TargetInstrInfo:
Change produceSameValue() to take MachineRegisterInfo as an optional argument.
When in SSA form, targets can use it to make more aggressive equality analysis.
Machine LICM:
1. Eliminate isLoadFromConstantMemory, use MI.isInvariantLoad instead.
2. Fix a bug which prevent CSE of instructions which are not re-materializable.
3. Use improved form of produceSameValue.
ARM:
1. Teach ARM produceSameValue to look pass some PIC labels.
2. Look for operands from different loads of different constant pool entries
which have same values.
3. Re-implement PIC GA materialization using movw + movt. Combine the pair with
a "add pc" or "ldr [pc]" to form pseudo instructions. This makes it possible
to re-materialize the instruction, allow machine LICM to hoist the set of
instructions out of the loop and make it possible to CSE them. It's a bit
hacky, but it significantly improve code quality.
4. Some minor bug fixes as well.
With the fixes, using movw + movt to materialize GAs significantly outperform the
load from constantpool method. 186.crafty and 255.vortex improved > 20%, 254.gap
and 176.gcc ~10%.
llvm-svn: 123905
movw r0, :lower16:(L_foo$non_lazy_ptr-(LPC0_0+4))
movt r0, :upper16:(L_foo$non_lazy_ptr-(LPC0_0+4))
LPC0_0:
add r0, pc, r0
It's not yet enabled by default as some tests are failing. I suspect bugs in
down stream tools.
llvm-svn: 123619
in the right direction. It eliminated some hacks and will unblock codegen
work. But it's far from being done. It doesn't reject illegal expressions,
e.g. (FOO - :lower16:BAR). It also doesn't work in Thumb2 mode at all.
llvm-svn: 123369
as a "long" direct branch. While the mnemonics are the same, they encode the branch offset differently, and
the Darwin assembler appears to prefer the "long" form for direct branches. Thus, in the name of bitwise
equivalence, provide encoding and fixup support for it.
llvm-svn: 121710
Thumb2 encoding to share code with the ARM encoding, which gets use fixup support for free.
It also allows us to fold away at least one codegen-only pattern.
llvm-svn: 120481
and xor. The 32-bit move immediates can be hoisted out of loops by machine
LICM but the isel hacks were preventing them.
Instead, let peephole optimization pass recognize registers that are defined by
immediates and the ARM target hook will fold the immediates in.
Other changes include 1) do not fold and / xor into cmp to isel TST / TEQ
instructions if there are multiple uses. This happens when the 'and' is live
out, machine sink would have sinked the computation and that ends up pessimizing
code. The peephole pass would recognize situations where the 'and' can be
toggled to define CPSR and eliminate the comparison anyway.
2) Move peephole pass to after machine LICM, sink, and CSE to avoid blocking
important optimizations.
rdar://8663787, rdar://8241368
llvm-svn: 119548
'db', 'ib', 'da') instead of having that mode as a separate field in the
instruction. It's more convenient for the asm parser and much more readable for
humans.
<rdar://problem/8654088>
llvm-svn: 119310
ins/outs list that isn't specified by their asmstring. Previously
the asmmatcher would just force a 0 register into it, which clearly
isn't right. Mark a bunch of ARM instructions that use this as
isCodeGenOnly. Some of them are clearly pseudo instructions (like
t2TBB) others use a weird hasExtraSrcRegAllocReq thing that will
either need to be removed or the asmmatcher will need to be taught
about it (someday).
llvm-svn: 118119
Instead of silently ignoring these instructions, emit a hard error and
force the target author to either refactor the target or mark the
instruction 'isCodeGenOnly'.
Mark a few instructions in ARM and MBlaze as isCodeGenOnly the are
doing this.
llvm-svn: 117858
There were a number of issues to fix up here:
* The "device" argument of the llvm.memory.barrier intrinsic should be
used to distinguish the "Full System" domain from the "Inner Shareable"
domain. It has nothing to do with using DMB vs. DSB instructions.
* The compiler should never need to emit DSB instructions. Remove the
ARMISD::SYNCBARRIER node and also remove the instruction patterns for DSB.
* Merge the separate DMB/DSB instructions for options only used for the
disassembler with the default DMB/DSB instructions. Add the default
"full system" option ARM_MB::SY to the ARM_MB::MemBOpt enum.
* Add a separate ARMISD::MEMBARRIER_MCR node for subtargets that implement
a data memory barrier using the MCR instruction.
* Fix up encodings for these instructions (except MCR).
I also updated the tests and added a few new ones to check for DMB options
that were not currently being exercised.
llvm-svn: 117756
- For now, loads of [r, r] addressing mode is the same as the
[r, r lsl/lsr/asr #] variants. ARMBaseInstrInfo::getOperandLatency() should
identify the former case and reduce the output latency by 1.
- Also identify [r, r << 2] case. This special form of shifter addressing mode
is "free".
llvm-svn: 117519
allow target to correctly compute latency for cases where static scheduling
itineraries isn't sufficient. e.g. variable_ops instructions such as
ARM::ldm.
This also allows target without scheduling itineraries to compute operand
latencies. e.g. X86 can return (approximated) latencies for high latency
instructions such as division.
- Compute operand latencies for those defined by load multiple instructions,
e.g. ldm and those used by store multiple instructions, e.g. stm.
llvm-svn: 115755
1. Cortex-a9 8-bit and 16-bit loads / stores AGU cycles are 1 cycle longer than 32-bit ones.
2. Cortex-a9 is out-of-order so model all read cycles as cycle 1.
3. Lots of other random fixes for A8 and A9.
llvm-svn: 115121
passed the root of the match, even though only a few patterns
actually needed this (one in X86, several in ARM [which should
be refactored anyway], and some in CellSPU that I don't feel
like detangling). Instead of requiring all ComplexPatterns to
take the dead root, have targets opt into getting the root by
putting SDNPWantRoot on the ComplexPattern.
llvm-svn: 114471
int x(int t) {
if (t & 256)
return -26;
return 0;
}
We generate this:
tst.w r0, #256
mvn r0, #25
it eq
moveq r0, #0
while gcc generates this:
ands r0, r0, #256
it ne
mvnne r0, #25
bx lr
Scandalous really!
During ISel time, we can look for this particular pattern. One where we have a
"MOVCC" that uses the flag off of a CMPZ that itself is comparing an AND
instruction to 0. Something like this (greatly simplified):
%r0 = ISD::AND ...
ARMISD::CMPZ %r0, 0 @ sets [CPSR]
%r0 = ARMISD::MOVCC 0, -26 @ reads [CPSR]
All we have to do is convert the "ISD::AND" into an "ARM::ANDS" that sets [CPSR]
when it's zero. The zero value will all ready be in the %r0 register and we only
need to change it if the AND wasn't zero. Easy!
llvm-svn: 112664
optional modified register (instead of reg0). Along with r112461 it will make
sure that the optional define of CPSR is marked as "def" and will thus mark the
instructions using these classes (t2ANDS*) as setting the 's' flag.
llvm-svn: 112462
comparison with 0. These two pieces of code should give identical results:
rsbs r1, r1, 0
cmp r0, r1
mov r0, #0
it ls
mov r0, #1
and:
cmn r0, r1
mov r0, #0
it ls
mov r0, #1
However, the CMN gives the *opposite* result when r1 is 0. This is because the
carry flag is set in the CMP case but not in the CMN case. In short, the CMP
instruction doesn't perform a truncate of the (logical) NOT of 0 plus the value
of r0 and the carry bit (because the "carry bit" parameter to AddWithCarry is
defined as 1 in this case, the carry flag will always be set when r0 >= 0). The
CMN instruction doesn't perform a NOT of 0 so there is never a "carry" when this
AddWithCarry is performed (because the "carry bit" parameter to AddWithCarry is
defined as 0).
The AddWithCarry in the CMP case seems to be relying upon the identity:
~x + 1 = -x
However when x is 0 and unsigned, this doesn't hold:
x = 0
~x = 0xFFFF FFFF
~x + 1 = 0x1 0000 0000
(-x = 0) != (0x1 0000 0000 = ~x + 1)
Therefore, we should disable *all* versions of CMN, especially when comparing
against zero, until we can limit when the CMN instruction is used (when we know
that the RHS is not 0) or when we have a hardware fix for this.
(See the ARM docs for the "AddWithCarry" pseudo-code.)
This is related to <rdar://problem/7569620>.
llvm-svn: 112176
memory and synchronization barrier dmb and dsb instructions.
- Change instruction names to something more sensible (matching name of actual
instructions).
- Added tests for memory barrier codegen.
llvm-svn: 110785
relatively expensive comparison analyzer on each instruction. Also rename the
comparison analyzer method to something more in line with what it actually does.
This pass is will eventually be folded into the Machine CSE pass.
llvm-svn: 110539
have 4 bits per register in the operand encoding), but have undefined
behavior when the operand value is 13 or 15 (SP and PC, respectively).
The trivial coalescer in linear scan sometimes will merge a copy from
SP into a subsequent instruction which uses the copy, and if that
instruction cannot legally reference SP, we get bad code such as:
mls r0,r9,r0,sp
instead of:
mov r2, sp
mls r0, r9, r0, r2
This patch adds a new register class for use by Thumb2 that excludes
the problematic registers (SP and PC) and is used instead of GPR
for those operands which cannot legally reference PC or SP. The
trivial coalescer explicitly requires that the register class
of the destination for the COPY instruction contain the source
register for the COPY to be considered for coalescing. This prevents
errant instructions like that above.
PR7499
llvm-svn: 109842
- This fixed a number of bugs in if-converter, tail merging, and post-allocation
scheduler. If-converter now runs branch folding / tail merging first to
maximize if-conversion opportunities.
- Also changed the t2IT instruction slightly. It now defines the ITSTATE
register which is read by instructions in the IT block.
- Added Thumb2 specific hazard recognizer to ensure the scheduler doesn't
change the instruction ordering in the IT block (since IT mask has been
finalized). It also ensures no other instructions can be scheduled between
instructions in the IT block.
This is not yet enabled.
llvm-svn: 106344
Thumb2 ADD and SUB instructions: allow RSB instructions be changed to set the
condition codes, and allow RSBS instructions to be predicated.
llvm-svn: 104582
such that the non-VFP versions have no implicit defs of VFP registers.
If any callee-saved VFP registers are marked as having been defined, the
prologue/epilogue code will try to save and restore them.
Radar 7770432.
llvm-svn: 100892
writebacks to the address register. This gets rid of the hack that the
first register on the list was the magic writeback register operand. There
was an implicit constraint that if that operand was not reg0 it had to match
the base register operand. The post-RA scheduler's antidependency breaker
did not understand that constraint and sometimes changed one without the
other. This also fixes Radar 7495976 and should help the verifier work
better for ARM code.
There are now new ld/st instructions explicit writeback operands and explicit
constraints that tie those registers together.
llvm-svn: 98409
immediate instructions cannot set the condition codes, so they do not have
the extra cc_out operand. We hit an assertion during tail duplication
because the instruction being duplicated had more operands that expected.
llvm-svn: 98001
Instruction (PLI) for disassembly only.
According to A8.6.120 PLI (immediate, literal), for example, different
instructions are generated for "pli [pc, #0]" and "pli [pc, #-0"]. The
disassembler solves it by mapping -0 (negative zero) to -1, -1 to -2, ..., etc.
llvm-svn: 97731
the opc string passed in, since it's a given from the class inheritance of T2sI.
The fixed the extra 's' in adcss & sbcss when disassembly printing.
llvm-svn: 97582
o Parallel addition and subtraction, signed/unsigned
o Miscellaneous operations: QADD, QDADD, QSUB, QDSUB
o Unsigned sum of absolute differences [and accumulate]: USAD8, USADA8
o Signed/Unsigned saturate: SSAT, SSAT16, USAT, USAT16
o Signed multiply accumulate long (halfwords): SMLAL<x><y>
o Signed multiply accumulate/subtract [long] (dual): SMLAD[x], SMLALD[X], SMLSD[X], SMLSLD[X]
o Signed dual multiply add/subtract [long]: SMUAD[X], SMUSD[X]
llvm-svn: 97276
to have the predicate on the pattern itself instead. Support for the new
ISel. Remove definitions of CarryDefIsUnused and CarryDefIsUsed since they are
no longer used anywhere.
llvm-svn: 96384
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
load of a GV from constantpool and then add pc. It allows the code sequence to
be rematerializable so it would be hoisted by machine licm.
- Add a late pass to break these pseudo instructions into a number of real
instructions. Also move the code in Thumb2 IT pass that breaks up t2MOVi32imm
to this pass. This is done before post regalloc scheduling to allow the
scheduler to proper schedule these instructions. It also allow them to be
if-converted and shrunk by later passes.
llvm-svn: 86304
bunch of associated comments, because it doesn't have anything to do
with DAGs or scheduling. This is another step in decoupling MachineInstr
emitting from scheduling.
llvm-svn: 85517
instruction. This makes it re-materializable.
Thumb2 will split it back out into two instructions so IT pass will generate the
right mask. Also, this expose opportunies to optimize the movw to a 16-bit move.
llvm-svn: 82982
several things other than Neon vector lane numbers. For inline assembly
operands with a "c" print code, check that they really are immediates.
llvm-svn: 79676
This patch takes pain to ensure all the PEI lowering code does the right thing when lowering frame indices, insert code to manipulate stack pointers, etc. It's also custom lowering dynamic stack alloc into pseudo instructions so we can insert the right instructions at scheduling time.
This fixes PR4659 and PR4682.
llvm-svn: 78361
instructions for calls since BL and BLX are always 32-bit long and BX is always
16-bit long.
Also, we should be using BLX to call external function stubs.
llvm-svn: 77756
- This change also makes it possible to switch between ARM / Thumb on a
per-function basis.
- Fixed thumb2 routine which expand reg + arbitrary immediate. It was using
using ARM so_imm logic.
- Use movw and movt to do reg + imm when profitable.
- Other code clean ups and minor optimizations.
llvm-svn: 77300
Before:
adr r12, #LJTI3_0_0
ldr pc, [r12, +r0, lsl #2]
LJTI3_0_0:
.long LBB3_24
.long LBB3_30
.long LBB3_31
.long LBB3_32
After:
adr r12, #LJTI3_0_0
add pc, r12, +r0, lsl #2
LJTI3_0_0:
b.w LBB3_24
b.w LBB3_30
b.w LBB3_31
b.w LBB3_32
This has several advantages.
1. This will make it easier to optimize this to a TBB / TBH instruction +
(smaller) table.
2. This eliminate the need for ugly asm printer hack to force the address
into thumb addresses (bit 0 is one).
3. Same codegen for pic and non-pic.
4. This eliminate the need to align the table so constantpool island pass
won't have to over-estimate the size.
Based on my calculation, the later is probably slightly faster as well since
ldr pc with shifter address is very slow. That is, it should be a win as long
as the HW implementation can do a reasonable job of branch predict the second
branch.
llvm-svn: 77024
Turns out this also fixed a poor code selection on Thumb1. I have no idea why we were using a mov + add to do the same thing as ADR before.
llvm-svn: 76889
