Shuffles that only move an element into position 0 of the vector are common in
the output of the loop vectorizer and often generate suboptimal code when SSSE3
is not available. Lower them to vector shifts if possible.
We still prefer palignr over psrldq because it has higher throughput on
sandybridge.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182102 91177308-0d34-0410-b5e6-96231b3b80d8
This patch implements the equivalent change to r182091/r182092
in the old-style code emitter. Instead of having two separate
16-bit immediate encoding routines depending on the instruction,
this patch introduces a single encoder that checks the machine
operand flags to decide whether the low or high half of a
symbol address is required.
Since now both encoders make no further distinction between
"symbolLo" and "symbolHi", the .td operand can now use a
single getS16ImmEncoding method.
Tested by running the old-style JIT tests on 32-bit Linux.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182097 91177308-0d34-0410-b5e6-96231b3b80d8
Now that fixup_ppc_ha16 and fixup_ppc_lo16 are being treated exactly
the same everywhere, it no longer makes sense to have two fixup types.
This patch merges them both into a single type fixup_ppc_half16,
and renames fixup_ppc_lo16_ds to fixup_ppc_half16ds for consistency.
(The half16 and half16ds names are taken from the description of
relocation types in the PowerPC ABI.)
No change in code generation expected.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182092 91177308-0d34-0410-b5e6-96231b3b80d8
The current PowerPC MC back end distinguishes between fixup_ppc_ha16
and fixup_ppc_lo16, which are determined by the instruction the fixup
applies to, and uses this distinction to decide whether a fixup ought
to resolve to the high or the low part of a symbol address.
This isn't quite correct, however. It is valid -if unusual- assembler
to use, e.g.
li 1, symbol@ha
or
lis 1, symbol@l
Whether the high or the low part of the address is used depends solely
on the @ suffix, not on the instruction.
In addition, both
li 1, symbol
and
lis 1, symbol
are valid, assuming the symbol address fits into 16 bits; again, both
will then refer to the actual symbol value (so li will load the value
itself, while lis will load the value shifted by 16).
To fix this, two places need to be adapted. If the fixup cannot be
resolved at assembler time, a relocation needs to be emitted via
PPCELFObjectWriter::getRelocType. This routine already looks at
the VK_ type to determine the relocation. The only problem is that
will reject any _LO modifier in a ha16 fixup and vice versa. This
is simply incorrect; any of those modifiers ought to be accepted
for either fixup type.
If the fixup *can* be resolved at assembler time, adjustFixupValue
currently selects the high bits of the symbol value if the fixup
type is ha16. Again, this is incorrect; see the above example
lis 1, symbol
Now, in theory we'd have to respect a VK_ modifier here. However,
in fact common code never even attempts to resolve symbol references
using any nontrivial VK_ modifier at assembler time; it will always
fall back to emitting a reloc and letting the linker handle it.
If this ever changes, presumably there'd have to be a target callback
to resolve VK_ modifiers. We'd then have to handle @ha etc. there.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182091 91177308-0d34-0410-b5e6-96231b3b80d8
Previously, three instructions were needed:
trunc.w.s $f0, $f2
mfc1 $4, $f0
sw $4, 0($2)
Now we need only two:
trunc.w.s $f0, $f2
swc1 $f0, 0($2)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182053 91177308-0d34-0410-b5e6-96231b3b80d8
This patch removes alias definition for addiu $rs,$imm
and instead uses the TwoOperandAliasConstraint field in
the ArithLogicI instruction class.
This way all instructions that inherit ArithLogicI class
have the same macro defined.
The usage examples are added to test files.
Patch by Vladimir Medic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182048 91177308-0d34-0410-b5e6-96231b3b80d8
Some IR-level instructions (such as FP <-> i64 conversions) are not chained
w.r.t. the mtctr intrinsic and yet may become function calls that clobber the
counter register. At the selection-DAG level, these might be reordered with the
mtctr intrinsic causing miscompiles. To avoid this situation, if an existing
preheader has instructions that might use the counter register, create a new
preheader for the mtctr intrinsic. This extra block will be remerged with the
old preheader at the MI level, but will prevent unwanted reordering at the
selection-DAG level.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182045 91177308-0d34-0410-b5e6-96231b3b80d8
invalid instruction sequence.
Rather than emitting an int-to-FP move instruction and an int-to-FP conversion
instruction during instruction selection, we emit a pseudo instruction which gets
expanded post-RA. Without this change, register allocation can possibly insert a
floating point register move instruction between the two instructions, which is not
valid according to the ISA manual.
mtc1 $f4, $4 # int-to-fp move instruction.
mov.s $f2, $f4 # move contents of $f4 to $f2.
cvt.s.w $f0, $f2 # int-to-fp conversion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182042 91177308-0d34-0410-b5e6-96231b3b80d8
This patch adds bnez and beqz instructions which represent alias definitions for bne and beq instructions as follows:
bnez $rs,$imm => bne $rs,$zero,$imm
beqz $rs,$imm => beq $rs,$zero,$imm
The corresponding test cases are added.
Patch by Vladimir Medic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182040 91177308-0d34-0410-b5e6-96231b3b80d8
This is the second part of the change to always return "true"
offset values from getPreIndexedAddressParts, tackling the
case of "memrix" type operands.
This is about instructions like LD/STD that only have a 14-bit
field to encode immediate offsets, which are implicitly extended
by two zero bits by the machine, so that in effect we can access
16-bit offsets as long as they are a multiple of 4.
The PowerPC back end currently handles such instructions by
carrying the 14-bit value (as it will get encoded into the
actual machine instructions) in the machine operand fields
for such instructions. This means that those values are
in fact not the true offset, but rather the offset divided
by 4 (and then truncated to an unsigned 14-bit value).
Like in the case fixed in r182012, this makes common code
operations on such offset values not work as expected.
Furthermore, there doesn't really appear to be any strong
reason why we should encode machine operands this way.
This patch therefore changes the encoding of "memrix" type
machine operands to simply contain the "true" offset value
as a signed immediate value, while enforcing the rules that
it must fit in a 16-bit signed value and must also be a
multiple of 4.
This change must be made simultaneously in all places that
access machine operands of this type. However, just about
all those changes make the code simpler; in many cases we
can now just share the same code for memri and memrix
operands.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182032 91177308-0d34-0410-b5e6-96231b3b80d8
On PPC32, i64 FP conversions are implemented using runtime calls (which clobber
the counter register). These must be excluded.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182023 91177308-0d34-0410-b5e6-96231b3b80d8
DAGCombiner::CombineToPreIndexedLoadStore calls a target routine to
decompose a memory address into a base/offset pair. It expects the
offset (if constant) to be the true displacement value in order to
perform optional additional optimizations; in particular, to convert
other uses of the original pointer into uses of the new base pointer
after pre-increment.
The PowerPC implementation of getPreIndexedAddressParts, however,
simply calls SelectAddressRegImm, which returns a TargetConstant.
This value is appropriate for encoding into the instruction, but
it is not always usable as true displacement value:
- Its type is always MVT::i32, even on 64-bit, where addresses
ought to be i64 ... this causes the optimization to simply
always fail on 64-bit due to this line in DAGCombiner:
// FIXME: In some cases, we can be smarter about this.
if (Op1.getValueType() != Offset.getValueType()) {
- Its value is truncated to an unsigned 16-bit value if negative.
This causes the above opimization to generate wrong code.
This patch fixes both problems by simply returning the true
displacement value (in its original type). This doesn't
affect any other user of the displacement.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182012 91177308-0d34-0410-b5e6-96231b3b80d8
getExceptionHandlingType is not ExceptionHandling::DwarfCFI on xcore, so
etFrameInstructions is never called. There is no point creating cfi
instructions if they are never used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181979 91177308-0d34-0410-b5e6-96231b3b80d8
This creates stubs that help Mips32 functions call Mips16
functions which have floating point parameters that are normally passed
in floating point registers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181972 91177308-0d34-0410-b5e6-96231b3b80d8
Increase the number of instructions LLVM recognizes as setting the ZF
flag. This allows us to remove test instructions that redundantly
recalculate the flag.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181937 91177308-0d34-0410-b5e6-96231b3b80d8
The old PPCCTRLoops pass, like the Hexagon pass version from which it was
derived, could only handle some simple loops in canonical form. We cannot
directly adapt the new Hexagon hardware loops pass, however, because the
Hexagon pass contains a fundamental assumption that non-constant-trip-count
loops will contain a guard, and this is not always true (the result being that
incorrect negative counts can be generated). With this commit, we replace the
pass with a late IR-level pass which makes use of SE to calculate the
backedge-taken counts and safely generate the loop-count expressions (including
any necessary max() parts). This IR level pass inserts custom intrinsics that
are lowered into the desired decrement-and-branch instructions.
The most fragile part of this new implementation is that interfering uses of
the counter register must be detected on the IR level (and, on PPC, this also
includes any indirect branches in addition to function calls). Also, to make
all of this work, we need a variant of the mtctr instruction that is marked
as having side effects. Without this, machine-code level CSE, DCE, etc.
illegally transform the resulting code. Hopefully, this can be improved
in the future.
This new pass is smaller than the original (and much smaller than the new
Hexagon hardware loops pass), and can handle many additional cases correctly.
In addition, the preheader-creation code has been copied from LoopSimplify, and
after we decide on where it belongs, this code will be refactored so that it
can be explicitly shared (making this implementation even smaller).
The new test-case files ctrloop-{le,lt,ne}.ll have been adapted from tests for
the new Hexagon pass. There are a few classes of loops that this pass does not
transform (noted by FIXMEs in the files), but these deficiencies can be
addressed within the SE infrastructure (thus helping many other passes as well).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181927 91177308-0d34-0410-b5e6-96231b3b80d8
We want the order to be deterministic on all platforms. NAKAMURA Takumi
fixed that in r181864. This patch is just two small cleanups:
* Move the function to the cpp file. It is only passed to array_pod_sort.
* Remove the ppc implementation which is now redundant
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181910 91177308-0d34-0410-b5e6-96231b3b80d8
This patch matches GCC behavior: the code used to only allow unaligned
load/store on ARM for v6+ Darwin, it will now allow unaligned load/store for
v6+ Darwin as well as for v7+ on other targets.
The distinction is made because v6 doesn't guarantee support (but LLVM assumes
that Apple controls hardware+kernel and therefore have conformant v6 CPUs),
whereas v7 does provide this guarantee (and Linux behaves sanely).
Overall this should slightly improve performance in most cases because of
reduced I$ pressure.
Patch by JF Bastien
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181897 91177308-0d34-0410-b5e6-96231b3b80d8
Now that applyFixup understands differently-sized fixups, we can define
fixup_ppc_lo16/fixup_ppc_lo16_ds/fixup_ppc_ha16 to properly be 2-byte
fixups, applied at an offset of 2 relative to the start of the
instruction text.
This has the benefit that if we actually need to generate a real
relocation record, its address will come out correctly automatically,
without having to fiddle with the offset in adjustFixupOffset.
Tested on both 64-bit and 32-bit PowerPC, using external and
integrated assembler.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181894 91177308-0d34-0410-b5e6-96231b3b80d8
The PPCAsmBackend::applyFixup routine handles the case where a
fixup can be resolved within the same object file. However,
this routine is currently hard-coded to assume the size of
any fixup is always exactly 4 bytes.
This is sort-of correct for fixups on instruction text; even
though it only works because several of what really would be
2-byte fixups are presented as 4-byte fixups instead (requiring
another hack in PPCELFObjectWriter::adjustFixupOffset to clean
it up).
However, this assumption breaks down completely for fixups
on data, which legitimately can be of any size (1, 2, 4, or 8).
This patch makes applyFixup aware of fixups of varying sizes,
introducing a new helper routine getFixupKindNumBytes (along
the lines of what the ARM back end does). Note that in order
to handle fixups of size 8, we also need to fix the return type
of adjustFixupValue to uint64_t to avoid truncation.
Tested on both 64-bit and 32-bit PowerPC, using external and
integrated assembler.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181891 91177308-0d34-0410-b5e6-96231b3b80d8
The transformation happening here is that we want to turn a
"mul(ext(X), ext(X))" into a "vmull(X, X)", stripping off the extension. We have
to make sure that X still has a valid vector type - possibly recreate an
extension to a smaller type. In case of a extload of a memory type smaller than
64 bit we used create a ext(load()). The problem with doing this - instead of
recreating an extload - is that an illegal type is exposed.
This patch fixes this by creating extloads instead of ext(load()) sequences.
Fixes PR15970.
radar://13871383
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181842 91177308-0d34-0410-b5e6-96231b3b80d8
