| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
which are not analyzed with SCEV techniques, which can require
brute-forcing through a large number of instructions. This
fixes a massive compile-time issue on 400.perlbench (in
particular, the loop in MD5Transform).
llvm-svn: 71259
|
|
|
|
|
|
| |
clients to use it.
llvm-svn: 71258
|
|
|
|
| |
llvm-svn: 71256
|
|
|
|
| |
llvm-svn: 71255
|
|
|
|
| |
llvm-svn: 71253
|
|
|
|
|
|
| |
separate function.
llvm-svn: 71252
|
|
|
|
|
|
|
| |
- This eliminates 5/1000 failures on return-types-32, on the current
ABITest config.
llvm-svn: 71250
|
|
|
|
| |
llvm-svn: 71248
|
|
|
|
|
|
| |
This fixes an old FIXME, and is needed by some upcoming changes.
llvm-svn: 71247
|
|
|
|
|
|
| |
Also, tweak default list of bit-fields to try.
llvm-svn: 71246
|
|
|
|
|
|
|
| |
exit block of a loop, if there is exactly one, similar to
getExitingBlock.
llvm-svn: 71245
|
|
|
|
|
|
| |
'autorelease'. This isn't used yet to flag errors.
llvm-svn: 71244
|
|
|
|
|
|
| |
use of dot-syntax expression. This is to match gcc's.
llvm-svn: 71243
|
|
|
|
| |
llvm-svn: 71242
|
|
|
|
| |
llvm-svn: 71241
|
|
|
|
| |
llvm-svn: 71240
|
|
|
|
| |
llvm-svn: 71239
|
|
|
|
| |
llvm-svn: 71238
|
|
|
|
| |
llvm-svn: 71237
|
|
|
|
| |
llvm-svn: 71236
|
|
|
|
| |
llvm-svn: 71235
|
|
|
|
| |
llvm-svn: 71233
|
|
|
|
|
|
| |
need to work a bit to combine things like (x+c1+c2) into x+c3.
llvm-svn: 71232
|
|
|
|
|
|
| |
the various notions of type size used in LLVM.
llvm-svn: 71230
|
|
|
|
| |
llvm-svn: 71227
|
|
|
|
|
|
|
|
| |
targets,
patch by Mikael Lepistö!
llvm-svn: 71226
|
|
|
|
|
|
|
|
|
| |
method, fixing a crash on PR4146. While the store will
ultimately overwrite the "padded size" number of bits in memory,
the stored value may be a subset of this size. This function
only wants to handle the case where all bits are stored.
llvm-svn: 71224
|
|
|
|
|
|
|
|
| |
This used to work, but I broke it when I modified the code to emit the same thing as GCC for message sends to super in classes."
Patch by David Chisnall!
llvm-svn: 71220
|
|
|
|
|
|
|
|
|
| |
of the underlying _N builtin, not the the type of the pointee of the
actual type. This ensures that atomics involving pointers end up
using the correct integer type when they are resolved, avoiding
aborts in codegen.
llvm-svn: 71218
|
|
|
|
| |
llvm-svn: 71216
|
|
|
|
| |
llvm-svn: 71214
|
|
|
|
|
|
| |
incomplete, do not compute its size and return the original region.
llvm-svn: 71213
|
|
|
|
|
|
|
|
|
| |
semantic rules that gcc and icc use. This implements the variadic
and concrete versions as builtins and has sema do the
disambiguation. There are probably a bunch of details to finish up
but this seems like a large monotonic step forward :)
llvm-svn: 71212
|
|
|
|
| |
llvm-svn: 71211
|
|
|
|
|
|
| |
preparation for the day we use null TargetData when no target is specified.
llvm-svn: 71210
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
unconditional branch to the outside of the loop. e.g.
/// A:
/// ...
/// <fallthrough to B>
///
/// B: --> loop header
/// ...
/// jcc <cond> C, [exit]
///
/// C:
/// ...
/// jmp B
///
/// ==>
///
/// A:
/// ...
/// jmp B
///
/// C: --> new loop header
/// ...
/// <fallthough to B>
///
/// B:
/// ...
/// jcc <cond> C, [exit]
llvm-svn: 71209
|
|
|
|
|
|
| |
build under gcc 4.3.
llvm-svn: 71208
|
|
|
|
|
|
| |
No functionality change.
llvm-svn: 71207
|
|
|
|
| |
llvm-svn: 71206
|
|
|
|
| |
llvm-svn: 71205
|
|
|
|
| |
llvm-svn: 71202
|
|
|
|
| |
llvm-svn: 71201
|
|
|
|
|
|
|
|
| |
implicitly be changed to unsigned values in GRSimpleVals.cpp. This can happen
when the comparison involves logic in specialized transfer functions (e.g.,
OSAtomicCompareAndSwap).
llvm-svn: 71200
|
|
|
|
| |
llvm-svn: 71199
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
"This patch is a first pass at adding support for exceptions for the GNU runtime. There are a few limitations at present:
- @synchronized() is not yet supported at all. gcc currently emits calls to runtime library functions that don't exist for this directive.
- Only id @catch statements are currently working. This is enough for NS_DURING and friends, but I need to spend more time reading the output from gcc -S to work out how it finds the class pointer to make arbitrary class type catch statements work.
- I've tested it with a few common cases[1] and the clang test suite (which doesn't test exceptions for the GNU runtime, but shows I haven't broken anything else), but there are probably a lot of cases I've missed."
Patch by David Chisnall!
llvm-svn: 71198
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
scavenger gets confused about register liveness if it doesn't see them.
I'm not thrilled with this solution, but it only comes up when there are dead
copies in the code, which is something that hopefully doesn't happen much.
Here is what happens in pr4100: As shown in the following excerpt from the
debug output of llc, the source of a move gets reloaded from the stack,
inserting a new load instruction before the move. Since that source operand
is a kill, the physical register is free to be reused for the destination
of the move. The move ends up being a no-op, copying R3 to R3, so it is
deleted. But, it leaves behind the load to reload %reg1028 into R3, and
that load is not updated to show that it's destination operand (R3) is dead.
The scavenger gets confused by that load because it thinks that R3 is live.
Starting RegAlloc of: %reg1025<def,dead> = MOVr %reg1028<kill>, 14, %reg0, %reg0
Regs have values:
Reloading %reg1028 into R3
Last use of R3[%reg1028], removing it from live set
Assigning R3 to %reg1025
Register R3 [%reg1025] is never used, removing it from live set
Alternative solutions might be either marking the load as dead, or zapping
the load along with the no-op copy. I couldn't see an easy way to do
either of those, though.
llvm-svn: 71196
|
|
|
|
| |
llvm-svn: 71194
|
|
|
|
|
|
|
| |
other changes. Instead of repeatedly checking for GC mode when
getting the return effect, just do this computation once.
llvm-svn: 71193
|
|
|
|
|
|
| |
- Support generating structures with bit-fields.
llvm-svn: 71192
|
|
|
|
|
|
| |
block pointer. Radar 6441502
llvm-svn: 71190
|