We want tuple-reading ops immediately follow tuple-generating op, so
that tuple values will not be spilled/copied.
The mechanism introduced in the previous CL cannot really avoid tuples
interleaving. In this CL we always emit tuple and their selectors together.
Maybe remove the tuple scores if it does not help on performance (todo).
Also let tighten not move tuple-reading ops across blocks.
In the previous CL a special case of regenerating flags with tuple-reading
pseudo-op is added, but it did not cover end-of-block case. This is fixed
in this CL and the condition is generalized.
Progress on SSA backend for ARM. Still not complete.
Updates #15365.
Change-Id: I8980b34e7a64eb98153540e9e19a3782e20406ff
Reviewed-on: https://go-review.googlesource.com/23792
Reviewed-by: David Chase <drchase@google.com>
Introduce dec64 rules to (generically) decompose 64-bit integer on
32-bit architectures. 64-bit integer is composed/decomposed with
Int64Make/Hi/Lo ops, as for complex types.
The idea of dealing with Add64 is the following:
(Add64 (Int64Make xh xl) (Int64Make yh yl))
->
(Int64Make
(Add32withcarry xh yh (Select0 (Add32carry xl yl)))
(Select1 (Add32carry xl yl)))
where Add32carry returns a tuple (flags,uint32). Select0 and Select1
read the first and the second component of the tuple, respectively.
The two Add32carry will be CSE'd.
Similarly for multiplication, Mul32uhilo returns a tuple (hi, lo).
Also add support of KeepAlive, to fix build after merge.
Tests addressed_ssa.go, array_ssa.go, break_ssa.go, chan_ssa.go,
cmp_ssa.go, ctl_ssa.go, map_ssa.go, and string_ssa.go in
cmd/compile/internal/gc/testdata passed.
Progress on SSA for ARM. Still not complete.
Updates #15365.
Change-Id: I7867c76785a456312de5d8398a6b3f7ca5a4f7ec
Reviewed-on: https://go-review.googlesource.com/23213
Reviewed-by: Keith Randall <khr@golang.org>
Fix hardcoded flag register mask in ssa/flagalloc.go by auto-generating
the mask.
Also fix a mistake (in previous CL) about conditional branches.
Progress on SSA backend for ARM. Still not complete. Now "container/ring"
package compiles and tests passed.
Updates #15365.
Change-Id: Id7c8805c30dbb8107baedb485ed0f71f59ed6ea8
Reviewed-on: https://go-review.googlesource.com/23093
Reviewed-by: Keith Randall <khr@golang.org>
Keep track of how many uses each Value has. Each appearance in
Value.Args and in Block.Control counts once.
The number of uses of a value is generically useful to
constrain rewrite rules. For instance, we might want to
prevent merging index operations into loads if the same
index expression is used lots of times.
But I have one use in particular for which the use count is required.
We must make sure we don't combine ops with loads if the load has
more than one use. Otherwise, we may split a single load
into multiple loads and that breaks perceived behavior in
the presence of races. In particular, the load of m.state
in sync/mutex.go:Lock can't be done twice. (I have a separate
CL which triggers the mutex failure. This CL has a test which
demonstrates a similar failure.)
Change-Id: Icaafa479239f48632a069d0c3f624e6ebc6b1f0e
Reviewed-on: https://go-review.googlesource.com/20790
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Todd Neal <todd@tneal.org>
Make sure we do any just-before-return cleanup on all paths out of a
function, including when recovering. Each exit path should include
deferreturn (if there are any defers) and then the exit
code (e.g. copying heap-escaping return values back to the stack).
Introduce a Defer SSA block type which has two outgoing edges - one the
fallthrough edge (the defer was queued successfully) and one which
immediately returns (the defer had a successful recover() call and
normal execution should resume at the return point).
Fixes#14725
Change-Id: Iad035c9fd25ef8b7a74dafbd7461cf04833d981f
Reviewed-on: https://go-review.googlesource.com/20486
Reviewed-by: David Chase <drchase@google.com>
The tree's pretty inconsistent about single space vs double space
after a period in documentation. Make it consistently a single space,
per earlier decisions. This means contributors won't be confused by
misleading precedence.
This CL doesn't use go/doc to parse. It only addresses // comments.
It was generated with:
$ perl -i -npe 's,^(\s*// .+[a-z]\.) +([A-Z]),$1 $2,' $(git grep -l -E '^\s*//(.+\.) +([A-Z])')
$ go test go/doc -update
Change-Id: Iccdb99c37c797ef1f804a94b22ba5ee4b500c4f7
Reviewed-on: https://go-review.googlesource.com/20022
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Dave Day <djd@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Flagalloc was recalculating flags is some situations
when it didn't need to. Fixed by using the same name
for the original flag calculation instruction throughout.
Change-Id: Ic0bf58f728a8d87748434dd25a67b0708755e1f8
Reviewed-on: https://go-review.googlesource.com/19237
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Rename StoreConst to ValAndOff so we can use it for other ops.
Make ValAndOff print nicely.
Add some notes & checks related to my aborted attempt to
implement combined CMP+load ops.
Change-Id: I2f901d12d42bc5a82879af0334806aa184a97e27
Reviewed-on: https://go-review.googlesource.com/18947
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: David Chase <drchase@google.com>
The x86 backend automatically rewrites MOV $0, AX to
XOR AX, AX. That rewrite isn't ok when the flags register
is live across the MOV. Keep track of which moves care
about preserving flags, then disable this rewrite for them.
On x86, Prog.Mark was being used to hold the length of the
instruction. We already store that in Prog.Isize, so no
need to store it in Prog.Mark also. This frees up Prog.Mark
to hold a bitmask on x86 just like all the other architectures.
Update #12405
Change-Id: Ibad8a8f41fc6222bec1e4904221887d3cc3ca029
Reviewed-on: https://go-review.googlesource.com/18861
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Reorder how register & stack allocation is done. We used to allocate
registers, then fix up merge edges, then allocate stack slots. This
lead to lots of unnecessary copies on merge edges:
v2 = LoadReg v1
v3 = StoreReg v2
If v1 and v3 are allocated to the same stack slot, then this code is
unnecessary. But at regalloc time we didn't know the homes of v1 and
v3.
To fix this problem, allocate all the stack slots before fixing up the
merge edges. That way, we know what stack slots values use so we know
what copies are required.
Use a good technique for shuffling values around on merge edges.
Improves performance of the go1 TimeParse benchmark by ~12%
Change-Id: I731f43e4ff1a7e0dc4cd4aa428fcdb97812b86fa
Reviewed-on: https://go-review.googlesource.com/17915
Reviewed-by: David Chase <drchase@google.com>
With the separate flagalloc pass, it should be fine to
allow CSE of control values. The worst that can happen
is that the comparison gets un-CSEd by flagalloc.
Fix bug in flagalloc where flag restores were getting
clobbered by rematerialization during register allocation.
Change-Id: If476cf98b69973e8f1a8eb29441136dd12fab8ad
Reviewed-on: https://go-review.googlesource.com/17760
Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
Spilling/restoring flag values is a pain to do during regalloc.
Instead, allocate the flag register in a separate pass. Regalloc then
operates normally on any flag recomputation instructions.
Change-Id: Ia1c3d9e6eff678861193093c0b48a00f90e4156b
Reviewed-on: https://go-review.googlesource.com/17694
Reviewed-by: David Chase <drchase@google.com>
Cherry Zhang
Keith Randall
Brad Fitzpatrick