 0% 1.53ns ± 0% ~ (all equal)
DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 4% -54.75% (p=0.008 n=5+5)
DivconstU32-4 1.78ns ± 1% 1.78ns ± 1% ~ (p=1.000 n=5+5)
DivconstI16-4 1.54ns ± 2% 1.53ns ± 0% ~ (p=0.333 n=5+4)
DivisiblePow2constI16-4 1.78ns ± 0% 0.79ns ± 1% -55.39% (p=0.000 n=4+5)
DivconstU16-4 1.00ns ± 5% 0.99ns ± 1% ~ (p=0.730 n=5+5)
DivconstI8-4 1.54ns ± 0% 1.53ns ± 0% ~ (p=0.714 n=4+5)
DivisiblePow2constI8-4 1.78ns ± 0% 0.80ns ± 0% -55.06% (p=0.000 n=5+4)
DivconstU8-4 0.93ns ± 1% 0.95ns ± 1% +1.72% (p=0.024 n=5+5)
A follow-up CL will address the general case of x%c == 0 for signed integers.
Updates #15806
Change-Id: I0d284863774b1bc8c4ce87443bbaec6103e14ef4
Reviewed-on: https://go-review.googlesource.com/c/go/+/168038
Reviewed-by: Keith Randall <khr@golang.org>
A lot of the naked for loops begin like:
for {
v := b.Control
if v.Op != OpConstBool {
break
}
...
return true
}
Instead, write them out in a more compact and readable way:
for v.Op == OpConstBool {
...
return true
}
This requires the addition of two bytes.Buffer writers, as this helps us
make a decision based on future pieces of generated code. This probably
makes rulegen slightly slower, but that's not noticeable; the code
generation still takes ~3.5s on my laptop, excluding build time.
The "v := b.Control" declaration can be moved to the top of each
function. Even though the rules can modify b.Control when firing, they
also make the function return, so v can't be used again.
While at it, remove three unnecessary lines from the top of each
rewriteBlock func.
In total, this results in ~4k lines removed from the generated code, and
a slight improvement in readability.
Change-Id: I317e4c6a4842c64df506f4513375475fad2aeec5
Reviewed-on: https://go-review.googlesource.com/c/go/+/167399
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This rewrite rule triggers only once, in math/big.quotToFloat64,
as part of converting a uint64 to a float64.
Nevertheless, it is cheap; let's add it.
Change-Id: I3ed4a197a559110fec1bc04b3a8abb4c7fcc2c89
Reviewed-on: https://go-review.googlesource.com/c/go/+/167500
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
We know that a & 31 is non-negative for all a, signed or not.
We can avoid checking that and needing to write out an
unreachable call to panicshift.
Change-Id: I32f32fb2c950d2b2b35ac5c0e99b7b2dbd47f917
Reviewed-on: https://go-review.googlesource.com/c/go/+/167499
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
This reduces the number of extra bounds check hints we need to
insert. For example, rather than producing:
_ = v.Args[2]
x := v.Args[0]
y := v.Args[1]
z := v.Args[2]
We now produce:
z := v.Args[2]
x := v.Args[0]
y := v.Args[1]
This gets rid of about 7000 lines of code from the rewrite rules.
Change-Id: I1291cf0f82e8d035a6d65bce7dee6cedee04cbcd
Reviewed-on: https://go-review.googlesource.com/c/go/+/167397
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The sheer length of the generated rules files makes my
editor and git client unhappy.
This change is a small step towards shortening them.
We recognize a few magic variables during rulegen: b, config, fe, typ.
Of these, only b appears prone to false positives.
By tightening the heuristic and fixing one case in MIPS.rules,
we can make the heuristic enough that it has no failures.
That allows us to remove the hedge assignments to _,
removing 3000 pointless lines of code.
Change-Id: I080cde5db28c8277cb3fd9ddcd829306c9a27785
Reviewed-on: https://go-review.googlesource.com/c/go/+/166979
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Make sure the argument to memmove is of pointer type before we try to
get the element type.
This has been noticed for code that uses unsafe+linkname so it can
call runtime.memmove. Probably not the best thing to allow, but the
code is out there and we'd rather not break it unnecessarily.
Fixes#30061
Change-Id: I334a8453f2e293959fd742044c43fbe93f0b3d31
Reviewed-on: https://go-review.googlesource.com/c/160826
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This makes it easier to track names of function arguments
for debugging purposes.
Change-Id: Ic34856fe0b910005e1c7bc051d769d489a4b158e
Reviewed-on: https://go-review.googlesource.com/c/150098
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
We have an existing optimization that recognizes
memory moves of the form A -> B -> C and converts
them into A -> C, in the hopes that the store to
B will be end up being dead and thus eliminated.
However, when A, B, and C are large types,
the front end sometimes emits VarDef ops for the moves.
This change adds an optimization to match that pattern.
This required changing an old compiler test.
The test assumed that a temporary was required
to deal with a large return value.
With this optimization in place, that temporary
ended up being eliminated.
Triggers 649 times during 'go build -a std cmd'.
Cuts 16k off cmd/go.
name old object-bytes new object-bytes delta
Template 507kB ± 0% 507kB ± 0% -0.15% (p=0.008 n=5+5)
Unicode 225kB ± 0% 225kB ± 0% ~ (all equal)
GoTypes 1.85MB ± 0% 1.85MB ± 0% ~ (all equal)
Flate 328kB ± 0% 328kB ± 0% ~ (all equal)
GoParser 402kB ± 0% 402kB ± 0% -0.00% (p=0.008 n=5+5)
Reflect 1.41MB ± 0% 1.41MB ± 0% -0.20% (p=0.008 n=5+5)
Tar 458kB ± 0% 458kB ± 0% ~ (all equal)
XML 601kB ± 0% 599kB ± 0% -0.21% (p=0.008 n=5+5)
Change-Id: I9b5f25c8663a0b772ad1ee51fa61f74b74d26dd3
Reviewed-on: https://go-review.googlesource.com/c/143479
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
This optimization is not sound if A, B, or C
might overlap with each other.
Thanks to Michael Munday for pointing this
out during the review of CL 143479.
This reduces the number of times this optimization
triggers during make.bash from 386 to 74.
This is unfortunate, but I don't see an obvious way around it,
short of souping up the disjointness analysis.
name old object-bytes new object-bytes delta
Template 507kB ± 0% 507kB ± 0% +0.13% (p=0.008 n=5+5)
Unicode 225kB ± 0% 225kB ± 0% ~ (all equal)
GoTypes 1.85MB ± 0% 1.85MB ± 0% +0.02% (p=0.008 n=5+5)
Flate 328kB ± 0% 328kB ± 0% ~ (all equal)
GoParser 402kB ± 0% 402kB ± 0% ~ (all equal)
Reflect 1.41MB ± 0% 1.41MB ± 0% ~ (all equal)
Tar 457kB ± 0% 458kB ± 0% +0.20% (p=0.008 n=5+5)
XML 600kB ± 0% 601kB ± 0% +0.03% (p=0.008 n=5+5)
Change-Id: Ida408cb627145ba9faf473a78606f050c2f3f51c
Reviewed-on: https://go-review.googlesource.com/c/145208
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
During development and debugging, I often want to
write noteRule(fmt.Sprintf(...)), and end up
manually adding the import to the generated code.
Let's just make it always available instead.
Change-Id: I1e2d47c98ba056e1b5da42e35fb6ad26f1d9cc3d
Reviewed-on: https://go-review.googlesource.com/c/145207
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
The goal of this change is to move work from walk to SSA,
and simplify things along the way.
This is hard to accomplish cleanly with small incremental changes,
so this large commit message aims to provide a roadmap to the diff.
High level description:
Prior to this change, walk was responsible for constructing (most of) the stack for function calls.
ascompatte gathered variadic arguments into a slice.
It also rewrote n.List from a list of arguments to a list of assignments to stack slots.
ascompatte was called multiple times to handle the receiver in a method call.
reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack.
adjustargs then made extra stack space for go/defer args as needed.
Node to SSA construction evaluated all the statements in n.List,
and issued the function call, assuming that the stack was correctly constructed.
Intrinsic calls had to dig around inside n.List to extract the arguments,
since intrinsics don't use the stack to make function calls.
This change moves stack construction to the SSA construction phase.
ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did.
It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries.
It does not, however, make any assignments to stack slots.
Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List.
(It would be better to use Ninit instead of List; future work.)
During SSA construction, after doing all the temporary assignments in n.List,
the function arguments are assigned to stack slots by
constructing the appropriate SSA Value, using (*state).storeArg.
SSA construction also now handles adjustments for go/defer args.
This change also simplifies intrinsic calls, since we no longer need to undo walk's work.
Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely.
Generated code differences:
There were a few optimizations applied along the way, the old way.
f(g()) was rewritten to do a block copy of function results to function arguments.
And reorder1 avoided introducing the final "save the stack" temporary in n.List.
The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed.
SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary.
The exception was when the temporary's type was not SSA-able;
in that case, we got a Move into an autotmp and then an immediate Move onto the stack,
with the autotmp never read or used again.
This change introduces a new rewrite rule to detect such pointless double Moves
and collapse them into a single Move.
This is actually more powerful than the original optimization,
since the original optimization relied on the imprecise Node.HasCall calculation.
The other significant difference in the generated code is that the stack is now constructed
completely in SP-offset order. Prior to this change, the stack was constructed somewhat
haphazardly: first the final argument that Node.HasCall deemed to require a temporary,
then other arguments, then the method receiver, then the defer/go args.
SP-offset is probably a good default order. See future work.
There are a few minor object file size changes as a result of this change.
I investigated some regressions in early versions of this change.
One regression (in archive/tar) was the addition of a single CMPQ instruction,
which would be eliminated were this TODO from flagalloc to be done:
// TODO: Remove original instructions if they are never used.
One regression (in text/template) was an ADDQconstmodify that is now
a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change
in the order in which arguments are written. The argument change
order can also now be luckier, so this appears to be a wash.
All in all, though there will be minor winners and losers,
this change appears to be performance neutral.
Future work:
Move loading the result of function calls to SSA construction; eliminate OINDREGSP.
Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass.
Among other benefits, this would make it easier to transition to a new calling convention.
This would require rethinking the handling of stack conflicts and is non-trivial.
Figure out some clean way to indicate that stack construction Stores/Moves
do not alias each other, so that subsequent passes may do things like
CSE+tighten shared stack setup, do DSE using non-first Stores, etc.
This would allow us to eliminate the minor text/template regression.
Possibly make assignments to stack slots not treated as statements by DWARF.
Compiler benchmarks:
name old time/op new time/op delta
Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48)
Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50)
GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48)
Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50)
SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50)
Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45)
GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49)
Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43)
Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50)
XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49)
[Geo mean] 382ms 378ms -0.89%
name old user-time/op new user-time/op delta
Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48)
Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49)
GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48)
Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46)
SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50)
Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48)
GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46)
Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48)
Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50)
XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48)
[Geo mean] 482ms 476ms -1.23%
name old alloc/op new alloc/op delta
Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5)
Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5)
GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5)
Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5)
SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5)
Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5)
GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5)
Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5)
Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5)
XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5)
[Geo mean] 82.8MB 81.1MB -2.08%
name old allocs/op new allocs/op delta
Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5)
Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5)
GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5)
Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5)
SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5)
Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5)
GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4)
Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5)
Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5)
XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5)
[Geo mean] 815k 804k -1.35%
name old object-bytes new object-bytes delta
Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5)
Unicode 224kB ± 0% 224kB ± 0% ~ (all equal)
GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5)
Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5)
GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5)
Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5)
Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5)
XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5)
Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143
Reviewed-on: https://go-review.googlesource.com/c/114797
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Do []byte(string) conversions more efficiently when the string
is a constant. Instead of calling stringtobyteslice, allocate
just the space we need and encode the initialization directly.
[]byte("foo") rewrites to the following pseudocode:
var s [3]byte // on heap or stack, depending on whether b escapes
s = *(*[3]byte)(&"foo"[0]) // initialize s from the string
b = s[:]
which generates this assembly:
0x001d 00029 (tmp1.go:9) LEAQ type.[3]uint8(SB), AX
0x0024 00036 (tmp1.go:9) MOVQ AX, (SP)
0x0028 00040 (tmp1.go:9) CALL runtime.newobject(SB)
0x002d 00045 (tmp1.go:9) MOVQ 8(SP), AX
0x0032 00050 (tmp1.go:9) MOVBLZX go.string."foo"+2(SB), CX
0x0039 00057 (tmp1.go:9) MOVWLZX go.string."foo"(SB), DX
0x0040 00064 (tmp1.go:9) MOVW DX, (AX)
0x0043 00067 (tmp1.go:9) MOVB CL, 2(AX)
// Then the slice is b = {AX, 3, 3}
The generated code is still not optimal, as it still does load/store
from read-only memory instead of constant stores. Next CL...
Update #26498Fixes#10170
Change-Id: I4b990b19f9a308f60c8f4f148934acffefe0a5bd
Reviewed-on: https://go-review.googlesource.com/c/140698
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
In some optimization rules the type of generated OffPtr was
incorrectly set to the type of the pointee, instead of the
pointer. When the OffPtr value is spilled, this may generate
a spill of the wrong type, e.g. a floating point spill of an
integer (pointer) value. On Wasm, this leads to invalid
bytecode.
Fixes#27961.
Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc
Reviewed-on: https://go-review.googlesource.com/c/139257
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Richard Musiol <neelance@gmail.com>
That optimization is not valid if x == -0.
The test is a bit tricky because 0 == -0. We distinguish
0 from -0 with 1/0 == inf, 1/-0 == -inf.
This has been a bug since CL 24790 in Go 1.8. Probably doesn't
warrant a backport.
Fixes#27718
Note: the optimization x-0 -> x is actually valid.
But it's probably best to take it out, so as to not confuse readers.
Change-Id: I99f16a93b45f7406ec8053c2dc759a13eba035fa
Reviewed-on: https://go-review.googlesource.com/135701
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Use the new custom truncate/extension code when storing or extracting
float32 values from AuxInts to avoid the value being changed by the
host platform's floating point conversion instructions (e.g. sNaN ->
qNaN).
Updates #27516.
Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974
Reviewed-on: https://go-review.googlesource.com/134855
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Signalling NaNs were being converted to quiet NaNs during constant
propagation through integer <-> float store-to-load forwarding.
This occurs because we store float32 constants as float64
values and CPU hardware 'quietens' NaNs during conversion between
the two.
Eventually we want to move to using float32 values to store float32
constants, however this will be a big change since both the compiler
and the assembler expect float64 values. So for now this is a small
change that will fix the immediate issue.
Fixes#27193.
Change-Id: Iac54bd8c13abe26f9396712bc71f9b396f842724
Reviewed-on: https://go-review.googlesource.com/132956
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
A signed right shift before an unsigned right shift by register width-1
(extracts the sign bit) is superflous.
trigger counts during ./make.bash
0 (Rsh8U (Rsh8 x _) 7 ) -> (Rsh8U x 7 )
0 (Rsh16U (Rsh16 x _) 15 ) -> (Rsh16U x 15)
2 (Rsh32U (Rsh32 x _) 31 ) -> (Rsh32U x 31)
251 (Rsh64U (Rsh64 x _) 63 ) -> (Rsh64U x 63)
Changes the instructions generated on AMD64 for x / 2 where
x is a signed integer from:
MOVQ AX, CX
SARQ $63, AX
SHRQ $63, AX
ADDQ CX, AX
SARQ $1, AX
to:
MOVQ AX, CX
SHRQ $63, AX
ADDQ CX, AX
SARQ $1, AX
Change-Id: I86321ae8fc9dc24b8fa9eb80aa5c7299eff8c9dc
Reviewed-on: https://go-review.googlesource.com/115956
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
