RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-20 19:42:54 +02:00
Code Issues Projects Releases Wiki Activity
72,381 Commits 4 Branches 5 Tags 704 MiB
b38506e3a7
Commit Graph

47229 Commits

Author SHA1 Message Date
Charles Davis
b38506e3a7 Implement .seh_stackalloc and .seh_pushframe parsing. 
I haven't implemented any of the ones that take registers yet. The problem is
that for x86-64 the streamer methods expect a native x86 register number (note:
%r8-%r15 want 8-15 instead of 0-7; same for %xmm8-%xmm15). I haven't figured
out exactly how I want to do that yet.

llvm-svn: 131899
 2011-05-23 16:43:09 +00:00
Chris Lattner
c64a4d1170 fix a really nasty basicaa mod/ref calculation bug that was causing miscompilation of 
UnitTests/ObjC/messages-2.m with the recent optimizer improvements.

llvm-svn: 131897
 2011-05-23 05:15:43 +00:00
Cameron Zwarich
5a416bda73 Fix <rdar://problem/9476260> by having tail calls always generate 32-bit branches 
in Darwin Thumb2 code. Tail calls are already disabled on Thumb1.

llvm-svn: 131894
 2011-05-23 01:57:17 +00:00
Chris Lattner
bee56202ba rearrange two transforms, since one subsumes the other. Make the shift-exactness 
xform recurse.

llvm-svn: 131888
 2011-05-23 00:32:19 +00:00
Chris Lattner
ec35f49b3e Transform any logical shift of a power of two into an exact/NUW shift when 
in a known-non-zero context.

llvm-svn: 131887
 2011-05-23 00:21:50 +00:00
Chris Lattner
498f516575 use the valuetracking isPowerOfTwo function, which is more powerful than checking 
for a constant directly.  Thanks to Duncan for pointing this out.

llvm-svn: 131885
 2011-05-23 00:09:55 +00:00
Chris Lattner
c277b19e97 Teach valuetracking that byval arguments with a specified alignment are 
aligned.

Teach memcpyopt to not give up all hope when confonted with an underaligned
memcpy feeding an overaligned byval.  If the *source* of the memcpy can be
determined to be adequeately aligned, or if it can be forced to be, we can
eliminate the memcpy.

This addresses PR9794.  We now compile the example into:

define i32 @f(%struct.p* nocapture byval align 8 %q) nounwind ssp {
entry:
  %call = call i32 @g(%struct.p* byval align 8 %q) nounwind
  ret i32 %call
}

in both x86-64 and x86-32 mode.  We still don't get a tailcall though,
because tailcalls apparently can't handle byval.

llvm-svn: 131884
 2011-05-23 00:03:39 +00:00
Chris Lattner
5fd8380a35 add a helper method to get the byval alignment of an argument. 
llvm-svn: 131883
 2011-05-22 23:57:23 +00:00
Chris Lattner
ca9723d80b Eliminate some temporary variables, and don't call getByValTypeAlignment 
when we're just going to throw the result away.  No functionality change.

llvm-svn: 131880
 2011-05-22 23:23:02 +00:00
Chris Lattner
5442c034a8 add a missing alias to make us more bug compatible with gcc, PR9378 
llvm-svn: 131874
 2011-05-22 22:31:57 +00:00
Chris Lattner
78663aa9f5 implement PR9315, constant folding exp2 in terms of pow (since hosts without 
C99 runtimes don't have exp2).

llvm-svn: 131872
 2011-05-22 22:22:35 +00:00
Renato Golin
759db3cbe3 RTABI chapter 4.3.4 specifies __eabi_mem* calls. Specifically, __eabi_memset accepts parameters (ptr, size, value) in a different order than GNU's memset (ptr, value, size), therefore the special lowering in AAPCS mode. Implementation by Evzen Muller. 
llvm-svn: 131868
 2011-05-22 21:41:23 +00:00
Charles Davis
a4187a96c9 Add methods to parse the SEH directives to the COFFAsmParser. Implement some 
of them, particularly the ones that don't take arguments. Also implement
.seh_proc and .seh_handler.

llvm-svn: 131866
 2011-05-22 21:12:15 +00:00
Chris Lattner
3066a7e5ad add a note. 
llvm-svn: 131863
 2011-05-22 18:28:46 +00:00
Chris Lattner
84f101ea45 add some random notes. 
llvm-svn: 131862
 2011-05-22 18:26:48 +00:00
Evan Cheng
a2e8770368 Teach ValueTracking about x86 crc32 intrinsics. 
llvm-svn: 131861
 2011-05-22 18:25:30 +00:00
Chris Lattner
8ed794f599 Carve out a place in instcombine to put transformations which work knowing that their 
result is non-zero.  Implement an example optimization (PR9814), which allows us to
transform:
  A / ((1 << B) >>u 2)
into:
  A >>u (B-2)

which we compile into:

_divu3:                                 ## @divu3
	leal	-2(%rsi), %ecx
	shrl	%cl, %edi
	movl	%edi, %eax
	ret

instead of:

_divu3:                                 ## @divu3
	movb	%sil, %cl
	movl	$1, %esi
	shll	%cl, %esi
	shrl	$2, %esi
	movl	%edi, %eax
	xorl	%edx, %edx
	divl	%esi, %eax
	ret

llvm-svn: 131860
 2011-05-22 18:18:41 +00:00
Johnny Chen
6e8aeaa2f4 Fix Bug 9386 - ARM disassembler failed to disassemble conditional bx 
Modified the patch to .td file supplied by Jyun-Yan You.  Add a test case and
modified ARMDisassemblerCore.cpp a little bit.

llvm-svn: 131859
 2011-05-22 17:51:04 +00:00
Chris Lattner
44516c3ba3 Fix PR9815: I was trying to get out of "generating code and then 
failing to form a memset, then having to delete it" but my approximation
isn't safe for self recurrent loops.  Instead of doign a hack, just
do it the right way.

llvm-svn: 131858
 2011-05-22 17:39:56 +00:00
Frits van Bommel
df58ece78a Add a parameter to ConstantFoldTerminator() that callers can use to ask it to also clean up the condition of any conditional terminator it folds to be unconditional, if that turns the condition into dead code. This just means it calls RecursivelyDeleteTriviallyDeadInstructions() in strategic spots. It defaults to the old behavior. 
I also changed -simplifycfg, -jump-threading and -codegenprepare to use this to produce slightly better code without any extra cleanup passes (AFAICT this was the only place in -simplifycfg where now-dead conditions of replaced terminators weren't being cleaned up). The only other user of this function is -sccp, but I didn't read that thoroughly enough to figure out whether it might be holding pointers to instructions that could be deleted by this.

llvm-svn: 131855
 2011-05-22 16:24:18 +00:00
Chris Lattner
98ce5cd957 fix PR9856, an incorrectly conservative assertion: a global can be 
"stored once" even if its address is compared.

llvm-svn: 131849
 2011-05-22 07:15:13 +00:00
Chris Lattner
6cce3b63ab fix PR9841 by having GVN not process dead loads. This was 
causing it to get into infinite loops when it would widen a 
load (which can necessarily leave around dead loads).

llvm-svn: 131847
 2011-05-22 07:03:34 +00:00
Chris Lattner
75adf06d44 fix a bug for hosts without round, PR8893. 
llvm-svn: 131842
 2011-05-22 06:03:53 +00:00
Chris Lattner
9359dea740 move PR9408 here. 
llvm-svn: 131841
 2011-05-22 05:45:06 +00:00
Nick Lewycky
9f7644d549 Teach the inliner to emit llvm.lifetime.start/end, to scope the local variables 
of the inlinee to the code representing the original function.

llvm-svn: 131838
 2011-05-22 05:22:10 +00:00
Chris Lattner
7eb3367809 switch to using a smallvector to avoid allocations for most normal size instructions. 
llvm-svn: 131837
 2011-05-22 04:53:24 +00:00
Chris Lattner
b58a888c57 tidy some things up. 
llvm-svn: 131836
 2011-05-22 04:52:24 +00:00
Charles Davis
d74119d139 Implement emission of all Win64 exception tables. Make the COFF streamer emit 
these tables.

llvm-svn: 131833
 2011-05-22 04:15:07 +00:00
Charles Davis
c23700bb65 Make the COFF streamer emit unwind info when processing a .seh_handlerdata 
directive.

Implement emission of Win64 EH unwind info.

Pull in <cassert> in MCWin64EH.h so it can use the assert() macro.

llvm-svn: 131832
 2011-05-22 03:01:05 +00:00
Charles Davis
a537e589c9 Don't allow unaligned offsets and sizes in the Win64 EH directives. 
Also, fix threshold for 'Big' register saves.

llvm-svn: 131830
 2011-05-22 00:56:20 +00:00
Chris Lattner
69a52dd280 random comment cleanups. 
llvm-svn: 131829
 2011-05-22 00:50:53 +00:00
Chris Lattner
03d83a7b9b remove StandardPasses, it has been replaced with PassManagerBuilder 
llvm-svn: 131827
 2011-05-22 00:30:45 +00:00
Chris Lattner
c383dd2c87 eliminate dependence on StandardPasses.h. The code generator's pass pipeline 
should eventually convert to PMBuilder, but I don't plan to do this.

llvm-svn: 131819
 2011-05-22 00:13:44 +00:00
Nick Lewycky
0539f44b9f Add CreateLifetimeStart and CreateLifetimeEnd to the IRBuilder, with plans to 
use these soon.

llvm-svn: 131812
 2011-05-21 23:14:36 +00:00
Duncan Sands
c228e6bdea Revert commit 131781, to see if it fixes the x86-64 dragonegg buildbot. 
Original log message:
When BasicAA can determine that two pointers have the same base but
differ by a dynamic offset, return PartialAlias instead of MayAlias.
See the comment in the code for details. This fixes PR9971.

llvm-svn: 131809
 2011-05-21 20:54:46 +00:00
Chris Lattner
777808dcaf add a copy ctor to TargetLibraryInfo. 
llvm-svn: 131806
 2011-05-21 20:09:13 +00:00
Eli Friedman
58beee1262 PR7952: Make isa<> use the same logic as cast<>, so that they both work 
consistently.

llvm-svn: 131803
 2011-05-21 19:13:10 +00:00
Benjamin Kramer
85e86083d5 X86: smulo -> add is now done target-independently in DAGCombiner, remove the patterns. 
llvm-svn: 131801
 2011-05-21 18:32:01 +00:00
Benjamin Kramer
df3070e83d Implement mulo x, 2 -> addo x, x in DAGCombiner. 
llvm-svn: 131800
 2011-05-21 18:31:55 +00:00
Benjamin Kramer
24f75ab769 Revert "InstCombine: Turn mul.with.overflow(X, 2) into the cheaper add.with.overflow(X, X)" 
It's better to do this in codegen, mul.with.overflow(X, 2) is more canonical because it has only one use on "X".

llvm-svn: 131798
 2011-05-21 18:31:42 +00:00
Charles Davis
525855fe48 Don't allow .seh_handler and .seh_handlerdata directives inside chained 
unwind areas.

llvm-svn: 131795
 2011-05-21 17:36:25 +00:00
Charles Davis
7f9c007dce A handler for a function in the Win64 EH scheme can be both an unwind handler 
and an exception handler. Handle that case.

Also, add an 'Emitted' member to the MCWin64EHUnwindInfo struct. It will be
needed later.

llvm-svn: 131791
 2011-05-21 15:57:49 +00:00
Benjamin Kramer
51d1eac4bc InstCombine: Turn mul.with.overflow(X, 2) into the cheaper add.with.overflow(X, X) 
llvm-svn: 131789
 2011-05-21 09:22:06 +00:00
Cameron Zwarich
28ea8de263 Fix PR9978 by adding RIP to GR64_TC so it can be used as an address in PIC code. It 
is already in GR64 for the same reasons. Since it isn't allocatable it can't cause
any problems.

llvm-svn: 131787
 2011-05-21 04:13:49 +00:00
Akira Hatanaka
af8143cdde Use the correct register to access stack frame objects. 
llvm-svn: 131785
 2011-05-21 03:01:03 +00:00
Akira Hatanaka
6605a1e1a6 Insert instructions that copy $sp to or from $fp at the right locations. 
llvm-svn: 131784
 2011-05-21 02:29:26 +00:00
Dan Gohman
048c261e5d When BasicAA can determine that two pointers have the same base but 
differ by a dynamic offset, return PartialAlias instead of MayAlias.
See the comment in the code for details. This fixes PR9971.

llvm-svn: 131781
 2011-05-21 01:05:08 +00:00
Andrew Trick
9f867fd2dc Have Program::Wait return -2 for crashed and timeouts instead of embedding 
info in the error message. Per Dan's request.

llvm-svn: 131780
 2011-05-21 00:56:46 +00:00
Charles Davis
520128d52f Set the "LastFrameInst" field when streaming a .seh_setframe directive. This 
is needed for an upcoming patch.

llvm-svn: 131772
 2011-05-21 00:09:04 +00:00
Cameron Zwarich
00ec3d1f9f Fix PR9962 by properly constraining register classes in RemoveCopyByCommutingDef(). This 
actually fixes most of the VerifyCoalescing failures in test-suite.

llvm-svn: 131768
 2011-05-20 23:25:36 +00:00