optimization level.
This only really affects llc for now because both the llvm-gcc and clang front
ends override the default register allocator. I intend to remove that code later.
llvm-svn: 104904
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.
llvm-svn: 104415
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.
llvm-svn: 104380
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.
llvm-svn: 104321
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.
llvm-svn: 104262
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.
llvm-svn: 104094
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.
llvm-svn: 103946
beneficial cases. See the changes in test/CodeGen/X86/tail-opts.ll and
test/CodeGen/ARM/ifcvt2.ll for details.
The fix is to change HashEndOfMBB to hash at most one instruction,
instead of trying to apply heuristics about when it will be profitable to
consider more than one instruction. The regular tail-merging heuristics
are already prepared to handle the same cases, and they're more precise.
Also, make test/CodeGen/ARM/ifcvt5.ll and
test/CodeGen/Thumb2/thumb2-branch.ll slightly more complex so that they
continue to test what they're intended to test.
And, this eliminates the problem in
test/CodeGen/Thumb2/2009-10-15-ITBlockBranch.ll, the testcase from
PR5204. Update it accordingly.
llvm-svn: 102907
the intrinsics. The reason for those i8* types is that the intrinsics are
overloaded on the vector type and we don't have a way to declare an intrinsic
where one argument is an overloaded vector type and another argument is a
pointer to the vector element type. The bitcasts added here will match what
the frontend will typically generate when these intrinsics are used.
llvm-svn: 101840
instructions to help disassembly.
We also changed the output of the addressing modes to omit the '+' from the
assembler syntax #+/-<imm> or +/-<Rm>. See, for example, A8.6.57/58/60.
And modified test cases to not expect '+' in +reg or #+num. For example,
; CHECK: ldr.w r9, [r7, #28]
llvm-svn: 98745
U test/CodeGen/ARM/tls2.ll
U test/CodeGen/ARM/arm-negative-stride.ll
U test/CodeGen/ARM/2009-10-30.ll
U test/CodeGen/ARM/globals.ll
U test/CodeGen/ARM/str_pre-2.ll
U test/CodeGen/ARM/ldrd.ll
U test/CodeGen/ARM/2009-10-27-double-align.ll
U test/CodeGen/Thumb2/thumb2-strb.ll
U test/CodeGen/Thumb2/ldr-str-imm12.ll
U test/CodeGen/Thumb2/thumb2-strh.ll
U test/CodeGen/Thumb2/thumb2-ldr.ll
U test/CodeGen/Thumb2/thumb2-str_pre.ll
U test/CodeGen/Thumb2/thumb2-str.ll
U test/CodeGen/Thumb2/thumb2-ldrh.ll
U utils/TableGen/TableGen.cpp
U utils/TableGen/DisassemblerEmitter.cpp
D utils/TableGen/RISCDisassemblerEmitter.h
D utils/TableGen/RISCDisassemblerEmitter.cpp
U Makefile.rules
U lib/Target/ARM/ARMInstrNEON.td
U lib/Target/ARM/Makefile
U lib/Target/ARM/AsmPrinter/ARMInstPrinter.cpp
U lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
U lib/Target/ARM/AsmPrinter/ARMInstPrinter.h
D lib/Target/ARM/Disassembler
U lib/Target/ARM/ARMInstrFormats.td
U lib/Target/ARM/ARMAddressingModes.h
U lib/Target/ARM/Thumb2ITBlockPass.cpp
llvm-svn: 98640
(RISCDisassemblerEmitter) which emits the decoder functions for ARM and Thumb,
and the disassembler core which invokes the decoder function and builds up the
MCInst based on the decoded Opcode.
Added sub-formats to the NeonI/NeonXI instructions to further refine the NEONFrm
instructions to help disassembly.
We also changed the output of the addressing modes to omit the '+' from the
assembler syntax #+/-<imm> or +/-<Rm>. See, for example, A8.6.57/58/60.
And modified test cases to not expect '+' in +reg or #+num. For example,
; CHECK: ldr.w r9, [r7, #28]
llvm-svn: 98637
an undef value. This is only going to come up for bugpoint-reduced tests --
correct programs will not access memory at undefined addresses -- so it's not
worth the effort of doing anything more aggressive.
llvm-svn: 97745
greater-than-or-equal SELECT_CCs to NEON vmin/vmax instructions. This is
only allowed when UnsafeFPMath is set or when at least one of the operands
is known to be nonzero.
llvm-svn: 97065
branch in ARM v4 code, since it gets clobbered by the return address before
it is used. Instead of adding a new register class containing all the GPRs
except LR, just use the existing tGPR class.
llvm-svn: 96360
bug fixes, and with improved heuristics for analyzing foreign-loop
addrecs.
This change also flattens IVUsers, eliminating the stride-oriented
groupings, which makes it easier to work with.
llvm-svn: 95975
legalization even when the IR-level optimizer has removed dead phis, such
as when the high half of an i64 value is unused on a 32-bit target.
I had to adjust a few test cases that had dead phis.
This is a partial fix for Radar 7627077.
llvm-svn: 95816
only run for x86 with fastisel. I've found it being very effective in
eliminating some obvious dead code as result of formal parameter lowering
especially when tail call optimization eliminated the need for some of the loads
from fixed frame objects. It also shrinks a number of the tests. A couple of
tests no longer make sense and are now eliminated.
llvm-svn: 95493
Even if they are suported by the core, they can be disabled
(this is just a configuration bit inside some register).
Allow unaligned memops on darwin and conservatively disallow them otherwise.
llvm-svn: 94889
This new version is much more aggressive about doing "full" reduction in
cases where it reduces register pressure, and also more aggressive about
rewriting induction variables to count down (or up) to zero when doing so
reduces register pressure.
It currently uses fairly simplistic algorithms for finding reuse
opportunities, but it introduces a new framework allows it to combine
multiple strategies at once to form hybrid solutions, instead of doing
all full-reduction or all base+index.
llvm-svn: 94061
adding an "i" to the suffix, indicating that the elements are integers, is
accepted but not part of the standard syntax. This helps us pass a few more
of the Neon tests from gcc.
llvm-svn: 93677
different BlockAddress labels, but nothing semantically important.
Add a FIXME that BlockAddress codegen is broken if the LLVM BB has
an empty name (e.g. strip was run).
llvm-svn: 93303
The change in SelectionDAGBuilder is needed to allow using bitcasts to convert
between f64 (the default type for ARM "d" registers) and 64-bit Neon vector
types. Radar 7457110.
llvm-svn: 91649
both source operands. In the canonical form, the 2nd operand is changed to an
undef and the shuffle mask is adjusted to only reference elements from the 1st
operand. Radar 7434842.
llvm-svn: 90417
than doing the same via constpool:
1. Load from constpool costs 3 cycles on A9, movt/movw pair - just 2.
2. Load from constpool might stall up to 300 cycles due to cache miss.
3. Movt/movw does not use load/store unit.
4. Less constpool entries => better compiler performance.
This is only enabled on ELF systems, since darwin does not have needed
relocations (yet).
llvm-svn: 89720
D0<def,dead> = ...
...
= S0<use, kill>
S0<def> = ...
...
D0<def> =
The first D0 def is correctly marked dead, however, livevariables should have
added an implicit def of S0 or we end up with a use without a def.
llvm-svn: 88690
will not accept negative values for these. LLVM's default operand printing
sign extends values, so that valid unsigned values appear as negative
immediates. Print all VMOV immediate operands as hex values to resolve this.
Radar 7372576.
llvm-svn: 86301
the loop preheader. Add instructions which are already in the preheader block that
may be common expressions of those that are hoisted out. These does get a few more
instructions CSE'ed.
llvm-svn: 85799
void f (int a1, int a2, int a3, int a4, int a5,...)
In ARMTargetLowering::LowerFormalArguments if the function has 4 or
more regular arguments we used to set VarArgsFrameIndex using an
offset of 0, which is only correct if the function has exactly 4
regular arguments.
llvm-svn: 85590
physical registers. This is especially critical for the later two since they
start the live interval of a super-register. e.g.
%DO<def> = INSERT_SUBREG %D0<undef>, %S0<kill>, 1
If this instruction is eliminated, the register scavenger will not be happy as
D0 is not defined previously.
This fixes PR5055.
llvm-svn: 82968
- Allocate MachineMemOperands and MachineMemOperand lists in MachineFunctions.
This eliminates MachineInstr's std::list member and allows the data to be
created by isel and live for the remainder of codegen, avoiding a lot of
copying and unnecessary translation. This also shrinks MemSDNode.
- Delete MemOperandSDNode. Introduce MachineSDNode which has dedicated
fields for MachineMemOperands.
- Change MemSDNode to have a MachineMemOperand member instead of its own
fields with the same information. This introduces some redundancy, but
it's more consistent with what MachineInstr will eventually want.
- Ignore alignment when searching for redundant loads for CSE, but remember
the greatest alignment.
Target-specific code which previously used MemOperandSDNodes with generic
SDNodes now use MemIntrinsicSDNodes, with opcodes in a designated range
so that the SelectionDAG framework knows that MachineMemOperand information
is available.
llvm-svn: 82794
For the AAPCS ABI, SP must always be 4-byte aligned, and at any "public
interface" it must be 8-byte aligned. For the older ARM APCS ABI, the stack
alignment is just always 4 bytes. For X86, we currently align SP at
entry to a function (e.g., to 16 bytes for Darwin), but no stack alignment
is needed at other times, such as for a leaf function.
After discussing this with Dan, I decided to go with the approach of adding
a new "TransientStackAlignment" field to TargetFrameInfo. This value
specifies the stack alignment that must be maintained even in between calls.
It defaults to 1 except for ARM, where it is 4. (Some other targets may
also want to set this if they have similar stack requirements. It's not
currently required for PPC because it sets targetHandlesStackFrameRounding
and handles the alignment in target-specific code.) The existing StackAlignment
value specifies the alignment upon entry to a function, which is how we've
been using it anyway.
llvm-svn: 82767
LiveVariables add implicit kills to correctly track partial register kills. This works well enough and is fairly accurate. But coalescer can make it impossible to maintain these markers. e.g.
BL <ga:sss1>, %R0<kill,undef>, %S0<kill>, %R0<imp-def>, %R1<imp-def,dead>, %R2<imp-def,dead>, %R3<imp-def,dead>, %R12<imp-def,dead>, %LR<imp-def,dead>, %D0<imp-def>, ...
...
%reg1031<def> = FLDS <cp#1>, 0, 14, %reg0, Mem:LD4[ConstantPool]
...
%S0<def> = FCPYS %reg1031<kill>, 14, %reg0, %D0<imp-use,kill>
When reg1031 and S0 are coalesced, the copy (FCPYS) will be eliminated the the implicit-kill of D0 is lost. In this case it's possible to move the marker to the FLDS. But in many cases, this is not possible. Suppose
%reg1031<def> = FOO <cp#1>, %D0<imp-def>
...
%S0<def> = FCPYS %reg1031<kill>, 14, %reg0, %D0<imp-use,kill>
When FCPYS goes away, the definition of S0 is the "FOO" instruction. However, transferring the D0 implicit-kill to FOO doesn't work since it is the def of D0 itself. We need to fix this in another time by introducing a "kill" pseudo instruction to track liveness.
Disabling the assertion is not ideal, but machine verifier is doing that job now. It's important to know double-def is not a miscomputation since it means a register should be free but it's not tracked as free. It's a performance issue instead.
llvm-svn: 82677
of the defs are processed.
Also fix a implicit_def propagation bug: a implicit_def of a physical register
should be applied to uses of the sub-registers.
llvm-svn: 82616
variable increment / decrement slighter high priority.
This has major impact on some micro-benchmarks. On MultiSource/Applications
and spec tests, it's a minor win. It also reduce 256.bzip instruction count
by 8%, 55 on 164.gzip on i386 / Darwin.
llvm-svn: 82485
tied to different source registers, the TwoAddressInstructionPass needs to
be smarter. Change it to check before replacing a source register whether
that source register is tied to a different destination register, and if so,
defer handling it until a subsequent iteration.
llvm-svn: 80654
makes an eggregious hack somewhat more palatable. Bringing the LSDA forward
and making it a GV available for reference would be even better, but is
beyond the scope of what I'm looking to solve at this point.
Objective C++ code could generate function names that broke the previous
scheme. This fixes that.
llvm-svn: 80649
This is derived from a patch by Anton Korzh. I modified it to recognize
the VEXT shuffles during legalization and lower them to a target-specific
DAG node.
llvm-svn: 79428
support unaligned mem access only for certain types. (Should it be size
instead?)
ARM v7 supports unaligned access for i16 and i32, some v6 variants support it
as well.
llvm-svn: 79127
It is legal for an inline asm operand to use an earlyclobber register if the
use operand is tied to the earlyclobber operand. The issue is discussed here:
http://gcc.gnu.org/ml/gcc/1999-04n/msg00431.html
We should perhaps let only the machine code verifier worry about these finer
details. EarlyClobber operands are not really interesting to the scavenger.
This fixes PR4528 for the third time.
llvm-svn: 79122
In a naked function, the flag is never set and getPristineRegs() returns an
empty list. That means naked functions are able to clobber callee saved
registers, but that is the whole point of naked functions.
This fixes PR4716.
llvm-svn: 79096
the overloaded vector types allowed floating-point or integer vector elements.
Most of these operations actually depend on the element type, so bitcasting
was not an option.
If you include the vpadd intrinsics that I updated earlier, this gets rid
of 20 intrinsics.
llvm-svn: 78646
Evan Cheng