At this point in the compiler we haven't assigned Xoffset values for
PAUTO variables anyway, so just immediately store the stack offsets
into Xoffset rather than into a global map.
Change-Id: I61eb471c857c8b145fd0895cbd98fd4e8d3c3365
Reviewed-on: https://go-review.googlesource.com/30081
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
After the removal of the old backend many types are no longer referenced
outside internal/gc. Make these functions private so that tools like
honnef.co/go/unused can spot when they become dead code. In doing so
this CL identified several previously public helpers which are no longer
used, so removes them.
This should be the last of the public functions.
Change-Id: I7e9c4e72f86f391b428b9dddb6f0d516529706c3
Reviewed-on: https://go-review.googlesource.com/29134
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Linker and reflect info generation (reflect.go) relies on formatting
of types (tconv). The fmt.Format based approach introduces extra
allocations, which matter in those cases. Resurrected sconv and tconv
code from commit c85b77c (fmt.go only); and adjusted it slightly.
The formatter-based approach is still used throughout the rest of the
compiler, but reflect.go now uses the tconv method that simply returns
the desired string.
(The timing data below may not be accurate; I've included it only for
comparison with the numbers in issue #16897).
name old time/op new time/op delta
Template 297ms ± 2% 288ms ± 3% -3.12% (p=0.000 n=27+29)
Unicode 155ms ± 5% 150ms ± 5% -3.26% (p=0.000 n=30+30)
GoTypes 1.00s ± 3% 0.95s ± 3% -4.51% (p=0.000 n=28+29)
name old alloc/op new alloc/op delta
Template 46.8MB ± 0% 46.5MB ± 0% -0.65% (p=0.000 n=28+30)
Unicode 37.9MB ± 0% 37.8MB ± 0% -0.24% (p=0.000 n=29+30)
GoTypes 144MB ± 0% 143MB ± 0% -0.68% (p=0.000 n=30+30)
name old allocs/op new allocs/op delta
Template 469k ± 0% 446k ± 0% -5.01% (p=0.000 n=29+30)
Unicode 375k ± 0% 369k ± 0% -1.62% (p=0.000 n=30+28)
GoTypes 1.47M ± 0% 1.37M ± 0% -6.29% (p=0.000 n=30+30)
The code for sconv/tconv in fmt.go now closely match the code from c85b77c
again; except that the functions are now methods. Removing the use of
the bytes.Buffer in tconv and special-caseing interface{} has helped a
small amount as well:
name old time/op new time/op delta
Template 299ms ± 3% 288ms ± 3% -3.83% (p=0.000 n=29+29)
Unicode 156ms ± 5% 150ms ± 5% -3.56% (p=0.000 n=30+30)
GoTypes 960ms ± 2% 954ms ± 3% -0.58% (p=0.037 n=26+29)
name old alloc/op new alloc/op delta
Template 46.6MB ± 0% 46.5MB ± 0% -0.22% (p=0.000 n=30+30)
Unicode 37.8MB ± 0% 37.8MB ± 0% ~ (p=0.075 n=30+30)
GoTypes 143MB ± 0% 143MB ± 0% -0.31% (p=0.000 n=30+30)
name old allocs/op new allocs/op delta
Template 447k ± 0% 446k ± 0% -0.28% (p=0.000 n=30+30)
Unicode 369k ± 0% 369k ± 0% -0.03% (p=0.032 n=30+28)
GoTypes 1.38M ± 0% 1.37M ± 0% -0.35% (p=0.000 n=29+30)
Comparison between c85b77c and now (see issue #16897):
name old time/op new time/op delta
Template 307ms ± 4% 288ms ± 3% -6.24% (p=0.000 n=29+29)
Unicode 164ms ± 4% 150ms ± 5% -8.20% (p=0.000 n=30+30)
GoTypes 1.01s ± 3% 0.95s ± 3% -5.72% (p=0.000 n=30+29)
name old alloc/op new alloc/op delta
Template 46.8MB ± 0% 46.5MB ± 0% -0.66% (p=0.000 n=29+30)
Unicode 37.8MB ± 0% 37.8MB ± 0% -0.13% (p=0.000 n=30+30)
GoTypes 143MB ± 0% 143MB ± 0% -0.11% (p=0.000 n=30+30)
name old allocs/op new allocs/op delta
Template 444k ± 0% 446k ± 0% +0.48% (p=0.000 n=30+30)
Unicode 369k ± 0% 369k ± 0% +0.09% (p=0.000 n=30+28)
GoTypes 1.35M ± 0% 1.37M ± 0% +1.47% (p=0.000 n=30+30)
There's still a small increase (< 1.5%) for GoTypes but pending a complete
rewrite of fmt.go, this seems ok again.
Fixes#16897.
Change-Id: I7e0e56cd1b9f981252eded917f5752259d402354
Reviewed-on: https://go-review.googlesource.com/29087
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
- also consistently use %v instead of %s when we have a (gc) Formatter
- rewrite done automatically using Formats test in -u (update) mode
- manual update of format strings that were not single string constants
- updated fmt.go, fmt_test.go accordingly
- fmt_test: permit "%T" always
Change-Id: I8f0704286aba5704600ad0c4a4484005b79b905d
Reviewed-on: https://go-review.googlesource.com/28954
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
First step towards cleaning up format use. Not yet enabled.
Change-Id: Ia8d76bf02fe05882fffb9d17c9a30dc38d28bf81
Reviewed-on: https://go-review.googlesource.com/28784
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Does not pass toolstash -cmp due to changed export data,
but the cmd/go binary (which doesn't contain export data)
is bit-for-bit identical.
Change-Id: I6b12f9de18cf7da528e9207dccbf8f08c969f142
Reviewed-on: https://go-review.googlesource.com/26753
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Consider a switch statement like:
switch x {
case 1:
// ...
case 2, 3, 4, 5, 6:
// ...
case 5:
// ...
}
Prior to this CL, the generated code treated
2, 3, 4, 5, and 6 independently in a binary search.
With this CL, the generated code checks whether
2 <= x && x <= 6.
walkinrange then optimizes that range check
into a single unsigned comparison.
Experiments suggest that the best min range size
is 2, using binary size as a proxy for optimization.
Binary sizes before/after this CL:
cmd/compile: 14209728 / 14165360
cmd/go: 9543100 / 9539004
Change-Id: If2f7fb97ca80468fa70351ef540866200c4c996c
Reviewed-on: https://go-review.googlesource.com/26770
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
For #15323.
Change-Id: I23192a05ce57012aa2f96909d90d6a33b913766b
Reviewed-on: https://go-review.googlesource.com/28151
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Starting point for uniform use of printer in fmt.go.
It provides a hook to store additional state (and
remove global variables) and should also be more
efficient and cleaner than the mix of string concatenation
and bytes.Buffer use we have now.
Change-Id: I72de14b01850cca32d407a1cb16c894179ea8848
Reviewed-on: https://go-review.googlesource.com/27916
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
The compiler was canonicalizing unnamed types of the form
struct { i int }
across packages, even though an unexported field i should not be
accessible from other packages.
The fix requires both qualifying the field name in the string used by
the compiler to distinguish the type, and ensuring the struct's pkgpath
is set in the rtype version of the data when the type being written is
not part of the localpkg.
Fixes#16616
Change-Id: Ibab160b8b5936dfa47b17dbfd48964a65586785b
Reviewed-on: https://go-review.googlesource.com/27791
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Some FmtSharp export formatting flag support was removed with
commit b4e9f70. Don't panic if FmtSharp is set, just ignore it.
Fixes#16820.
Change-Id: Ie0c3d3774bd55002f6d2781b1212d070f083e6b2
Reviewed-on: https://go-review.googlesource.com/27556
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Switch lowering splits each case expression out
into its own OCASE node.
Change-Id: Ifcb72b99975ed36da8540f6e43343e9aa2058572
Reviewed-on: https://go-review.googlesource.com/26769
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
The main check here is that liveness now crashes if it finds an instruction
using a variable that should be tracked but is not.
Comments and adjustments in nodarg to explain what's going on and
to remove the "-1" argument added a few months ago, plus a sketch
of a future simplification.
The need for n.Orig in the earlier CL seems to have been an intermediate
problem rather than fundamental: the new explanations in nodarg make
clear that nodarg is not causing the problem I thought, and in fact now
using n instead of n.Orig works fine in plive.go.
Change-Id: I3f5cf9f6e4438a6d27abac7d490e7521545cd552
Reviewed-on: https://go-review.googlesource.com/23450
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
As in the elimination of PHEAP|PPARAM in CL 23393,
this is something the front end can trivially take care of
and then not bother the back ends with.
It also eliminates some suspect (and only lightly exercised)
code paths in the back ends.
I don't have a smoking gun for this one but it seems
more clearly correct.
Change-Id: I3b3f5e669b3b81d091ff1e2fb13226a6f14c69d5
Reviewed-on: https://go-review.googlesource.com/23431
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
The liveness computation of parameters generally was never
correct, but forcing all parameters to be live throughout the
function covered up that problem. The new SSA back end is
too clever: even though it currently keeps the parameter values live
throughout the function, it may find optimizations that mean
the current values are not written back to the original parameter
stack slots immediately or ever (for example if a parameter is set
to nil, SSA constant propagation may replace all later uses of the
parameter with a constant nil, eliminating the need to write the nil
value back to the stack slot), so the liveness code must now
track the actual operations on the stack slots, exposing these
problems.
One small problem in the handling of arguments is that nodarg
can return ONAME PPARAM nodes with adjusted offsets, so that
there are actually multiple *Node pointers for the same parameter
in the instruction stream. This might be possible to correct, but
not in this CL. For now, we fix this by using n.Orig instead of n
when considering PPARAM and PPARAMOUT nodes.
The major problem in the handling of arguments is general
confusion in the liveness code about the meaning of PPARAM|PHEAP
and PPARAMOUT|PHEAP nodes, especially as contrasted with PAUTO|PHEAP.
The difference between these two is that when a local variable "moves"
to the heap, it's really just allocated there to start with; in contrast,
when an argument moves to the heap, the actual data has to be copied
there from the stack at the beginning of the function, and when a
result "moves" to the heap the value in the heap has to be copied
back to the stack when the function returns
This general confusion is also present in the SSA back end.
The PHEAP bit worked decently when I first introduced it 7 years ago (!)
in 391425ae. The back end did nothing sophisticated, and in particular
there was no analysis at all: no escape analysis, no liveness analysis,
and certainly no SSA back end. But the complications caused in the
various downstream consumers suggest that this should be a detail
kept mainly in the front end.
This CL therefore eliminates both the PHEAP bit and even the idea of
"heap variables" from the back ends.
First, it replaces the PPARAM|PHEAP, PPARAMOUT|PHEAP, and PAUTO|PHEAP
variable classes with the single PAUTOHEAP, a pseudo-class indicating
a variable maintained on the heap and available by indirecting a
local variable kept on the stack (a plain PAUTO).
Second, walkexpr replaces all references to PAUTOHEAP variables
with indirections of the corresponding PAUTO variable.
The back ends and the liveness code now just see plain indirected
variables. This may actually produce better code, but the real goal
here is to eliminate these little-used and somewhat suspect code
paths in the back end analyses.
The OPARAM node type goes away too.
A followup CL will do the same to PPARAMREF. I'm not sure that
the back ends (SSA in particular) are handling those right either,
and with the framework established in this CL that change is trivial
and the result clearly more correct.
Fixes#15747.
Change-Id: I2770b1ce3cbc93981bfc7166be66a9da12013d74
Reviewed-on: https://go-review.googlesource.com/23393
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Updates #15462
Unexport Jconv, Sconv, Fconv, Hconv, Bconv, and VConv as they are
not referenced outside internal/gc.
Econv was only called by EType.String, so merge it into that method.
Change-Id: Iad9b06078eb513b85a03a43cd9eb9366477643d1
Reviewed-on: https://go-review.googlesource.com/22531
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Updates #15462
Replace all use of oconv(op, FmtSharp) with fmt.Printf("%#v", op).
This removes all the callers of oconv.
Change-Id: Ic3bf22495147f8497c8bada01d681428e2405b0e
Reviewed-on: https://go-review.googlesource.com/22530
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Updates #15462
Automatic refactor with sed -e.
Replace all oconv(op, 0) to string conversion with the raw op value
which fmt's %v verb can print directly.
The remaining oconv(op, FmtSharp) will be replaced with op.GoString and
%#v in the next CL.
Change-Id: I5e2f7ee0bd35caa65c6dd6cb1a866b5e4519e641
Reviewed-on: https://go-review.googlesource.com/22499
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
