are targeting an ELF platform. Only fold gs-relative (and fs-relative) loads
if it is actually sensible to do so for the target platform.
This fixes PR13438.
llvm-svn: 160687
Corrected type for index of llvm.x86.avx2.gather.d.pd.256
from 256-bit to 128-bit.
Corrected types for src|dst|mask of llvm.x86.avx2.gather.q.ps.256
from 256-bit to 128-bit.
Support the following intrinsics:
llvm.x86.avx2.gather.d.q, llvm.x86.avx2.gather.q.q
llvm.x86.avx2.gather.d.q.256, llvm.x86.avx2.gather.q.q.256
llvm.x86.avx2.gather.d.d, llvm.x86.avx2.gather.q.d
llvm.x86.avx2.gather.d.d.256, llvm.x86.avx2.gather.q.d.256
llvm-svn: 159402
x86 addressing modes. This allows PIE-based TLS offsets to fit directly
into an addressing mode immediate offset, which is the last remaining
code quality issue from PR12380. With this patch, that PR is completely
fixed.
To understand why this patch is correct to match these offsets into
addressing mode immediates, break it down by cases:
1) 32-bit is trivially correct, and unmodified here.
2) 64-bit non-small mode is unchanged and never matches.
3) 64-bit small PIC code which is RIP-relative is handled specially in
the match to try to fit RIP into the base register. If it fails, it
now early exits. This behavior is unchanged by the patch.
4) 64-bit small non-PIC code which is not RIP-relative continues to work
as it did before. The reason these immediates are safe is because the
ABI ensures they fit in small mode. This behavior is unchanged.
5) 64-bit small PIC code which is *not* using RIP-relative addressing.
This is the only case changed by the patch, and the primary place you
see it is in TLS, either the win64 section offset TLS or Linux
local-exec TLS model in a PIC compilation. Here the ABI again ensures
that the immediates fit because we are in small mode, and any other
operations required due to the PIC relocation model have been handled
externally to the Wrapper node (extra loads etc are made around the
wrapper node in ISelLowering).
I've tested this as much as I can comparing it with GCC's output, and
everything appears safe. I discussed this with Anton and it made sense
to him at least at face value. That said, if there are issues with PIC
code after this patch, yell and we can revert it.
llvm-svn: 154304
This allows us to keep passing reduced masks to SimplifyDemandedBits, but
know about all the bits if SimplifyDemandedBits fails. This allows instcombine
to simplify cases like the one in the included testcase.
llvm-svn: 154011
This is a code change to add support for changing instruction sequences of the form:
load
inc/dec of 8/16/32/64 bits
store
into the appropriate X86 inc/dec through memory instruction:
inc[qlwb] / dec[qlwb]
The checks that were in X86DAGToDAGISel::Select(SDNode *Node)>>ISD::STORE have been extracted to isLoadIncOrDecStore and reworked to use the better
named wrappers for getOperand(unsigned) (e.g. getOffset()) and replaced Chain.getNode() with LoadNode. The comments have also been expanded.
llvm-svn: 153635
This is a code change to add support for changing instruction sequences of the form:
load
inc/dec of 8/16/32/64 bits
store
into the appropriate X86 inc/dec through memory instruction:
inc[qlwb] / dec[qlwb]
The checks that were in X86DAGToDAGISel::Select(SDNode *Node)>>ISD::STORE have been extracted to isLoadIncOrDecStore and reworked to use the better
named wrappers for getOperand(unsigned) (e.g. getOffset()) and replaced Chain.getNode() with LoadNode. The comments have also been expanded.
llvm-svn: 153617
If the DEC node had more than one user, it was doing this lowering but
leaving the original DEC node around and so decrementing twice.
Fixes PR11964.
llvm-svn: 150356
same pattern. We already had this pattern is a few places, but others
tried to make a rough approximation of an actual DAG structure. As not
everywhere went to this trouble, nothing could rely on this being done.
In fact, I've checked all references to these node Ids, and the ones
that are using the topo-sort properties are actually satisfied with
a strict-weak-ordering. The requirement appears to be that Use >= Def.
I've added a big blurb of comments to this bit of the transform to
clarify why the order is so important for the next reader of the code.
I'm starting with this change as it is very small, and trivially
reverted if something breaks or the >= above really does need to be >.
If that proves the case, we can hide the problem by reverting this
patch, but the problem exists elsewhere as well, and so a more
comprehensive solution will be needed.
llvm-svn: 148001
hoped this would revive one of the llvm-gcc selfhost build bots, but it
didn't so it doesn't appear that my transform is the culprit.
If anyone else is seeing failures, please let me know!
llvm-svn: 147957
strange build bot failures that look like a miscompile into an infloop.
I'll investigate this tomorrow, but I'd both like to know whether my
patch is the culprit, and get the bots back to green.
llvm-svn: 147945
mask+shift pairs at the beginning of the ISD::AND case block, and then
hoist the final pattern into a helper function, simplifying and
reflowing it appropriately. This should have no observable behavior
change, but several simplifications fell out of this such as directly
computing the new mask constant, etc.
llvm-svn: 147939
extracts and scaled addressing modes into its own helper function. No
functionality changed here, just hoisting and layout fixes falling out
of that hoisting.
llvm-svn: 147937
detect a pattern which can be implemented with a small 'shl' embedded in
the addressing mode scale. This happens in real code as follows:
unsigned x = my_accelerator_table[input >> 11];
Here we have some lookup table that we look into using the high bits of
'input'. Each entity in the table is 4-bytes, which means this
implicitly gets turned into (once lowered out of a GEP):
*(unsigned*)((char*)my_accelerator_table + ((input >> 11) << 2));
The shift right followed by a shift left is canonicalized to a smaller
shift right and masking off the low bits. That hides the shift right
which x86 has an addressing mode designed to support. We now detect
masks of this form, and produce the longer shift right followed by the
proper addressing mode. In addition to saving a (rather large)
instruction, this also reduces stalls in Intel chips on benchmarks I've
measured.
In order for all of this to work, one part of the DAG needs to be
canonicalized *still further* than it currently is. This involves
removing pointless 'trunc' nodes between a zextload and a zext. Without
that, we end up generating spurious masks and hiding the pattern.
llvm-svn: 147936
this substraction will result in small negative numbers at worst which
become very large positive numbers on assignment and are thus caught by
the <=4 check on the next line. The >0 check clearly intended to catch
these as negative numbers.
Spotted by inspection, and impossible to trigger given the shift widths
that can be used.
llvm-svn: 147773
fixes: Use a separate register, instead of SP, as the
calling-convention resource, to avoid spurious conflicts with
actual uses of SP. Also, fix unscheduling of calling sequences,
which can be triggered by pseudo-two-address dependencies.
llvm-svn: 143206
it fixes the dragonegg self-host (it looks like gcc is miscompiled).
Original commit messages:
Eliminate LegalizeOps' LegalizedNodes map and have it just call RAUW
on every node as it legalizes them. This makes it easier to use
hasOneUse() heuristics, since unneeded nodes can be removed from the
DAG earlier.
Make LegalizeOps visit the DAG in an operands-last order. It previously
used operands-first, because LegalizeTypes has to go operands-first, and
LegalizeTypes used to be part of LegalizeOps, but they're now split.
The operands-last order is more natural for several legalization tasks.
For example, it allows lowering code for nodes with floating-point or
vector constants to see those constants directly instead of seeing the
lowered form (often constant-pool loads). This makes some things
somewhat more complicated today, though it ought to allow things to be
simpler in the future. It also fixes some bugs exposed by Legalizing
using RAUW aggressively.
Remove the part of LegalizeOps that attempted to patch up invalid chain
operands on libcalls generated by LegalizeTypes, since it doesn't work
with the new LegalizeOps traversal order. Instead, define what
LegalizeTypes is doing to be correct, and transfer the responsibility
of keeping calls from having overlapping calling sequences into the
scheduler.
Teach the scheduler to model callseq_begin/end pairs as having a
physical register definition/use to prevent calls from having
overlapping calling sequences. This is also somewhat complicated, though
there are ways it might be simplified in the future.
This addresses rdar://9816668, rdar://10043614, rdar://8434668, and others.
Please direct high-level questions about this patch to management.
Delete #if 0 code accidentally left in.
llvm-svn: 143188
on every node as it legalizes them. This makes it easier to use
hasOneUse() heuristics, since unneeded nodes can be removed from the
DAG earlier.
Make LegalizeOps visit the DAG in an operands-last order. It previously
used operands-first, because LegalizeTypes has to go operands-first, and
LegalizeTypes used to be part of LegalizeOps, but they're now split.
The operands-last order is more natural for several legalization tasks.
For example, it allows lowering code for nodes with floating-point or
vector constants to see those constants directly instead of seeing the
lowered form (often constant-pool loads). This makes some things
somewhat more complicated today, though it ought to allow things to be
simpler in the future. It also fixes some bugs exposed by Legalizing
using RAUW aggressively.
Remove the part of LegalizeOps that attempted to patch up invalid chain
operands on libcalls generated by LegalizeTypes, since it doesn't work
with the new LegalizeOps traversal order. Instead, define what
LegalizeTypes is doing to be correct, and transfer the responsibility
of keeping calls from having overlapping calling sequences into the
scheduler.
Teach the scheduler to model callseq_begin/end pairs as having a
physical register definition/use to prevent calls from having
overlapping calling sequences. This is also somewhat complicated, though
there are ways it might be simplified in the future.
This addresses rdar://9816668, rdar://10043614, rdar://8434668, and others.
Please direct high-level questions about this patch to management.
llvm-svn: 143177
In 64-bit mode, sub_8bit_hi sub-registers can only be used by NOREX
instructions. The COPY created from the EXTRACT_SUBREG DAG node cannot
target all GR8 registers, only those in GR8_NOREX.
TO enforce this, we ensure that all instructions using the
EXTRACT_SUBREG are GR8_NOREX constrained.
This fixes PR11088.
llvm-svn: 141499
This tends to happen a lot with bitfield code generated by clang. A simple example for x86_64 is
uint64_t foo(uint64_t x) { return (x&1) << 42; }
which used to compile into bloated code:
shlq $42, %rdi ## encoding: [0x48,0xc1,0xe7,0x2a]
movabsq $4398046511104, %rax ## encoding: [0x48,0xb8,0x00,0x00,0x00,0x00,0x00,0x04,0x00,0x00]
andq %rdi, %rax ## encoding: [0x48,0x21,0xf8]
ret ## encoding: [0xc3]
with this patch we can fold the immediate into the and:
andq $1, %rdi ## encoding: [0x48,0x83,0xe7,0x01]
movq %rdi, %rax ## encoding: [0x48,0x89,0xf8]
shlq $42, %rax ## encoding: [0x48,0xc1,0xe0,0x2a]
ret ## encoding: [0xc3]
It's possible to save another byte by using 'andl' instead of 'andq' but I currently see no way of doing
that without making this code even more complicated. See the TODOs in the code.
llvm-svn: 129990
have their low bits set to zero. This allows us to optimize
out explicit stack alignment code like in stack-align.ll:test4 when
it is redundant.
Doing this causes the code generator to start turning FI+cst into
FI|cst all over the place, which is general goodness (that is the
canonical form) except that various pieces of the code generator
don't handle OR aggressively. Fix this by introducing a new
SelectionDAG::isBaseWithConstantOffset predicate, and using it
in places that are looking for ADD(X,CST). The ARM backend in
particular was missing a lot of addressing mode folding opportunities
around OR.
llvm-svn: 125470
into and/shift would cause nodes to move around and a dangling pointer
to happen. The code tried to avoid this with a HandleSDNode, but
got the details wrong.
llvm-svn: 123578
beginning of the "main" function. The assembler complains about the invalid
suffix for the 'call' instruction. The right instruction is "callq __main".
Patch by KS Sreeram!
llvm-svn: 122933
backend that they were all implemented except umul. This one fell back
to the default implementation that did a hi/lo multiply and compared the
top. Fix this to check the overflow flag that the 'mul' instruction
sets, so we can avoid an explicit test. Now we compile:
void *func(long count) {
return new int[count];
}
into:
__Z4funcl: ## @_Z4funcl
movl $4, %ecx ## encoding: [0xb9,0x04,0x00,0x00,0x00]
movq %rdi, %rax ## encoding: [0x48,0x89,0xf8]
mulq %rcx ## encoding: [0x48,0xf7,0xe1]
seto %cl ## encoding: [0x0f,0x90,0xc1]
testb %cl, %cl ## encoding: [0x84,0xc9]
movq $-1, %rdi ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
cmoveq %rax, %rdi ## encoding: [0x48,0x0f,0x44,0xf8]
jmp __Znam ## TAILCALL
instead of:
__Z4funcl: ## @_Z4funcl
movl $4, %ecx ## encoding: [0xb9,0x04,0x00,0x00,0x00]
movq %rdi, %rax ## encoding: [0x48,0x89,0xf8]
mulq %rcx ## encoding: [0x48,0xf7,0xe1]
testq %rdx, %rdx ## encoding: [0x48,0x85,0xd2]
movq $-1, %rdi ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
cmoveq %rax, %rdi ## encoding: [0x48,0x0f,0x44,0xf8]
jmp __Znam ## TAILCALL
Other than the silly seto+test, this is using the o bit directly, so it's going in the right
direction.
llvm-svn: 120935
by having X86DAGToDAGISel::SelectAddr get passed in the parent node
of the operand match (the load/store/atomic op) and having it get
the address space from that, instead of having special FS/GS addr
mode operations that require duplicating the entire instruction set
to support.
This makes FS and GS relative accesses *far* more predictable and
work much better. It also simplifies the X86 backend a bit, more
to come.
There is still a pending issue with nodes like ISD::PREFETCH and
X86ISD::FLD, which really should be MemSDNode's but aren't.
llvm-svn: 114491
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
like all other instructions, even though a segment is not
allowed. This resolves a bunch of gross hacks in the
encoder and makes LEA more consistent with the rest of the
instruction set.
No functionality change.
llvm-svn: 107934
CopyFromReg nodes for aliasing registers (AX and AL). This confuses the fast
register allocator.
Instead of CopyFromReg(AL), use ExtractSubReg(CopyFromReg(AX), sub_8bit).
This fixes PR7312.
llvm-svn: 106934
const_casts, and it reinforces the design of the Target classes being
immutable.
SelectionDAGISel::IsLegalToFold is now a static member function, because
PIC16 uses it in an unconventional way. There is more room for API
cleanup here.
And PIC16's AsmPrinter no longer uses TargetLowering.
llvm-svn: 101635
that they are not destination type specific. This allows
tblgen to factor them and the type check is redundant with
what the isel does anyway.
llvm-svn: 97629
CopyToReg/CopyFromReg/INLINEASM. These are annoying because
they have the same opcode before an after isel. Fix this by
setting their NodeID to -1 to indicate that they are selected,
just like what automatically happens when selecting things that
end up being machine nodes.
With that done, give IsLegalToFold a new flag that causes it to
ignore chains. This lets the HandleMergeInputChains routine be
the one place that validates chains after a match is successful,
enabling the new hotness in chain processing. This smarter
chain processing eliminates the need for "PreprocessRMW" in the
X86 and MSP430 backends and enables MSP to start matching it's
multiple mem operand instructions more aggressively.
I currently #if out the dead code in the X86 backend and MSP
backend, I'll remove it for real in a follow-on patch.
The testcase changes are:
test/CodeGen/X86/sse3.ll: we generate better code
test/CodeGen/X86/store_op_load_fold2.ll: PreprocessRMW was
miscompiling this before, we now generate correct code
Convert it to filecheck while I'm at it.
test/CodeGen/MSP430/Inst16mm.ll: Add a testcase for mem/mem
folding to make anton happy. :)
llvm-svn: 97596
DoInstructionSelection. Inline "SelectRoot" into it from DAGISelHeader.
Sink some other stuff out of DAGISelHeader into SDISel.
Eliminate the various 'Indent' stuff from various targets, which dates
to when isel was recursive.
17 files changed, 114 insertions(+), 430 deletions(-)
llvm-svn: 97555
it to follow the mode needed by the new isel. Instead of returning
the input and output chains, it just returns the (currently only one,
which is a silly limitation) node that has input and output chains.
Since we want the old thing to still work, add a new
SelectScalarSSELoad to emulate the old interface. The XXX suffix
and the wrapper will eventually go away.
llvm-svn: 96715
into a roundss intrinsic, producing a cyclic dag. The root cause
of this is badness handling ComplexPattern nodes in the old dagisel
that I noticed through inspection. Eliminate a copy of the of the
code that handled ComplexPatterns by making EmitChildMatchCode call
into EmitMatchCode.
llvm-svn: 96408
IsLegalToFold and IsProfitableToFold. The generic version of the later simply checks whether the folding candidate has a single use.
This allows the target isel routines more flexibility in deciding whether folding makes sense. The specific case we are interested in is folding constant pool loads with multiple uses.
llvm-svn: 96255
into TargetOpcodes.h. #include the new TargetOpcodes.h
into MachineInstr. Add new inline accessors (like isPHI())
to MachineInstr, and start using them throughout the
codebase.
llvm-svn: 95687
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
new AsmPrinter. This is perhaps less elegant than describing them
in terms of MOV32r0 and subreg operations, but it allows the
current register to rematerialize them.
llvm-svn: 93158
clear what information these functions are actually using.
This is also a micro-optimization, as passing a SDNode * around is
simpler than passing a { SDNode *, int } by value or reference.
llvm-svn: 92564
- Force NDEBUG on in any Release build. This drops the compile time to ~100s
from ~600s, in Release mode.
- This may just be a temporary workaround, I don't know the true nature of the
gcc-4.2 compile time performance problem.
llvm-svn: 86695
the testcase into:
_test1: ## @test1
## BB#0: ## %entry
leaq L_test1_bb6(%rip), %rax
jmpq *%rax
L_test1_bb: ## Address Taken
LBB1_1: ## %bb
movb $1, %al
ret
L_test1_bb6: ## Address Taken
LBB1_2: ## %bb6
movb $2, %al
ret
Note, it is very very strange that BlockAddressSDNode doesn't carry
around TargetFlags. Dan, please fix this.
llvm-svn: 85703
when one of the bits being tested would end up being the sign bit in the
narrower type, and a signed comparison is being performed, since this would
change the result of the signed comparison. This fixes PR5132.
llvm-svn: 83670
- 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
naming scheme used in SelectionDAG, where there are multiple kinds
of "target" nodes, but "machine" nodes are nodes which represent
a MachineInstr.
llvm-svn: 82790
trying to create RMW opportunities in the x86 backend. This can cause a
cycle to appear in the graph, since the other uses may eventually feed into
the TokenFactor we are sinking the load below.
llvm-svn: 81996
has multiple uses, as one of the other uses may be on a path
to a different node above the callseq_start, because that
leads to a cyclic graph. This problem is exposed when
-combiner-global-alias-analysis is used. This fixes PR4880.
llvm-svn: 81821
Craig Topper