b37fb6a0da
Implement InstCombine/vec_demanded_elts.ll:test2. This allows us to turn
...
unsigned test(float f) {
return _mm_cvtsi128_si32( (__m128i) _mm_set_ss( f*f ));
}
into:
_test:
movss 4(%esp), %xmm0
mulss %xmm0, %xmm0
movd %xmm0, %eax
ret
instead of:
_test:
movss 4(%esp), %xmm0
mulss %xmm0, %xmm0
xorps %xmm1, %xmm1
movss %xmm0, %xmm1
movd %xmm1, %eax
ret
GCC gets:
_test:
subl $28, %esp
movss 32(%esp), %xmm0
mulss %xmm0, %xmm0
xorps %xmm1, %xmm1
movss %xmm0, %xmm1
movaps %xmm1, %xmm0
movd %xmm0, 12(%esp)
movl 12(%esp), %eax
addl $28, %esp
ret
llvm-svn: 36020
2007-04-14 22:29:23 +00:00
a6b5660209
avoid copying sets and vectors around.
...
llvm-svn: 36017
2007-04-14 22:10:17 +00:00
6f58839b20
avoid iterator invalidation.
...
llvm-svn: 36002
2007-04-14 18:06:52 +00:00
4bd0fd367a
An even better fix.
...
llvm-svn: 35998
2007-04-14 17:18:29 +00:00
7233aa9369
Fix recent regression that broke several llvm-tests.
...
llvm-svn: 35996
2007-04-14 16:55:19 +00:00
49fa8d2bff
Implement a few missing xforms: printf("foo\n") -> puts. printf("x") -> putchar
...
printf("") -> noop. Still need to do the xforms for fprintf.
This implements Transforms/SimplifyLibCalls/Printf.ll
llvm-svn: 35984
2007-04-14 01:17:48 +00:00
02137eec8f
in addition to merging, constantmerge should also delete trivially dead globals,
...
in order to clean up after simplifylibcalls.
llvm-svn: 35982
2007-04-14 01:11:54 +00:00
efb33d28c6
Implement PR1201 and test/Transforms/InstCombine/malloc-free-delete.ll
...
llvm-svn: 35981
2007-04-14 00:20:02 +00:00
164b76565b
use an accessor to simplify code.
...
llvm-svn: 35979
2007-04-14 00:17:39 +00:00
efd3051d60
Now that codegen prepare isn't defeating me, I can finally fix what I set
...
out to do! :)
This fixes a problem where LSR would insert a bunch of code into each MBB
that uses a particular subexpression (e.g. IV+base+C). The problem is that
this code cannot be CSE'd back together if inserted into different blocks.
This patch changes LSR to attempt to insert a single copy of this code and
share it, allowing codegenprepare to duplicate the code if it can be sunk
into various addressing modes. On CodeGen/ARM/lsr-code-insertion.ll,
for example, this gives us code like:
add r8, r0, r5
str r6, [r8, #+4]
..
ble LBB1_4 @cond_next
LBB1_3: @cond_true
str r10, [r8, #+4]
LBB1_4: @cond_next
...
LBB1_5: @cond_true55
ldr r6, LCPI1_1
str r6, [r8, #+4]
instead of:
add r10, r0, r6
str r8, [r10, #+4]
...
ble LBB1_4 @cond_next
LBB1_3: @cond_true
add r8, r0, r6
str r10, [r8, #+4]
LBB1_4: @cond_next
...
LBB1_5: @cond_true55
add r8, r0, r6
ldr r10, LCPI1_1
str r10, [r8, #+4]
Besides being smaller and more efficient, this makes it immediately
obvious that it is profitable to predicate LBB1_3 now :)
llvm-svn: 35972
2007-04-13 20:42:26 +00:00
feee64e997
Completely rewrite addressing-mode related sinking of code. In particular,
...
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
llvm-svn: 35970
2007-04-13 20:30:56 +00:00
38705d5494
Remove use of SlowOperationInformer.
...
llvm-svn: 35967
2007-04-13 18:58:18 +00:00
b730fe57bf
Undo previous check-in.
...
llvm-svn: 35966
2007-04-13 18:35:15 +00:00
f929b86140
Hello uses LLVMSupport.a (SlowerOperationInformer)
...
llvm-svn: 35965
2007-04-13 18:28:23 +00:00
749e4668e7
Implement the "thread_local" keyword.
...
llvm-svn: 35950
2007-04-12 18:32:50 +00:00
c78d122a6a
Build Hello by default so it can be used in test cases.
...
llvm-svn: 35922
2007-04-11 21:03:37 +00:00
5ee4d0726a
Fix Transforms/ScalarRepl/union-pointer.ll
...
llvm-svn: 35906
2007-04-11 15:45:25 +00:00
74ff60ff84
Turn stuff like:
...
icmp slt i32 %X, 0 ; <i1>:0 [#uses=1]
sext i1 %0 to i32 ; <i32>:1 [#uses=1]
into:
%X.lobit = ashr i32 %X, 31 ; <i32> [#uses=1]
This implements InstCombine/icmp.ll:test[34]
llvm-svn: 35891
2007-04-11 06:57:46 +00:00
d0f7942e23
Simplify some comparisons to arithmetic, this implements:
...
Transforms/InstCombine/icmp.ll
llvm-svn: 35890
2007-04-11 06:53:04 +00:00
20f2372a7c
canonicalize (x <u 2147483648) -> (x >s -1) and (x >u 2147483647) -> (x <s 0)
...
llvm-svn: 35886
2007-04-11 06:12:58 +00:00
7ddbff090a
fix a miscompilation of:
...
define i32 @test(i32 %X) {
entry:
%Y = and i32 %X, 4 ; <i32> [#uses=1]
icmp eq i32 %Y, 0 ; <i1>:0 [#uses=1]
sext i1 %0 to i32 ; <i32>:1 [#uses=1]
ret i32 %1
}
by moving code out of commonIntCastTransforms into visitZExt. Simplify the
APInt gymnastics in it etc.
llvm-svn: 35885
2007-04-11 05:45:39 +00:00
32104034f8
fix a regression introduced by my last patch.
...
llvm-svn: 35879
2007-04-11 03:27:24 +00:00
daa012d1fb
Simplify SROA conversion to integer in some ways, make it more general in others.
...
We now tolerate small amounts of undefined behavior, better emulating what
would happen if the transaction actually occurred in memory. This fixes
SingleSource/UnitTests/2007-04-10-BitfieldTest.c on PPC, at least until
Devang gets a chance to fix the CFE from doing undefined things with bitfields :)
llvm-svn: 35875
2007-04-11 00:57:54 +00:00
467b69cabb
Strengthen the boundary conditions of this fold, implementing
...
InstCombine/set.ll:test25
llvm-svn: 35852
2007-04-09 23:52:13 +00:00
3c7867935e
Re-constify things that don't break the build. Last patch in this
...
series, I promise.
llvm-svn: 35848
2007-04-09 23:38:18 +00:00
3e9690f987
eliminate the last uses of some TLI methods.
...
llvm-svn: 35844
2007-04-09 23:29:07 +00:00
f1ca1376d3
Unconst-ify stuff that broke the build.
...
llvm-svn: 35843
2007-04-09 23:08:26 +00:00
5917716146
Const-ify some parameters, and some cosmetic cleanups. No functionality
...
change.
llvm-svn: 35842
2007-04-09 22:54:50 +00:00
e0ef5ac6bd
Tabs -> Spaces
...
llvm-svn: 35841
2007-04-09 22:31:43 +00:00
83efbc84f7
Improve some _slow_ behavior introduced in my patches the last few days.
...
llvm-svn: 35839
2007-04-09 22:25:09 +00:00
780c009756
switch LSR to use isLegalAddressingMode instead of other simpler hooks
...
llvm-svn: 35837
2007-04-09 22:20:14 +00:00
bca0d57179
Check _all_ PHINodes.
...
llvm-svn: 35836
2007-04-09 22:20:10 +00:00
8eb8eeada9
Insert new pre-header before new header. Original pre-header may
...
happen to be an entry, in such case, it is not a good idea to
insert new block before entry.
Also fix typo in assertion check.
llvm-svn: 35833
2007-04-09 21:40:43 +00:00
854197884b
Preserve canonical loop form.
...
llvm-svn: 35829
2007-04-09 20:19:46 +00:00
8436cdfda2
Don't link against System or Support library. These things will already
...
be in the opt tool.
llvm-svn: 35827
2007-04-09 19:17:47 +00:00
b9af5747a5
Do not create new pre-header. Reuse original pre-header.
...
llvm-svn: 35825
2007-04-09 19:04:21 +00:00
03d7ae3a74
Simpler for() loops.
...
llvm-svn: 35822
2007-04-09 17:09:13 +00:00
d6ba41e02d
Fix future bug. Of course, Chris spotted this.
...
Handle Argument or Undef as an incoming PHI value.
llvm-svn: 35821
2007-04-09 16:41:46 +00:00
b28a391a8d
More cosmetic changes.
...
llvm-svn: 35820
2007-04-09 16:21:29 +00:00
88bc2c6f82
Only cosmetic changes. Zero functionality Change.
...
llvm-svn: 35819
2007-04-09 16:11:48 +00:00
a87c9f6114
Fix PR1304 and Transforms/InstCombine/2007-04-08-SingleEltVectorCrash.ll
...
llvm-svn: 35792
2007-04-09 01:37:55 +00:00
4ca9cbb170
Eliminate useless insertelement instructions. This implements
...
Transforms/InstCombine/vec_insertelt.ll and fixes PR1286.
We now compile the code from that bug into:
_foo:
movl 4(%esp), %eax
movdqa (%eax), %xmm0
movl 8(%esp), %ecx
psllw (%ecx), %xmm0
movdqa %xmm0, (%eax)
ret
instead of:
_foo:
subl $4, %esp
movl %ebp, (%esp)
movl %esp, %ebp
movl 12(%ebp), %eax
movdqa (%eax), %xmm0
#IMPLICIT_DEF %eax
pinsrw $2, %eax, %xmm0
xorl %ecx, %ecx
pinsrw $3, %ecx, %xmm0
pinsrw $4, %eax, %xmm0
pinsrw $5, %ecx, %xmm0
pinsrw $6, %eax, %xmm0
pinsrw $7, %ecx, %xmm0
movl 8(%ebp), %eax
movdqa (%eax), %xmm1
psllw %xmm0, %xmm1
movdqa %xmm1, (%eax)
movl %ebp, %esp
popl %ebp
ret
woo :)
llvm-svn: 35788
2007-04-09 01:11:16 +00:00
ae39ca037a
Cleanup some from my DomSet-removal changes. Add a new
...
isReachableFromEntry
test to ETForest to factor a common test out of code.
llvm-svn: 35786
2007-04-09 00:52:49 +00:00
aa8ad10c2f
Fix a typo that broke SimplifyLibCalls/SPrintF.ll (pr1315)
...
llvm-svn: 35768
2007-04-08 18:11:26 +00:00
c8d3788f71
reenable this xform, whoops :)
...
llvm-svn: 35765
2007-04-08 08:01:49 +00:00
7621a031d8
Fix regression on Instcombine/apint-or2.ll
...
llvm-svn: 35763
2007-04-08 07:55:22 +00:00
1150df9cc4
Generalize the code that handles (A&B)|(A&C) to work where B/C are not constants.
...
Add a new xform to simplify (A&B)|(~A&C). THis implements InstCombine/or2.ll:test1
llvm-svn: 35760
2007-04-08 07:47:01 +00:00
5717981e5d
implement a fixme: move optimizations for fwrite out of fputs into a new
...
fwrite optimizer.
llvm-svn: 35758
2007-04-08 07:00:35 +00:00
e6c64466c7
Remove DominatorSet usage from LoopSimplify. Patch from Owen Anderson.
...
llvm-svn: 35757
2007-04-08 01:04:30 +00:00
182a945fb5
Significantly simplify the clients of GetConstantStringInfo, by having it
...
just return the string itself.
llvm-svn: 35755
2007-04-07 21:58:02 +00:00
08c0b8b3c8
Fix problems in the sprintf optimizer
...
llvm-svn: 35754
2007-04-07 21:17:51 +00:00
bed184cbcf
Change CastToCStr to take a pointer instead of a reference.
...
Fix some miscompilations in fprintf optimizer.
llvm-svn: 35753
2007-04-07 21:04:50 +00:00
898d698d9f
Fix an off-by-one error that broke Prolangs/deriv2 with llc on x86
...
and Prolangs-C/cdecl
llvm-svn: 35749
2007-04-07 20:19:08 +00:00
f7ebea1b9f
Add DomSet back, and revert the changes to LoopSimplify. Apparently the
...
ETForest updating mechanisms don't work as I thought they did. These changes
will be reapplied once the issue is worked out.
llvm-svn: 35741
2007-04-07 18:23:27 +00:00
d4f51a8ae3
Add support for cast instructions.
...
llvm-svn: 35734
2007-04-07 15:48:32 +00:00
8763ba1b88
Completely purge DomSet. This is the (hopefully) final patch for PR1171.
...
llvm-svn: 35731
2007-04-07 07:17:27 +00:00
706e97049d
Completely purge DomSet from LoopSimplify. This is part of the
...
continuing work on PR1171.
llvm-svn: 35730
2007-04-07 06:56:47 +00:00
d03a646f06
BreakCriticalEdges does still preserve DominatorTree.
...
llvm-svn: 35729
2007-04-07 05:57:09 +00:00
b39d9ca902
Expunge DomSet from BreakCriticalEdges. This is part of the continuing
...
work for PR 1171.
llvm-svn: 35728
2007-04-07 05:49:29 +00:00
f095bf3ac4
Expunge DomSet from CodeExtractor. This is part of the continuing work
...
on PR1171.
llvm-svn: 35726
2007-04-07 05:31:27 +00:00
93f541057b
Support NE inequality in ValueRanges.
...
llvm-svn: 35724
2007-04-07 04:49:12 +00:00
910419596e
Expunge a bunch of uses of DomSet from LoopSimplify. Many more remain.
...
This is the beginning of work for PR1171.
llvm-svn: 35720
2007-04-07 04:37:14 +00:00
3bb6de85d1
Cleanup. Refactor out the applying of value ranges to its own method.
...
llvm-svn: 35719
2007-04-07 03:36:51 +00:00
12d44abe0f
Use TargetData to find the size of a type.
...
llvm-svn: 35718
2007-04-07 03:16:12 +00:00
eeb01b41ef
Strengthen icmp snuggling by doing 'compare-or-equal-to' to 'compare'
...
first and then range testing second.
llvm-svn: 35715
2007-04-07 02:30:14 +00:00
f42389ffe5
Add loop rotation pass.
...
llvm-svn: 35714
2007-04-07 01:25:15 +00:00
0f1509511e
fix a miscompilation in printf optimizer.
...
llvm-svn: 35713
2007-04-07 01:18:36 +00:00
6a36d636e9
trunc to bool no longer compares against zero
...
llvm-svn: 35712
2007-04-07 01:03:46 +00:00
e8829aa9dd
cleanups for strlen optimizer
...
llvm-svn: 35711
2007-04-07 01:02:00 +00:00
485b6415b1
Introduce a new ReplaceCallWith method, which simplifies a lot of code.
...
llvm-svn: 35710
2007-04-07 00:42:32 +00:00
6a6c1f1c30
fixes for strcpy optimizer
...
llvm-svn: 35709
2007-04-07 00:26:18 +00:00
f9ee647e86
Fix bugs in strncmp.
...
llvm-svn: 35708
2007-04-07 00:06:57 +00:00
c9ccc30212
fix 3 miscompilations and several compielr crashes in strcmp optimizer.
...
llvm-svn: 35707
2007-04-07 00:01:51 +00:00
39f0bb9670
Fix several nasty bugs in the strchr optimizer, this fixes
...
SimplifyLibCalls/2007-04-06-strchr-miscompile.ll and PR1307
llvm-svn: 35706
2007-04-06 23:38:55 +00:00
56b7fc7768
clean up strcat optimizer, no functionality change.
...
llvm-svn: 35704
2007-04-06 22:59:33 +00:00
9b2b8abd20
rename getConstantStringLength -> GetConstantStringInfo. Make it return
...
the start index of the array as well as the length. No functionality change.
llvm-svn: 35703
2007-04-06 22:54:17 +00:00
3dbe65f80a
implement Transforms/InstCombine/malloc2.ll and PR1313
...
llvm-svn: 35700
2007-04-06 18:57:34 +00:00
1a9a760318
Fix Transforms/GlobalOpt/2007-04-05-Crash.ll
...
llvm-svn: 35689
2007-04-05 21:09:42 +00:00
108083edff
Use a worklist-driven algorithm instead of a recursive one.
...
llvm-svn: 35680
2007-04-05 01:27:02 +00:00
7c2001d014
Prevent transformConstExprCastCall from generating conversions that assert
...
elsewhere.
llvm-svn: 35668
2007-04-04 19:16:42 +00:00
5a1c750f31
Fix 2007-04-04-BadFoldBitcastIntoMalloc.ll
...
llvm-svn: 35665
2007-04-04 16:58:57 +00:00
f01a47c93c
Fix comment.
...
llvm-svn: 35655
2007-04-04 06:42:45 +00:00
e5bbb3cb1a
Fix a bug I introduced with my patch yesterday which broke Qt (I converted
...
some constant exprs to apints).
Thanks to Anton for tracking down a small testcase that triggered this!
llvm-svn: 35633
2007-04-03 23:29:39 +00:00
a74deafb13
reinstate the previous two patches, with a bugfix :)
...
ldecod now passes.
llvm-svn: 35626
2007-04-03 17:43:25 +00:00
7511fa280d
Reverting back to 1.723. The last two commits broke JM (and possibily others) on ARM.
...
llvm-svn: 35620
2007-04-03 08:11:50 +00:00
81e0707552
split some code out into a helper function
...
llvm-svn: 35615
2007-04-03 05:11:24 +00:00
64c764cebc
Split a whole ton of code out of visitICmpInst into visitICmpInstWithInstAndIntCst.
...
llvm-svn: 35614
2007-04-03 04:46:52 +00:00
8b2ec5f506
Fix PR1253 and xor2.ll:test[01]
...
llvm-svn: 35612
2007-04-03 01:47:41 +00:00
f3197a7d53
allow -1 strides to reuse "1" strides.
...
llvm-svn: 35607
2007-04-02 22:51:58 +00:00
9bc8ab100d
1. Make use of APInt operation instead of using ConstantExpr::getXXX.
...
2. Use cheaper APInt methods.
llvm-svn: 35594
2007-04-02 13:45:30 +00:00
56cda95658
Use uint32_t for bitwidth instead of unsigned.
...
llvm-svn: 35593
2007-04-02 08:20:41 +00:00
28e0e4e11e
Pass the type of the store access, not the type of the store, into the
...
target hook. This allows us to codegen a loop as:
LBB1_1: @cond_next
mov r2, #0
str r2, [r0, +r3, lsl #2 ]
add r3, r3, #1
cmn r3, #1
bne LBB1_1 @cond_next
instead of:
LBB1_1: @cond_next
mov r2, #0
str r2, [r0], #+4
add r3, r3, #1
cmn r3, #1
bne LBB1_1 @cond_next
This looks the same, but has one fewer induction variable (and therefore,
one fewer register) live in the loop.
llvm-svn: 35592
2007-04-02 06:34:44 +00:00
9d5aacee92
Wrap long line
...
llvm-svn: 35588
2007-04-02 05:48:58 +00:00
50490d54f2
use more obvious function name.
...
llvm-svn: 35587
2007-04-02 05:42:22 +00:00
b24acc7bee
simplify (x+c)^signbit as (x+c+signbit), pointed out by PR1288. This implements
...
test/Transforms/InstCombine/xor.ll:test28
llvm-svn: 35584
2007-04-02 05:36:22 +00:00
b7b75145f1
reduce use of std::set
...
llvm-svn: 35576
2007-04-02 01:44:59 +00:00
c3748562bd
Various passes before isel split edges and do other CFG-restructuring changes.
...
isel has its own particular features that it wants in the CFG, in order to
reduce the number of times a constant is computed, etc. Make sure that we
clean up the CFG before doing any other things for isel. Doing so can
dramatically reduce the number of split edges and reduce the number of
places that constants get computed. For example, this shrinks
CodeGen/Generic/phi-immediate-factoring.ll from 44 to 37 instructions on X86,
and from 21 to 17 MBB's in the output. This is primarily a code size win,
not a performance win.
This implements CodeGen/Generic/phi-immediate-factoring.ll and PR1296.
llvm-svn: 35575
2007-04-02 01:35:34 +00:00
8fe3cbe6bd
print the type of an inserted IV in -debug mode.
...
llvm-svn: 35563
2007-04-01 22:21:39 +00:00
c3eeb42809
simplify this code, make it work for ap ints
...
llvm-svn: 35561
2007-04-01 20:57:36 +00:00
150f3bbab2
Avoid unnecessary APInt construction.
...
llvm-svn: 35555
2007-04-01 17:13:37 +00:00
Chris Lattner