llgo supports the application of LLVM attributes to global objects
and functions. This "feature" is undocumented and untested. As
discusses in D27442, it should be removed.
Differential Revision: https://reviews.llvm.org/D27474
llvm-svn: 288843
Synthesized functions do not have valid locations,
and so we were not setting a debug location. Recent
changes to DI require that function call instructions
within a function having DI must have a location set.
We just set the debug location to line=0, col=0.
Reviewers: pcc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D22905
llvm-svn: 288772
A recent commit (r286087) to the LLVM Go bindings that
changed things over to use the new attribute API broke
llgo. This commit updates llgo accordingly.
llvm-svn: 288769
Switch gofrontend to using go.googlesource.com, and
update to 81eb6a3f425b2158c67ee32c0cc973a72ce9d6be.
There are various changes required to update to the
go 1.5 runtime:
typemap.go is changed to accommodate the change in representation for equal/hash algorithms, and the removal of the zero value/type.
CMakeLists.txt is updated to add the build tree to the package search path, so internal packages, which are not installed, are found.
various files changes due to removal of __go_new_nopointers; the same change as in D11863, but with NoUnwindAttribute added to the added runtime functions which are called with "callOnly".
minor cleanups in ssa.go while investigating issues with unwinding/panic handling.
Differential Revisision: http://reviews.llvm.org/D15188
llvm-svn: 263536
Summary:
Another attempt at resolving the runtime assertion
in llgoi due to data layout mismatch between module
and execution engine.
The X86 data layout constant appears to be unnecessary,
and does not match what the execution engine picks.
Using the registered Target, we pick the same data
layout as the execution engine.
While I was in the vicinity, I deleted the last
remnants of PNaCl support.
Reviewers: pcc
Subscribers: jfb, llvm-commits, dschuff
Differential Revision: http://reviews.llvm.org/D12630
llvm-svn: 248565
Summary:
If a function requires a landing pad, set the personality function.
Requires D11116.
Reviewers: pcc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11117
llvm-svn: 242290
This takes us to Go 1.4. Also includes a couple of changes to the test
suite, both in the runtime package:
- Disable TestSetPanicOnFault. We cannot support this scenario at all,
due to LLVM's lack of non-call exceptions.
- Tweak TestFinalizerType. This test only passes with two GC runs.
Differential Revision: http://reviews.llvm.org/D8828
llvm-svn: 234134
Most importantly, this gives us https://go-review.googlesource.com/7533 which
fixes a bug in go/ssa that caused test failures in the Go 1.4 standard library.
Also remove the go1.4 tag workaround. We no longer need it now that we
ship llgo-go.
Differential Revision: http://reviews.llvm.org/D8827
llvm-svn: 234133
go/loader creates a fresh package map for each source package it imports. In
llgoi this caused binary imported packages to be imported anew for every input
line, resulting in spurious type errors and panics in go/ssa when encountering
previously imported types. Fix this by setting types.Config.Packages to our
internal package map.
Differential Revision: http://reviews.llvm.org/D8409
llvm-svn: 232617
Summary:
The debug metadata we generate is wrong, and is
now causing build failures. This revision disables
the only llvm.dbg.declare calls we make.
(There is also a drive-by fix to CMakeLists.txt,
adding in a missing .go dependency.)
Fixes http://llvm.org/bugs/show_bug.cgi?id=22330
Reviewers: pcc
Reviewed By: pcc
Subscribers: dblaikie, llvm-commits
Differential Revision: http://reviews.llvm.org/D7222
llvm-svn: 227403
At the same time, perform a number of simplifications:
- Rename go.tools directory to gotools.
- Import only the go directory; all required Go analysis code and
its dependencies have now been moved to this directory.
llvm-svn: 225825
Summary:
With this patch, llgo uses ssautil.Switches
to reconstitute (and synthesise) switches,
which can then be lowered to lookup tables,
trees, etc.
We currently only handle integer const case
switches. We erase the comparison blocks (other
than the initial block), and generate a switch
instruction at the end of the block starting
the if-else-if chain. ssautil.Switches does
not remove duplicate const cases (e.g. same
operands for "||"), so we do this in llgo for
now.
Test Plan: lit test added
Reviewers: pcc
Reviewed By: pcc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6831
llvm-svn: 225433
Summary: If a receive case in a select statement is not assigned to a named variable, then we can eliminate the alloca and copy at runtime.
Test Plan: lit test added
Reviewers: pcc
Reviewed By: pcc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6785
llvm-svn: 225033
The new ABI is simpler for use cases such as dynamically loaded packages.
The calling convention for import functions is similar to what go/ssa would
produce if BareInits were cleared. However, simply clearing this flag causes
two additional issues:
1) We would need to special case the 'init$guard' variable (see
discussion in https://codereview.appspot.com/78780043/).
2) The call to __go_register_gc_roots needs to appear in the right
place, i.e. after the guard check. Making this check appear
in the right place with non-bare inits seems unreliable at best.
So we keep BareInits set and generate the necessary code manually.
It is still possible to get the old ABI by specifying a path to a gccgo
installation.
Differential Revision: http://reviews.llvm.org/D6804
llvm-svn: 225030
This is useful for clients that need to use llgo's mangling of the package
path to look up a specific function within a given package.
Differential Revision: http://reviews.llvm.org/D6801
llvm-svn: 225027
This change allows clients to generate IR using "files" received from locations
other than the file system. The regular file parser is moved to a new library,
"driver", which is intended to eventually contain much of the logic from
the existing driver.
Differential Revision: http://reviews.llvm.org/D6794
llvm-svn: 225026
Patch by Andrew Wilkins!
canAvoidElementLoad and canAvoidLoad were incorrectly
eliding loads when an index expression is used as an
another array index expression. This led to a panic.
See comments on https://github.com/go-llvm/llgo/issues/175
Test Plan: lit test added
Differential Revision: http://reviews.llvm.org/D6676
llvm-svn: 224420
If we use the receiver's package, we can end up with identical manglings
for different functions. Consider:
package p
type U struct{}
func (U) f()
package q
import "p"
type T struct { p.U }
func (T) f()
The method set of *T has two synthetic methods named (*T).f(); one forwards to
(T).f(), and the other to (U).f(). Previously, we were distinguishing them
by the receiver's package, and in this case because both methods have the
same receiver, they received the same name.
The methods are correctly distinguished by the package owning the identifier
"f", which is available via f.Object().Pkg().
Differential Revision: http://reviews.llvm.org/D6673
llvm-svn: 224357
Peter Collingbourne