fc9063e9cd
Replace DEBUG(std::cerr with DOUT. Removed some iostream #includes.
...
llvm-svn: 31811
2006-11-17 07:33:59 +00:00
3216551491
Removed unneeded <iostream> #include.
...
llvm-svn: 31810
2006-11-17 07:10:51 +00:00
21eba2da26
If an indvar with a variable stride is used by the exit condition, go ahead
...
and handle it like constant stride vars. This fixes some bad codegen in
variable stride cases. For example, it compiles this:
void foo(int k, int i) {
for (k=i+i; k <= 8192; k+=i)
flags2[k] = 0;
}
to:
LBB1_1: #bb.preheader
movl %eax, %ecx
addl %ecx, %ecx
movl L_flags2$non_lazy_ptr, %edx
LBB1_2: #bb
movb $0, (%edx,%ecx)
addl %eax, %ecx
cmpl $8192, %ecx
jle LBB1_2 #bb
LBB1_5: #return
ret
or (if the array is local and we are in dynamic-nonpic or static mode):
LBB3_2: #bb
movb $0, _flags2(%ecx)
addl %eax, %ecx
cmpl $8192, %ecx
jle LBB3_2 #bb
and:
lis r2, ha16(L_flags2$non_lazy_ptr)
lwz r2, lo16(L_flags2$non_lazy_ptr)(r2)
slwi r3, r4, 1
LBB1_2: ;bb
li r5, 0
add r6, r4, r3
stbx r5, r2, r3
cmpwi cr0, r6, 8192
bgt cr0, LBB1_5 ;return
instead of:
leal (%eax,%eax,2), %ecx
movl %eax, %edx
addl %edx, %edx
addl L_flags2$non_lazy_ptr, %edx
xorl %esi, %esi
LBB1_2: #bb
movb $0, (%edx,%esi)
movl %eax, %edi
addl %esi, %edi
addl %ecx, %esi
cmpl $8192, %esi
jg LBB1_5 #return
and:
lis r2, ha16(L_flags2$non_lazy_ptr)
lwz r2, lo16(L_flags2$non_lazy_ptr)(r2)
mulli r3, r4, 3
slwi r5, r4, 1
li r6, 0
add r2, r2, r5
LBB1_2: ;bb
li r5, 0
add r7, r3, r6
stbx r5, r2, r6
add r6, r4, r6
cmpwi cr0, r7, 8192
ble cr0, LBB1_2 ;bb
This speeds up Benchmarks/Shootout/sieve from 8.533s to 6.464s and
implements LoopStrengthReduce/var_stride_used_by_compare.ll
llvm-svn: 31809
2006-11-17 06:17:33 +00:00
0727eeaeac
New testcase.
...
llvm-svn: 31808
2006-11-17 06:09:49 +00:00
4b8067f37b
Undo removal of the runtime libraries. While this may have been a bit
...
premature, these libraries will be going away for the 2.0 release. Other
arrangements for profiling, gc, etc. should be made in the next few months.
llvm-svn: 31807
2006-11-17 03:32:33 +00:00
9d46fcd59c
More removal of std::cerr and DEBUG, replacing with DOUT instead.
...
llvm-svn: 31806
2006-11-17 02:09:07 +00:00
3b7261b18e
implement a todo: change a map into a vector
...
llvm-svn: 31805
2006-11-17 01:52:23 +00:00
71adba6dce
Add opcode to TargetInstrDescriptor.
...
llvm-svn: 31804
2006-11-17 01:46:27 +00:00
ca6f6a43e0
#ifndef NDEBUG the "if (Stream) ..." stuff.
...
llvm-svn: 31803
2006-11-17 01:43:48 +00:00
84dd67e08a
Add opcode to TargetInstrDescriptor.
...
llvm-svn: 31802
2006-11-17 01:36:01 +00:00
fc01016c75
Adjusted the // comments so that doxygen picks them up.
...
llvm-svn: 31801
2006-11-17 01:13:12 +00:00
d108b9a60d
Replace std::cerr uses of the "DEBUG" macro with "DOUT" instead. Removes
...
a #include of iostream.
llvm-svn: 31800
2006-11-17 00:50:36 +00:00
be1a4d80b3
fix typo
...
llvm-svn: 31799
2006-11-17 00:49:36 +00:00
d7fda04420
Added "DOUT" macro. This is used as a replacement for the std::cerr
...
stream. It centralizes the use of std::cerr so that static c'tor/d'tors
aren't scattered around all over the place. The way to use it is like this:
DOUT << "This is a status line: " << Var << "\n";
If "-debug" is specified, it will print. Otherwise, it'll not print. If
NDEBUG is defined, the DOUT does nothing.
llvm-svn: 31798
2006-11-17 00:49:12 +00:00
a715288b40
implicit_def_vrrc doesn't generate code.
...
llvm-svn: 31797
2006-11-16 23:49:52 +00:00
538e9cafae
New test case.
...
llvm-svn: 31796
2006-11-16 23:38:21 +00:00
572dc9cb4e
Correct instructions for moving data between GR64 and SSE registers; also correct load i64 / store i64 from v2i64.
...
llvm-svn: 31795
2006-11-16 23:33:25 +00:00
7ae482c52a
Fix a potential bug: MOVPDI2DI, etc. are not copy instructions.
...
llvm-svn: 31794
2006-11-16 23:22:26 +00:00
3f97d03f86
In LLVM 2.0 we won't use the runtime libraries as llvm-gcc3 support will
...
be dropped. This patch pertains to removing the runtime directory from LLVM.
llvm-svn: 31793
2006-11-16 23:17:27 +00:00
48850c10c0
This is a general clean up of the PowerPC ABI. Address several problems and
...
bugs including making sure that the TOS links back to the previous frame,
that the maximum call frame size is not included twice when using frame
pointers, no longer growing the frame on calls, double storing of SP and
a cleaner/faster dynamic alloca.
llvm-svn: 31792
2006-11-16 22:43:37 +00:00
30055b9208
fix a regression that I introduced. stdu should scale the offset by 4
...
before printing it.
llvm-svn: 31791
2006-11-16 21:45:30 +00:00
c5e769710e
Align stubs on 4 byte boundary. This fixes 447.dealII.
...
llvm-svn: 31790
2006-11-16 20:13:34 +00:00
8969ebca44
*** empty log message ***
...
llvm-svn: 31789
2006-11-16 20:11:33 +00:00
e03ca9b0b4
Allow target to specify alignment for function stub.
...
llvm-svn: 31788
2006-11-16 20:04:54 +00:00
9fd73b8a7f
Match MachineCodeEmitter changes.
...
llvm-svn: 31787
2006-11-16 20:04:04 +00:00
b3030c42e0
Use a release version of bugpoint, if found. Include gccld passes.
...
llvm-svn: 31786
2006-11-16 18:32:47 +00:00
96d7386006
add a statistic
...
llvm-svn: 31785
2006-11-16 18:13:49 +00:00
ae2a3097e6
Cleaned up the document. Added "doc_code" divisions for code examples.
...
llvm-svn: 31784
2006-11-16 09:31:19 +00:00
d811cc6818
Don't recompute getNumOperands for each iteration.
...
llvm-svn: 31783
2006-11-16 07:35:18 +00:00
cc9816c87b
Added "removeRange", which takes and removes an entire LiveRange.
...
llvm-svn: 31781
2006-11-16 02:43:32 +00:00
c7969fd1e7
Added a new method "CreateNewLiveInterval" which, given a list of
...
LiveRanges, creates a new LiveInterval from them. The LiveRanges should
have existed already in another LiveInterval, but removed.
llvm-svn: 31780
2006-11-16 02:41:50 +00:00
2ccd16c0af
make this harder
...
llvm-svn: 31779
2006-11-16 01:22:52 +00:00
a7ff5162b0
fix broken encoding
...
llvm-svn: 31778
2006-11-16 01:01:28 +00:00
e0a70388c2
enable the branch folding pass for the JIT.
...
llvm-svn: 31777
2006-11-16 01:00:07 +00:00
e742d9a4b7
add ppc64 r+i stores with update.
...
llvm-svn: 31776
2006-11-16 00:57:19 +00:00
6f5840c409
add patterns for ppc32 preinc stores. ppc64 next.
...
llvm-svn: 31775
2006-11-16 00:41:37 +00:00
3a494989a6
switch these back to the 'bad old way'
...
llvm-svn: 31774
2006-11-16 00:33:34 +00:00
f64da389f8
Fix an incorrectly inverted condition.
...
llvm-svn: 31773
2006-11-16 00:08:20 +00:00
8ae58b5828
ChangeToRegister should clear IsImp bit.
...
llvm-svn: 31772
2006-11-15 23:55:03 +00:00
b4e911d15a
Fix ppc64 epilog bug.
...
llvm-svn: 31771
2006-11-15 23:35:30 +00:00
5771156be0
Stop using isTwoAddress, switching to operand constraints instead.
...
Tell the codegen emitter that specific operands are not to be encoded, fixing
JIT regressions w.r.t. pre-inc loads and stores (e.g. lwzu, which we generate
even when general preinc loads are not enabled).
llvm-svn: 31770
2006-11-15 23:24:18 +00:00
78a403f90e
Remove the isTwoAddress property from the CodeGenInstruction class. It should
...
not be used for anything other than backwards compat constraint handling.
Add support for a new DisableEncoding property which contains a list of
registers that should not be encoded by the generated code emitter. Convert
the codeemitter generator to use this, fixing some PPC JIT regressions.
llvm-svn: 31769
2006-11-15 23:23:02 +00:00
5c23899c7d
add a new field needed by the code emitter generator.
...
llvm-svn: 31768
2006-11-15 22:55:04 +00:00
7cab3de0da
Fix typo pointed out by Bryan Adams
...
llvm-svn: 31766
2006-11-15 21:04:15 +00:00
dc2c8748a7
Properly transfer kill / dead info.
...
llvm-svn: 31765
2006-11-15 20:58:11 +00:00
bbbcac39e6
Kill / dead info has been moved to MI's.
...
llvm-svn: 31764
2006-11-15 20:56:39 +00:00
c57819d4a3
commuteInstruction should propagate kill / dead info.
...
llvm-svn: 31763
2006-11-15 20:56:03 +00:00
0ffff1c629
Match live variable changes.
...
llvm-svn: 31762
2006-11-15 20:55:15 +00:00
ebb0357b4c
Minor updates.
...
llvm-svn: 31761
2006-11-15 20:54:29 +00:00
12ea213de9
A register def can be partially dead when the whole register has use(s) but a subreg does not.
...
llvm-svn: 31760
2006-11-15 20:54:11 +00:00
70ec52896e
Do away with kill / dead maps. Move kill / dead info onto MI's.
...
llvm-svn: 31759
2006-11-15 20:51:59 +00:00
5b536a0be9
Add copyKillDeadInfo to copy kill / dead info; other minor updates.
...
llvm-svn: 31758
2006-11-15 20:48:17 +00:00
474b5b7c95
fix ldu/stu jit encoding. Swith 64-bit preinc load instrs to use memri
...
addrmodes.
llvm-svn: 31757
2006-11-15 19:55:13 +00:00
9f3083eef0
Add run(Function &F) support in FunctionPassManager_New
...
llvm-svn: 31756
2006-11-15 19:39:54 +00:00
2775aba51d
Simplify IntrinsicLowering and clarify that it is only for use by the
...
CBE and interpreter.
llvm-svn: 31755
2006-11-15 18:00:10 +00:00
60c2a0154a
Remove unneeded forward decls
...
llvm-svn: 31754
2006-11-15 17:53:13 +00:00
30d08801ef
remove dead #include
...
llvm-svn: 31753
2006-11-15 17:51:15 +00:00
0e117c7e9d
Fix the PPC regressions last night
...
llvm-svn: 31752
2006-11-15 17:40:51 +00:00
e3a63d136d
Fix a gcc 4.2 warning.
...
llvm-svn: 31751
2006-11-15 04:53:24 +00:00
8f73ba7e05
Add a method to get the bit width of a packed type.
...
llvm-svn: 31750
2006-11-15 03:02:41 +00:00
1396961e85
Switch loads over to use memri as the operand instead of a reg/imm operand
...
pair for cleanliness. Add instructions for PPC32 preinc-stores with commented
out patterns. More improvement is needed to enable the patterns, but we're
getting close.
llvm-svn: 31749
2006-11-15 02:43:19 +00:00
c94f214d22
ADd support for adding constraints to suboperands
...
llvm-svn: 31748
2006-11-15 02:38:17 +00:00
af87fa3a63
Add doInitialization and doFinalization support in FunctionManager_New.
...
llvm-svn: 31747
2006-11-15 02:07:25 +00:00
4a3fa4ff26
Tidy up marking of last analysis user pass.
...
llvm-svn: 31746
2006-11-15 01:48:14 +00:00
ff631ae600
Add doInitialization and doFinalization support in FunctionManager_New.
...
llvm-svn: 31745
2006-11-15 01:27:05 +00:00
42add710fa
Do not derive CommonPassManagerImpl from Pass.
...
Now BasicBlockPassManager_New is a FunctionPass,
FunctionPassManager_New is a ModulePass
llvm-svn: 31744
2006-11-15 01:11:27 +00:00
c75081b7a8
restore some 'magic' code that I removed: it is needed. Add comments explaining
...
why.
llvm-svn: 31743
2006-11-14 22:17:10 +00:00
c5dd5b11a4
minimal hack to get patterns whose result type is iPTR to be selected.
...
llvm-svn: 31742
2006-11-14 21:50:27 +00:00
07f4f58945
Copy initializeAnalysisImpl() implementation from PassManagerT.
...
Update LastUser to recursively walk required transitive set.
llvm-svn: 31741
2006-11-14 21:49:36 +00:00
75165d07a0
remove some dead code
...
llvm-svn: 31740
2006-11-14 21:41:35 +00:00
7c9f48cefb
Add support for nodes that return iPTR.
...
llvm-svn: 31739
2006-11-14 21:32:01 +00:00
110c777f2c
changes to get ptr_rc to be accepted in patterns. This is needed for ppc preinc
...
stores.
llvm-svn: 31738
2006-11-14 21:18:40 +00:00
cf772f7eca
Revert. This wasn't meant to be checked in.
...
llvm-svn: 31737
2006-11-14 19:20:33 +00:00
e79a451475
group load and store instructions together. No functionality change.
...
llvm-svn: 31736
2006-11-14 19:19:53 +00:00
58e8bed48e
Fix predicates for unindexed stores so they don't accidentally match indexed
...
stores.
llvm-svn: 31735
2006-11-14 19:13:39 +00:00
44dbdbe5cf
Rework PPC64 calls. Now we have a LR8/CTR8 register which the PPC64 calls
...
clobber. This allows LR8 to be save/restored correctly as a 64-bit quantity,
instead of handling it as a 32-bit quantity. This unbreaks ppc64 codegen when
the code is actually located above the 4G boundary.
llvm-svn: 31734
2006-11-14 18:44:47 +00:00
b542925b22
remove a ton of custom selection logic no longer needed
...
llvm-svn: 31733
2006-11-14 18:43:11 +00:00
f8c06b335d
Fix a bug handling nodes with variable arguments. The code was fixed to assume
...
that there were two input operands before the variable operand portion. This
*happened* to be true for all call instructions, which took a chain and a
destination, but was not true for the PPC BCTRL instruction, whose destination
is implicit.
Making this code more general allows elimination of the custom selection logic
for BCTRL.
llvm-svn: 31732
2006-11-14 18:41:38 +00:00
4916acbd0f
it would be nice of ctlz were lowered to bsf etc.
...
llvm-svn: 31730
2006-11-14 08:08:46 +00:00
f05d69ae72
implement InstCombine/shift-simplify.ll by transforming:
...
(X >> Z) op (Y >> Z) -> (X op Y) >> Z
for all shifts and all ops={and/or/xor}.
llvm-svn: 31729
2006-11-14 07:46:50 +00:00
c348c50c38
new testcase
...
llvm-svn: 31728
2006-11-14 07:45:36 +00:00
d12a4bf799
implement InstCombine/and-compare.ll:test1. This compiles:
...
typedef struct { unsigned prefix : 4; unsigned code : 4; unsigned unsigned_p : 4; } tree_common;
int foo(tree_common *a, tree_common *b) { return a->code == b->code; }
into:
_foo:
movl 4(%esp), %eax
movl 8(%esp), %ecx
movl (%eax), %eax
xorl (%ecx), %eax
# TRUNCATE movb %al, %al
shrb $4, %al
testb %al, %al
sete %al
movzbl %al, %eax
ret
instead of:
_foo:
movl 8(%esp), %eax
movb (%eax), %al
shrb $4, %al
movl 4(%esp), %ecx
movb (%ecx), %cl
shrb $4, %cl
cmpb %al, %cl
sete %al
movzbl %al, %eax
ret
saving one cycle by eliminating a shift.
llvm-svn: 31727
2006-11-14 06:06:06 +00:00
9adf8294ef
new testcase
...
llvm-svn: 31726
2006-11-14 06:04:35 +00:00
97ff46b3cc
lower "X = seteq Y, Z" to '(shr (ctlz (xor Y, Z)), 5)' instead of
...
'(shr (ctlz (sub Y, Z)), 5)'.
The use of xor better exposes the operation to bit-twiddling logic in the
dag combiner. For example, this:
typedef struct {
unsigned prefix : 4;
unsigned code : 4;
unsigned unsigned_p : 4;
} tree_common;
int foo(tree_common *a, tree_common *b) {
return a->code == b->code;
}
Now compiles to:
_foo:
lwz r2, 0(r4)
lwz r3, 0(r3)
xor r2, r3, r2
rlwinm r2, r2, 28, 28, 31
cntlzw r2, r2
srwi r3, r2, 5
blr
instead of:
_foo:
lbz r2, 3(r4)
lbz r3, 3(r3)
srwi r2, r2, 4
srwi r3, r3, 4
subf r2, r2, r3
cntlzw r2, r2
srwi r3, r2, 5
blr
saving a cycle.
llvm-svn: 31725
2006-11-14 05:28:08 +00:00
0ad1e3845b
A shim over other AA impls to catch incorrect uses
...
llvm-svn: 31724
2006-11-14 05:21:04 +00:00
1b89a7bcc1
Discard code that supported old bytecode formats. This makes the Bytecode
...
Reader code much easier to read and maintain. Backwards compatibility from
version 5 format has been retained. Older formats will produce an error.
llvm-svn: 31723
2006-11-14 04:47:22 +00:00
ca1892699c
Implement removeDeadPasses().
...
llvm-svn: 31722
2006-11-14 03:05:08 +00:00
3f0832a3b5
Keep track of Last user of analysis phase.
...
llvm-svn: 31721
2006-11-14 02:54:23 +00:00
f60b5d9676
Use std::map to map AnalysisID and Pass.
...
llvm-svn: 31720
2006-11-14 01:59:59 +00:00
6dc22332bf
add a note
...
llvm-svn: 31719
2006-11-14 01:57:53 +00:00
683712583a
minor tweaks, reject vector preinc.
...
llvm-svn: 31717
2006-11-14 01:38:31 +00:00
050ec72ae4
Manage analysis passes during run.
...
llvm-svn: 31716
2006-11-14 01:23:29 +00:00
9ce0379da2
Updating configure script for 2.0
...
llvm-svn: 31715
2006-11-14 00:59:52 +00:00
dafa4ddc46
Update comments.
...
llvm-svn: 31713
2006-11-14 00:03:04 +00:00
dbd3d294e6
Matches MachineInstr changes.
...
llvm-svn: 31712
2006-11-13 23:36:35 +00:00
77af6ac5e8
- Let MachineInstr ctors add implicit def and use operands. Other operands
...
will be inserted before these operands. If the opcode changes (by
setOpcode), the implicit operands are updated as well.
- Added IsKill, IsDead fields to MachineOperand in preparation for changes
that move kill / dead info to MachineInstr's.
llvm-svn: 31711
2006-11-13 23:34:06 +00:00
7086844634
Update PassManagerImpl_New::analysisCurrentlyAvailable to check all
...
managed passmanagers.
llvm-svn: 31710
2006-11-13 22:53:19 +00:00
ebba970a6d
Check currently available anlysis in active managers.
...
llvm-svn: 31709
2006-11-13 22:40:09 +00:00
1121e50f7e
teach the g5 hazard recognizer about update loads. This fixes
...
Ptrdist/anagram among others.
llvm-svn: 31708
2006-11-13 20:11:06 +00:00
3d1027193f
For PR950:
...
Changes for new cast instructions that are backwards compatible.
llvm-svn: 31707
2006-11-13 18:00:52 +00:00
Bill Wendling