Linking modules containing aliases to GEPs is still not quite right. GEPs that are equivalent to bitcasts will be replaced by bitcasts, GEPs that are not will just break. Aliases to GEPs that are not equivalent to bitcasts are horribly broken anyway (it might be worth adding an assert when creating the alias to prevent these being created; they just cause problems later).
llvm-svn: 93052
std::vectors.
- MacroBuilder wraps a raw_ostream so it can easily write to any buffer
supported by raw_ostream.
- MacroBuilder's method take Twines for easy string concatenation (this was done
with sprintf and temporary buffers before).
- Targets still use std::vector as they don't have access to the builder.
llvm-svn: 93051
really does need to be a vector type, because
TargetLowering::getOperationAction for SIGN_EXTEND_INREG uses that type,
and it needs to be able to distinguish between vectors and scalars.
Also, fix some more issues with legalization of vector casts.
llvm-svn: 93043
when the default case is winnowed down to be infeasible. When all
cases were ruled out (and the analysis state for the default case
would be infeasible) we would still consider the default case
possible. This fixes PR 5969.
llvm-svn: 93017
putting relocations into the constant pool - this isn't needed
for correctness and in the rare occasion it happens would pull
us out of fast isel for the block.
If fast-isel application startup time ever becomes an issue we
can add better support for these addresses instead of bailing.
llvm-svn: 92995
not just the viable ones. This is reasonable because the most common use of
deleted functions is to exclude some implicit conversion during calls; users
therefore will want to figure out why some other options were excluded.
Started sorting overload results. Right now it just sorts by location in the
translation unit (after putting viable functions first), but we can do better than
that.
Changed bool OnlyViable parameter to PrintOverloadCandidates to an enum for better
self-documentation.
llvm-svn: 92990
1. CMPXCHG8B and CMPXCHG16B did not specify implicit physical register defs and uses.
2. LCMPXCHG8B is loading 64 bit memory, not 32 bit.
llvm-svn: 92985
suggestions follow recovery. Additionally, add a note to these
diagnostics which suggests a fix-it for changing the behavior to what
the user probably meant. Examples:
t.cpp:2:9: warning: & has lower precedence than ==; == will be evaluated first
[-Wparentheses]
if (i & j == k) {
^~~~~~~~
( )
t.cpp:2:9: note: place parentheses around the & expression to evaluate it first
if (i & j == k) {
^
( )
t.cpp:14:9: warning: using the result of an assignment as a condition
without
parentheses [-Wparentheses]
if (i = f()) {
~~^~~~~
( )
t.cpp:14:9: note: use '==' to turn this assignment into an equality
comparison
if (i = f()) {
^
==
llvm-svn: 92975
Compare the dominance information calculated using a dominance tree walk to the
information calculated based on DFS numbers, if XDEBUG is enabled.
llvm-svn: 92969
Remove a FIXME and unify code that was necessary to work around broken
updateDFSNumbers(). Before updateDFSNumbers() did not work correctly for post
dominators.
llvm-svn: 92968
The DFS number calculation for postdominators was broken. In the case of
multiple exits that form the post dominator root nodes, do not iterate over
all exits, but start from the virtual root node. Otherwise bbs, that are not
post dominated by any exit but by the virtual root node, will never be assigned
a DFS number.
llvm-svn: 92967
Dan Gohman