Commit Graph

4043 Commits

Author SHA1 Message Date
Devang Patel
80250686d5 Simplify. Eliminate unneeded debug_loc entry.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104785 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-26 23:55:23 +00:00
Devang Patel
394427b014 Update debug info when live-in reg is copied into a vreg.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104732 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-26 20:18:50 +00:00
Dale Johannesen
b09e793bf9 Testcase for 104624/104619/PR7191/8023512.
Reduced from one provided by Duncan Sands, thanks!



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104710 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-26 17:55:45 +00:00
Dale Johannesen
854f30d965 Removing test; Chris thinks it's better to have the
bug go untested than have a testcase this large.  So be it.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104632 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-25 20:40:10 +00:00
Dale Johannesen
86234c30a7 Fix another variant of PR 7191. Also add a testcase
Mon Ping provided; unfortunately bugpoint failed to
reduce it, but I think it's important to have a test for
this in the suite.  8023512.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104624 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-25 18:47:23 +00:00
Bob Wilson
bb7ecb2bf5 Thumb2 RSBS instructions were being printed without the 'S' suffix.
Fix it by changing the T2I_rbin_s_is multiclass to handle the CPSR
output and 'S' suffix in the same way as T2I_bin_s_irs.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104531 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-24 18:44:06 +00:00
Evan Cheng
c7cf10c97e LR is in GPR, not tGPR even in Thumb1 mode.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104518 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-24 18:00:18 +00:00
Evan Cheng
2457f2c661 Implement @llvm.returnaddress. rdar://8015977.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104421 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-22 01:47:14 +00:00
Eric Christopher
1e6d3ac709 This test is darwin only. Make it so(tm).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104418 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-22 00:55:55 +00:00
Bob Wilson
be751cfe9c Recognize more BUILD_VECTORs and VECTOR_SHUFFLEs that can be implemented by
copying VFP subregs.  This exposed a bunch of dead code in the *spill-q.ll
tests, so I tweaked those tests to keep that code from being optimized away.
Radar 7872877.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104415 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-22 00:23:12 +00:00
Eric Christopher
8116ca5134 Add full bss data support for darwin tls variables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104414 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-22 00:10:22 +00:00
Bob Wilson
78f006acdf Change CodeGen/ARM/2009-11-02-NegativeLane.ll to use 16-bit vector elements
so that it will continue to test what it was meant to test when I commit a
separate change for better support of BUILD_VECTOR and VECTOR_SHUFFLE for Neon.
Fix a DAG combiner crash exposed by this test change.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104380 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-21 21:05:32 +00:00
Chris Lattner
a26a8471bd now that fp reg kill insertion stuff happens as a separate
pass after isel instead of being interlaced with it, we can
trust that all the code for a function has been isel'd before
it is run.

The practical impact of this is that we can scan for machine
instr phis instead of doing a fuzzy match on the LLVM BB for
phi nodes.  Doing the fuzzy match required knowing when isel
would produce an fp reg stack phi which was gross.  It was
also wrong in cases where select got lowered to a branch
tree because cmovs aren't available (PR6828).

Just do the scan on machine phis which is simpler, faster
and more correct.  This fixes PR6828.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104333 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-21 18:17:54 +00:00
Jakob Stoklund Olesen
2afb7505c5 Teach VirtRegRewriter to handle spilling in instructions that have multiple
definitions of the virtual register.

This happens when spilling the registers produced by REG_SEQUENCE:

%reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0

The rewriter would spill the register multiple times, dead store elimination
tried to keep up, but ended up cutting the branch it was sitting on.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104321 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-21 16:36:13 +00:00
Dale Johannesen
7d07b48b26 Fix i64->f64 conversion, x86-64, -no-sse. A bit
tricky since there's a 3rd 64-bit type, MMX vectors.
PR 7135.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104308 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-21 00:52:33 +00:00
Evan Cheng
f7d87ee158 Change ARM scheduling default to list-hybrid if the target supports floating point instructions (and is not using soft float).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104307 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-21 00:43:17 +00:00
Dan Gohman
f182b23f8f When canonicalizing icmp operand order to put the loop invariant
operand on the left, the interesting operand is on the right. This
fixes a bug where LSR was failing to recognize ICmpZero uses,
which led it to be unable to reverse the induction variable in the
attached testcase.

Delete test/CodeGen/X86/stack-color-with-reg-2.ll, because its test
is extremely fragile and hard to meaningfully update.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104262 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-20 19:26:52 +00:00
Bob Wilson
63b8845e78 Handle Neon v2f64 and v2i64 vector shuffles as register copies.
This fixes the remaining issue with pr7167.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104257 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-20 18:39:53 +00:00
Dan Gohman
a2086b3483 Teach LSR how to cope better with unrolled loops on targets where
the addressing modes don't make this trivially easy. This allows
it to avoid falling into the less precise heuristics in more
cases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104186 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-19 23:43:12 +00:00
Jakob Stoklund Olesen
3437352887 TwoAddressInstructionPass doesn't really know how to merge live intervals when
lowering REG_SEQUENCE instructions.

Insert copies for REG_SEQUENCE sources not killed to avoid breaking later passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104146 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-19 20:08:00 +00:00
Bob Wilson
29e7e32e08 Testcase to go with 104141.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104142 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-19 18:58:37 +00:00
Evan Cheng
9085f98b32 t2LEApcrel and tLEApcrel are re-materializable. This makes it possible to hoist more loads during machine LICM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104115 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-19 07:28:01 +00:00
Evan Cheng
0a942dbb1e Intrinsics which do a vector compare (results are all zero or all ones) are modeled as icmp / fcmp + sext. This is turned into a vsetcc by dag combine (yes, not a good long term solution). The targets can then isel the vsetcc to the appropriate instruction.
The trouble arises when the result of a vector cmp + sext is then and'ed with all ones. Instcombine will turn it into a vector cmp + zext, dag combiner will miss turning it into a vsetcc and hell breaks loose after that.

Teach dag combine to turn a vector cpm + zest into a vsetcc + and 1. This fixes rdar://7923010.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104094 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-19 01:08:17 +00:00
Jakob Stoklund Olesen
dcf7708ad9 Remember to update VirtRegLastUse when spilling without killing before a call.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104074 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 22:20:09 +00:00
Dan Gohman
e5efbafdac When converting a test to a cmp to fold a load, use the cmp that has an
8-bit immediate field rather than one with a wider immediate field.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104064 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 21:42:03 +00:00
Evan Cheng
28dad2a5ca Sink dag combine's post index load / store code that swap base ptr and index into the target hook. Only the target knows whether the swap is safe. In Thumb2 mode, the offset must be an immediate. rdar://7998649
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104060 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 21:31:17 +00:00
Evan Cheng
27e4840e03 Fix PR7162: Use source register classes and sub-indices to determine the correct register class of the definitions of REG_SEQUENCE.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104050 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 20:03:28 +00:00
Daniel Dunbar
3f40b31256 MC/X86: Implement custom lowering to make sure we match things like
X86::ADC32ri $0, %eax
to
  X86::ADC32i32 $0

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104030 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 17:22:24 +00:00
Evan Cheng
a083988c8a FIX PR7158. SimplifyVBinOp was asserting when it fails to constant fold (op (build_vector), (build_vector)).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104004 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-18 00:03:40 +00:00
Evan Cheng
c6dcce3ba5 Fix PR7175. Insert copies of a REG_SEQUENCE source if it is used by other REG_SEQUENCE instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103994 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 23:24:12 +00:00
Evan Cheng
44bfdd3d78 Fix PR7156. If the sources of a REG_SEQUENCE are all IMPLICIT_DEF's. Replace it with an IMPLICIT_DEF rather than deleting it or else it would be left without a def.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103984 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 22:09:49 +00:00
Evan Cheng
53c779bb3a Careful with reg_sequence coalescing to not to overwrite sub-register indices.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103971 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 20:57:12 +00:00
Evan Cheng
6206124250 Turn on -neon-reg-sequence by default.
Using NEON load / store multiple instructions will no longer create gobs of vmov of D registers!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 19:51:20 +00:00
Jakob Stoklund Olesen
aa4b0159da Avoid allocating the same physreg to multiple virtregs in one instruction.
While that approach works wonders for register pressure, it tends to break
everything.

This should unbreak the arm-linux builder and fix a number of miscompilations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103946 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 17:18:59 +00:00
Jakob Stoklund Olesen
0c9e4f5f3f Only use clairvoyance when defining a register, and then only if it has one use.
This makes allocation independent on the ordering of use-def chains.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103935 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-17 04:50:57 +00:00
Dale Johannesen
88dc976102 Removing as part of previous reversion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103915 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-16 20:19:40 +00:00
Dale Johannesen
f7f5a2760a Revert 103911; it broke a test that expects bitconvert
<1xi64> -> i64 to work in MMX registers on hosts where -no-sse
is the default (not mine).  The right thing is
to accept this and make i64->f64 conversions go through memory,
but I don't have time right now.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103914 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-16 20:19:04 +00:00
Dale Johannesen
f9b2242927 Make x86-64 64-bit bitconvert work when SSE is not available.
(This worked as of about 6 months ago and I didn't track down
exactly what broke it; I think this fix is appropriate.)



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103911 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-16 18:22:38 +00:00
Anton Korobeynikov
ded05e34b6 Add support for thiscall calling convention.
Patch by Charles Davis and Steven Watanabe!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103902 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-16 09:08:45 +00:00
Anton Korobeynikov
a9790d739a Some cheap DAG combine goodness for multiplication with a particular constant.
This can be extended later on to handle more "complex" constants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103881 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-15 18:16:59 +00:00
Evan Cheng
06b666c705 Allow TargetLowering::getRegClassFor() to be called on illegal types. Also
allow target to override it in order to map register classes to illegal
but synthesizable types. e.g. v4i64, v8i64 for ARM / NEON.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103854 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-15 02:18:07 +00:00
Bill Wendling
23ead99283 SystemZ really does mean "has calls" and not just "adjusts stack." Go ahead and
replace the check with the appropriate predicate. Modify the testcase to reflect
the correct code. (It should be saving callee-saved registers on the stack
allocated by the calling fuction.)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103829 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-14 22:17:42 +00:00
Jakob Stoklund Olesen
4ed1082683 Simplify the handling of physreg defs and uses in RegAllocFast.
This adds extra security against using clobbered physregs, and it adds kill
markers to physreg uses.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103784 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-14 18:03:25 +00:00
Jakob Stoklund Olesen
4bf4bafcce Take allocation hints from copy instructions to/from physregs.
This causes way more identity copies to be generated, ripe for coalescing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103686 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-13 00:19:43 +00:00
Jakob Stoklund Olesen
804291e316 Make sure to add kill flags to the last use of a virtreg when it is redefined.
The X86 floating point stack pass and others depend on good kill flags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103635 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-12 18:46:03 +00:00
Jakob Stoklund Olesen
85f4fdaed4 Enable a bunch more -regalloc=fast tests
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103531 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-12 00:11:24 +00:00
Jakob Stoklund Olesen
76b4d5a021 Keep track of the last place a live virtreg was used.
This allows us to add accurate kill markers, something the scavenger likes.
Add some more tests from ARM that needed this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103521 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-11 23:24:45 +00:00
Jakob Stoklund Olesen
a44c1223cd One more -regalloc=fast test
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103509 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-11 20:51:07 +00:00
Jakob Stoklund Olesen
82b07dc499 Simplify the tracking of used physregs to a bulk bitor followed by a transitive
closure after allocating all blocks.

Add a few more test cases for -regalloc=fast.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103500 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-11 20:30:28 +00:00
Jakob Stoklund Olesen
bbf33b38aa Mostly rewrite RegAllocFast.
Sorry for the big change. The path leading up to this patch had some TableGen
changes that I didn't want to commit before I knew they were useful. They
weren't, and this version does not need them.

The fast register allocator now does no liveness calculations. Instead it relies
on kill flags provided by isel. (Currently those kill flags are also ignored due
to isel bugs). The allocation algorithm is supposed to work with any subset of
valid kill flags. More kill flags simply means fewer spills inserted.

Registers are allocated from a working set that contains no aliases. That means
most allocations can be done directly without expensive alias checks. When the
working set runs out of registers we do the full alias check to find new free
registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103488 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-11 18:54:45 +00:00