Commits

Show all
Author Commit Message Labels Comments Date
Evan Cheng
Honor user-defined section specification of a global, ignores whether its initializer is null.
Tags
2 tags
Dale Johannesen
Fix PR 1688. See comments there.
Devang Patel
Update aux. info associated with an instruction before erasing instruction.
Devang Patel
Do not override user specified section.
Devang Patel
Don't increment invalid iterator.
Duncan Sands
Testcase for PR1678.
Evan Cheng
Disable if-conversion for this test.
Evan Cheng
-enable-arm-if-conversion is gone.
Dale Johannesen
Fix dumb regression in constant folding (Regression/C/casts)
Gordon Henriksen
Regenerate.
Gordon Henriksen
Incorporating review feedback for GC verifier patch.
Dan Gohman
Fix several more entries in the x86 reload/remat folding tables.
Gabor Greif
use typenames equivalent to (u)intval, because latter are not present in older caml/mlvalues.h (e.g. 2004/07/07, 1.48.6.1) Using this as a workaround for now, until --without-ocaml works or we settle on a better solution
Dale Johannesen
another long double buglet
Nick Lewycky
Excuse me.
Nick Lewycky
Fix optimization. %x = sub %x, %y does not imply that %y is zero.
Evan Cheng
Enable if-conversion for ARM by default.
Dale Johannesen
More long double fixes. x86_64 should build now.
Evan Cheng
Avoid referencing deleted instruction.
Ted Kremenek
Previously GraphWriter assumed operator*() for node_iterator always returned a reference type. This patch allows operator*() to return a non-reference type while still maintaining the old behavior when it does return a reference type. This patch was motivated when I tried to use "df_iterator" (see llvm/ADT/DepthFirstIterator.h) as a "node_iterator", as df_iterator does not return a reference type and thus we would get a compilation error when trying to take the address of a temporary.
Devang Patel
Avoid unsafe promotion.
Evan Cheng
PSHUFDmi, etc. are actually folding a load, not a store.
Dale Johannesen
Fix longdouble -> uint conversion.
Chris Lattner
reject things like "declare internal @foo"
Owen Anderson
Add a flag to mark a dirty cache entry. This is not yet used, but will eventually help non-local memdep caching.
Dale Johannesen
Fix some long double issues.
Duncan Sands
Improve comment.
Duncan Sands
Partial fix for PR1678: correct some parts of constant fold that were missed in the fix for PR1646. Probably this null/not-null logic should be factorized somewhere.
Duncan Sands
A global variable with external weak linkage can be null, while an alias could alias such a global variable.
Gabor Greif
include alloca.h if available. this helps Solaris, but intnat and uintnat types are still undefined, causing errors
  1. Prev
  2. Next