For some reason, adding definitions for these load and store
instructions changed whether some of the build bots matched
comparisons as signed or unsigned.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191663 91177308-0d34-0410-b5e6-96231b3b80d8
The only thing this does on its own is make the definitions of RISB[HL]G
a bit more precise. Those instructions are only used by the MC layer at
the moment, so no behavioral change is intended. The class is needed by
later patches though.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191660 91177308-0d34-0410-b5e6-96231b3b80d8
Use subreg_hNN and subreg_lNN for the high and low NN bits of a register.
List the low registers first, so that subreg_l32 also means the low 32
bits of a 128-bit register.
Floats are stored in the upper 32 bits of a 64-bit register, so they
should use subreg_h32 rather than subreg_l32.
No behavioral change intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191659 91177308-0d34-0410-b5e6-96231b3b80d8
I'm about to add support for high-word operations, so it seemed better
for the low-word registers to have names like R0L rather than R0W.
No behavioral change intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191655 91177308-0d34-0410-b5e6-96231b3b80d8
Add VEX_LIG to scalar FMA4 instructions.
Use VEX_LIG in some of the inheriting checks in disassembler table generator.
Make use of VEX_L_W, VEX_L_W_XS, VEX_L_W_XD contexts.
Don't let VEX_L_W, VEX_L_W_XS, VEX_L_W_XD, VEX_L_W_OPSIZE inherit from their non-L forms unless VEX_LIG is set.
Let VEX_L_W, VEX_L_W_XS, VEX_L_W_XD, VEX_L_W_OPSIZE inherit from all of their non-L or non-W cases.
Increase ranking on VEX_L_W, VEX_L_W_XS, VEX_L_W_XD, VEX_L_W_OPSIZE so they get chosen over non-L/non-W forms.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191649 91177308-0d34-0410-b5e6-96231b3b80d8
We were completely ignoring the unorder/ordered attributes of condition
codes and also incorrectly lowering seto and setuo.
Reviewed-by: Vincent Lejeune<vljn at ovi.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191603 91177308-0d34-0410-b5e6-96231b3b80d8
SelectionDAG will now attempt to inverse an illegal conditon in order to
find a legal one and if that doesn't work, it will attempt to swap the
operands using the inverted condition.
There are no new test cases for this, but a nubmer of the existing R600
tests hit this path.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191602 91177308-0d34-0410-b5e6-96231b3b80d8
This is useful for targets like R600, which only support GT, GE, NE, and EQ
condition codes as it removes the need to handle unsupported condition
codes in target specific code.
There are no tests with this commit, but R600 has been updated to take
advantage of this new feature, so its existing selectcc tests are now
testing the swapped operands path.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191601 91177308-0d34-0410-b5e6-96231b3b80d8
of loops.
Previously, two consecutive calls to function "func" would result in the
following sequence of instructions:
1. load $16, %got(func)($gp) // load address of lazy-binding stub.
2. move $25, $16
3. jalr $25 // jump to lazy-binding stub.
4. nop
5. move $25, $16
6. jalr $25 // jump to lazy-binding stub again.
With this patch, the second call directly jumps to func's address, bypassing
the lazy-binding resolution routine:
1. load $25, %got(func)($gp) // load address of lazy-binding stub.
2. jalr $25 // jump to lazy-binding stub.
3. nop
4. load $25, %got(func)($gp) // load resolved address of func.
5. jalr $25 // directly jump to func.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191591 91177308-0d34-0410-b5e6-96231b3b80d8
The backend tries to use block operations like MVC, NC, OC and XC for
simple scalar operations. For correctness reasons, it rejects any case
in which the regions might partially overlap. However, for performance
reasons, it should also reject cases where the regions might be equal,
since the instruction might then not use the fast path.
This fixes a performance regression seen in bzip2. We may want to limit
the optimisation even more in future, or even remove it entirely, but I'll
try with this for now.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191525 91177308-0d34-0410-b5e6-96231b3b80d8
The backend previously folded offsets into PC-relative addresses
whereever possible. That's the right thing to do when the address
can be used directly in a PC-relative memory reference (using things
like LRL). But if we have a register-based memory reference and need
to load the PC-relative address separately, it's better to use an anchor
point that could be shared with other accesses to the same area of the
variable.
Fixes a FIXME.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191524 91177308-0d34-0410-b5e6-96231b3b80d8
This intrinsic is lowered into an equivalent INSERT_VECTOR_ELT which is
further lowered into a sequence of insert.w's on MIPS32.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191521 91177308-0d34-0410-b5e6-96231b3b80d8
As specified in A8.8.72/A8.8.73/A8.8.74 in the ARM ARM, all variants of the ARM LDRD instruction have the following two constraints:
LDRD<c> <Rt>, <Rt2>, ...
(a) Rt must be even-numbered and not r14
(b) Rt2 must be R(t+1)
If those two constraints are not met the result of executing the instruction will be unpredictable.
Constraint (b) was already enforced, this commit adds support for constraint (a).
Fixes rdar://14479793.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191520 91177308-0d34-0410-b5e6-96231b3b80d8
This intrinsic is lowered into an equivalent BUILD_VECTOR which is further
lowered into a sequence of insert.w's on MIPS32.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191519 91177308-0d34-0410-b5e6-96231b3b80d8
For v4f32 and v2f64, INSERT_VECTOR_ELT is matched by a pseudo-insn which is
later expanded to appropriate insve.[wd] insns.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191515 91177308-0d34-0410-b5e6-96231b3b80d8
For v4f32 and v2f64, EXTRACT_VECTOR_ELT is matched by a pseudo-insn which may
be expanded to subregister copies and/or instructions as appropriate.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191514 91177308-0d34-0410-b5e6-96231b3b80d8
This file contains notes about the instruction selection for MSA. For example,
it notes that ilvl.d is cannot be selected because ilvev.d covers the same
cases and is selected instead of ilvl.d.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191507 91177308-0d34-0410-b5e6-96231b3b80d8
LDRD<c> <Rt>, <Rt2>, <label>
LDRD<c> <Rt>, <Rt2>, [<Rn>{, #+/-<imm>}]
LDRD<c> <Rt>, <Rt2>, [<Rn>], #+/-<imm>
LDRD<c> <Rt>, <Rt2>, [<Rn>, #+/-<imm>]!
As specified in A8.8.72/A8.8.73 in the ARM ARM, the T1 encoding has a constraint which enforces that Rt != Rt2.
If this constraint is not met the result of executing the instruction will be unpredictable.
Fixes rdar://14479780.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191504 91177308-0d34-0410-b5e6-96231b3b80d8
lowerMSABinaryIntr, lowerMSABinaryImmIntr, lowerMSABranchIntr,
and lowerMSAUnaryIntr were trivially small functions. Inlined them into
their callers.
lowerMSASplat now takes its callers SDLoc instead of making a new one.
No functional change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191503 91177308-0d34-0410-b5e6-96231b3b80d8
When generating code for shared libraries, even local calls may be
intercepted, so we need a nop after the call for the linker to fix up the
TOC. Test case adapted from the one provided in PR17354.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191440 91177308-0d34-0410-b5e6-96231b3b80d8
When asked to pad an irregular number of bytes, we should fill with
zeros. This is consistent with the behavior specified in the AIX
Assembler Language Reference as well as other LLVM and binutils
assemblers.
N.B. There is a small deviation from binutils' PPC assembler:
when handling pads which are greater than 4 bytes but not mod 4,
binutils will not emit any NOP sequences at all and only use zeros.
This may or may not be a bug but there is no excellent rationale as to
why that behavior is important to emulate. If that behavior is needed,
we can change writeNopData() to behave in the same way.
This fixes PR17352.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191426 91177308-0d34-0410-b5e6-96231b3b80d8
Encodings were checked against the Power ISA documents and double
checked against binutils.
This fixes PR17350.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191419 91177308-0d34-0410-b5e6-96231b3b80d8
This is the first set of instructions with a ".b" modifier thus we need to add the required code to disassemble a MSA128B register class.
Patch by Matheus Almeida
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191415 91177308-0d34-0410-b5e6-96231b3b80d8
In more detail, this patch adds the ability to parse, encode and decode MSA registers ($w0-$w31). The format of 2RF instructions (MipsMSAInstrFormat.td) was updated so that we could attach a test case to this patch i.e., the test case parses, encodes and decodes 2 MSA instructions. Following patches will add the remainder of the instructions.
Note that DecodeMSA128BRegisterClass is missing from MipsDisassembler.td because it's not yet required at this stage and having it would cause a compiler warning (unused function).
Patch by Matheus Almeida
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191412 91177308-0d34-0410-b5e6-96231b3b80d8
Generally, it is desirable to distribute (a + b) * c to a*c + b*c for
ARM with VMLx forwarding, where a, b and c are vectors.
However, for (a + b)*(a + b), distribution will result in one extra
instruction.
With distribution:
x = a + b (add)
y = a * x (mul)
z = y + b * y (mla)
Without distribution:
x = a + b (add)
z = x * x (mul)
This patch checks if a mul is a square of add/sub. If yes, skip
distribution.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191410 91177308-0d34-0410-b5e6-96231b3b80d8
Ideally, the machinel model is added at the time the instructions are
defined. But many instructions in X86InstrSSE.td still need a model.
Without this workaround the scheduler asserts because x86 already has
itinerary classes for these instructions, indicating they should be
modeled by the scheduler. Since we use the new machine model for other
instructions, it expects a new machine model for these too.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191391 91177308-0d34-0410-b5e6-96231b3b80d8
Another patch to avoid duplication of encoding information. Things like
NILF, NILL and NILH are used as both 32-bit and 64-bit instructions.
Here the 64-bit versions are defined as aliases of the 32-bit ones.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191369 91177308-0d34-0410-b5e6-96231b3b80d8
The binutils assembler supports a mode called DOLLAR_DOT which treats
the dollar sign token as a reference to the current program counter if
the dollar sign doesn't precede a constant or identifier.
This commit adds a new MCAsmInfo flag stating whether or not a given
target supports this interpretation of the dollar sign token; by
default, this flag is not enabled.
Further, enable this flag for PPC. The system assembler for AIX and
binutils both support using the dollar sign in this manner.
This fixes PR17353.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191368 91177308-0d34-0410-b5e6-96231b3b80d8
Similar to r191364, but for calls. This patch also removes the shortening
of BRASL to BRAS within a TU. Doing that was a bit controversial internally,
since there's a strong expectation with the z assembler that WYWIWYG.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191366 91177308-0d34-0410-b5e6-96231b3b80d8
Another patch to reduce the duplication of encoding information.
Rather than define separate patterns for truncating 64-bit stores,
use the 32-bit stores with a subreg. No behavioral changed intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191365 91177308-0d34-0410-b5e6-96231b3b80d8
This is the first of a few patches to reduce the dupliation of encoding
information. The return instruction is a normal BR in which one of the
registers is fixed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191364 91177308-0d34-0410-b5e6-96231b3b80d8
When loading immediates into a GR32, the port prefered LHI, followed by
LLILH or LLILL, followed by IILF. LHI and IILF are natural 32-bit
operations, but LLILH and LLILL also clear the upper 32 bits of the register.
This was represented as taking a 32-bit subreg of a 64-bit assignment.
Using subregs for something as simple as a move immediate was probably
a bad idea. Also, I have patches to add support for the high-word facility,
and we don't want something like LLILH and LLILL to stop the high word of
the same GPR from being used.
This patch therefore uses LHI and IILF to begin with and adds a late
machine-specific pass to use LLILH and LLILL if the other half of the
register is not live. The high-word patches extend this behavior to
IIHF, LLIHL and LLIHH.
No behavioral change intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191363 91177308-0d34-0410-b5e6-96231b3b80d8
This is being disabled because it is no longer needed for
performance. It is only used by postRAscheduler which is also planned
for removal, and it is implemented with an out-dated view of register
liveness. It consideres aliases instead of register units, assumes
valid kill flags, and assumes implicit uses on partial register
defs. Kill flags and implicit operands are error prone and impossible
to verify. We should gradually eliminate dependence on them in the
postRA phases.
Targets that still benefit from this should move to the MI
scheduler. If that doesn't solve the problem, then we should add a
hook to regalloc to optimize reload placement.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191348 91177308-0d34-0410-b5e6-96231b3b80d8
Most constant BUILD_VECTOR's are matched using ComplexPatterns which cover
bitcasted as well as normal vectors. However, it doesn't seem to be possible to
match ldi.[bhwd] in a type-agnostic manner (e.g. to support the widest range of
immediates, it should be possible to use ldi.b to load v2i64) using TableGen so
ldi.[bhwd] is matched using custom code in MipsSEISelDAGToDAG.cpp
This made the majority of the constant splat BUILD_VECTOR lowering redundant.
The only transformation remaining for constant splats is when an (up-to) 32-bit
constant splat is possible but the value does not fit into a 10-bit signed
integer. In this case, the BUILD_VECTOR is transformed into a bitcasted
BUILD_VECTOR so that fill.[bhw] can be used to splat the vector from a GPR32
register (which is initialized using the usual lui/addui sequence).
There are no additional tests since this is a re-implementation of previous
functionality. The change is intended to make it easier to implement some of
the upcoming instruction selection patches since they can rely on existing
support for BUILD_VECTOR's in the DAGCombiner.
compare_float.ll changed slightly because a BITCAST is no longer
introduced during legalization.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191299 91177308-0d34-0410-b5e6-96231b3b80d8
MIPS SelectionDAG changes:
* Added VCEQ, VCL[ET]_[SU] nodes to represent vector comparisons that produce a bitmask.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191286 91177308-0d34-0410-b5e6-96231b3b80d8
Make sure that the code that handles the constant addresses is run for the
GEPs. This just refactors that code and then calls it for the GEPs that are
collected during the iteration.
<rdar://problem/12445434>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191281 91177308-0d34-0410-b5e6-96231b3b80d8
Patch by Ana Pazos.
1.Added support for v1ix and v1fx types.
2.Added Scalar Pairwise Reduce instructions.
3.Added initial implementation of Scalar Arithmetic instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191263 91177308-0d34-0410-b5e6-96231b3b80d8
The recursive nature of the address selection code can cause the stack to
explode if there is a long chain of GEPs. Convert the recursive bit into a
iterative method to avoid this.
<rdar://problem/12445434>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191252 91177308-0d34-0410-b5e6-96231b3b80d8
Changes to MIPS SelectionDAG:
* Added nodes VEXTRACT_[SZ]EXT_ELT to represent extract and extend in a single
operation and implemented the DAG combines necessary to fold sign/zero
extends into the extract.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191199 91177308-0d34-0410-b5e6-96231b3b80d8
An unrelated change crept in because 'svn revert' isn't recursive by default.
The unrelated changes have been reverted.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191193 91177308-0d34-0410-b5e6-96231b3b80d8
Note: There's a later patch on my branch that re-implements this to select
build_vector without the custom SelectionDAG nodes. The future patch avoids
the constant-folding problems stemming from the custom node (i.e. it doesn't
need to re-implement all the DAG combines related to BUILD_VECTOR).
Changes to MIPS specific SelectionDAG nodes:
* Added VSPLAT
This is a special case of BUILD_VECTOR that covers the case the
BUILD_VECTOR is a splat operation.
* Added VSPLATD
This is a special case of VSPLAT that handles the cases when v2i64 is legal
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191191 91177308-0d34-0410-b5e6-96231b3b80d8
Previously, the DAGISel function WalkChainUsers was spotting that it
had entered already-selected territory by whether a node was a
MachineNode (amongst other things). Since it's fairly common practice
to insert MachineNodes during ISelLowering, this was not the correct
check.
Looking around, it seems that other nodes get their NodeId set to -1
upon selection, so this makes sure the same thing happens to all
MachineNodes and uses that characteristic to determine whether we
should stop looking for a loop during selection.
This should fix PR15840.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191165 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
LLVM would crash when trying to come up with a relocation type for
assembly like:
movabsq $V@TPOFF, %rax
Instead, we say the relocation type is R_X86_64_TPOFF64.
Fixes PR17274.
Reviewers: dblaikie, nrieck, rafael
CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1717
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191163 91177308-0d34-0410-b5e6-96231b3b80d8
Pre-increment loads are microcoded on the A2, and the address increment occurs
only after the load completes. As a result, the latency of the GPR address
update is an additional 2 cycles on top of the load latency.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191156 91177308-0d34-0410-b5e6-96231b3b80d8
In AVX 256bit vectors are valid vectors and therefore the Type Legalizer doesn't
split the VSELECT and SETCC nodes. AVX only supports MIN/MAX on 128bit vectors
and this fix enables vector splitting for this special case in the X86 DAG
Combiner.
This fix is related to PR16695, PR17002, and <rdar://problem/14594431>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191131 91177308-0d34-0410-b5e6-96231b3b80d8
The Type Legalizer recognizes that VSELECT needs to be split, because the type
is to wide for the given target. The same does not always apply to SETCC,
because less space is required to encode the result of a comparison. As a result
VSELECT is split and SETCC is unrolled into scalar comparisons.
This commit fixes the issue by checking for VSELECT-SETCC patterns in the DAG
Combiner. If a matching pattern is found, then the result mask of SETCC is
promoted to the expected vector mask for the given target. This mask has usually
te same size as the VSELECT return type (except for Intel KNL). Now the type
legalizer will split both VSELECT and SETCC.
This allows the following X86 DAG Combine code to sucessfully detect the MIN/MAX
pattern. This fixes PR16695, PR17002, and <rdar://problem/14594431>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191130 91177308-0d34-0410-b5e6-96231b3b80d8
The global registry is used to allow command line override of the
scheduler selection, but does not work well as the normal selection
API. For example, the same LLVM process should be able to target
multiple targets or subtargets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191071 91177308-0d34-0410-b5e6-96231b3b80d8
When selecting the DAG (add (WrapperRIP ...), (FrameIndex ...)), X86 code had
spotted the FrameIndex possibility and was working out whether it could fold
the WrapperRIP into this.
The test for forming a %rip version is notionally whether we already have a
base or index register (%rip precludes both), but we were forgetting to account
for the register that would be inserted later to access the frame.
rdar://problem/15024520
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190995 91177308-0d34-0410-b5e6-96231b3b80d8
1) make sure that the first two instructions of the sequence cannot
separate from each other. The linker requires that they be sequential.
If they get separated, it can still work but it will not work in all
cases because the first of the instructions mostly involves the hi part
of the pc relative offset and that part changes slowly. You would have
to be at the right boundary for this to matter.
2) make sure that this sequence begins on a longword boundary.
There appears to be a bug in binutils which makes some of these calculations
get messed up if the instruction sequence does not begin on a longword
boundary. This is being investigated with the appropriate binutils folks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190966 91177308-0d34-0410-b5e6-96231b3b80d8
XCore target: Add XCoreTargetTransformInfo
This is where getNumberOfRegisters() resides, which in turn returns the
number of vector registers (=0).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190936 91177308-0d34-0410-b5e6-96231b3b80d8
For some reason I never got around to adding these at the same time as
the signed versions. No idea why.
I'm not sure whether this SystemZII::BranchC* stuff is useful, or whether
it should just be replaced with an "is normal" flag. I'll leave that
for later though.
There are some boundary conditions that can be tweaked, such as preferring
unsigned comparisons for equality with [128, 256), and "<= 255" over "< 256",
but again I'll leave those for a separate patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190930 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
We indicate that the object files are safe by emitting a @feat.00
absolute address symbol. The address is presumably interpreted as a
bitfield of features that the compiler would like to enable. Bit 0 is
documented in the PE COFF spec to opt in to "registered SEH", which is
what /safeseh enables.
LLVM's object files are safe by default because LLVM doesn't know how to
produce SEH handlers.
Reviewers: Bigcheese
CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1691
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190898 91177308-0d34-0410-b5e6-96231b3b80d8
Documenting a design choice to generate only medium model sequences for TLS
addresses at this time. Small and large code models could be supported if
necessary.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190883 91177308-0d34-0410-b5e6-96231b3b80d8
Large code model on PPC64 requires creating and referencing TOC entries when
using the addis/ld form of addressing. This was not being done in all cases.
The changes in this patch to PPCAsmPrinter::EmitInstruction() fix this. Two
test cases are also modified to reflect this requirement.
Fast-isel was not creating correct code for loading floating-point constants
using large code model. This also requires the addis/ld form of addressing.
Previously we were using the addis/lfd shortcut which is only applicable to
medium code model. One test case is modified to reflect this requirement.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190882 91177308-0d34-0410-b5e6-96231b3b80d8
Add llvm.x86.* intrinsics for all of the Intel SHA Extensions instructions, as
well as tests. Also remove mayLoad and hasSideEffects, which can be inferred
from the instruction patterns.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190864 91177308-0d34-0410-b5e6-96231b3b80d8
Fast-isel generates a COPY_TO_REGCLASS for widening f32 to f64, which
is a nop on PPC64. This is needed to keep the register class system
happy, but on the fast-isel path it is not removed before emit as it
is for DAG select. Ignore this op when emitting instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190795 91177308-0d34-0410-b5e6-96231b3b80d8
The port originally had special patterns for extload, mapping them to the
same instructions as sextload. It seemed neater to have patterns that
match "an extension that is allowed to be signed" and "an extension that
is allowed to be unsigned".
This was originally meant to be a clean-up, but it does improve the handling
of promoted integers a little, as shown by args-06.ll.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190777 91177308-0d34-0410-b5e6-96231b3b80d8
This is a re-commit of r190764, with an extra check to make sure that we're not
performing the transformation on illegal types (a small test case has been
added for this as well).
Original commit message:
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190771 91177308-0d34-0410-b5e6-96231b3b80d8
This is causing test-suite failures.
Original commit message:
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190765 91177308-0d34-0410-b5e6-96231b3b80d8
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190764 91177308-0d34-0410-b5e6-96231b3b80d8
Implements Instruction scheduler latencies for Silvermont,
using latencies from the Intel Silvermont Optimization Guide.
Auto detects SLM.
Turns on post RA scheduler when generating code for SLM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190717 91177308-0d34-0410-b5e6-96231b3b80d8
Previously we modelled VPR128 and VPR64 as essentially identical
register-classes containing V0-V31 (which had Q0-Q31 as "sub_alias"
sub-registers). This model is starting to cause significant problems
for code generation, particularly writing EXTRACT/INSERT_SUBREG
patterns for converting between the two.
The change here switches to classifying VPR64 & VPR128 as
RegisterOperands, which are essentially aliases for RegisterClasses
with different parsing and printing behaviour. This fits almost
exactly with their real status (VPR128 == FPR128 printed strangely,
VPR64 == FPR64 printed strangely).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190665 91177308-0d34-0410-b5e6-96231b3b80d8
When a structure is passed by value, and that structure contains a vector
member, according to the PPC ABI, the structure will receive enhanced alignment
(so that the vector within the structure will always be aligned).
This should resolve PR16641.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190636 91177308-0d34-0410-b5e6-96231b3b80d8
In fast-math mode sqrt(x) is calculated using the fast expansion of the
reciprocal of the reciprocal sqrt expansion. The reciprocal and reciprocal
sqrt expansions use the associated estimate instructions along with some Newton
iterations. Unfortunately, as a result, sqrt(0) was being calculated as NaN,
which is not correct. Now we explicitly return a result of zero if the input is
zero.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190624 91177308-0d34-0410-b5e6-96231b3b80d8
Add basic assembly/disassembly support for the first Intel SHA
instruction 'sha1rnds4'. Also includes feature flag, and test cases.
Support for the remaining instructions will follow in a separate patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190611 91177308-0d34-0410-b5e6-96231b3b80d8
Use the new instruction deprecation feature to mark mftb (now replaced with
mfspr) and dst (along with the other Altivec cache control instructions) as
deprecated when targeting cores supporting at least ISA v2.03.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190605 91177308-0d34-0410-b5e6-96231b3b80d8
The 'Deprecated' class allows you to specify a SubtargetFeature that the
instruction is deprecated on.
The 'ComplexDeprecationPredicate' class allows you to define a custom
predicate that is called to check for deprecation.
For example:
ComplexDeprecationPredicate<"MCR">
would mean you would have to define the following function:
bool getMCRDeprecationInfo(MCInst &MI, MCSubtargetInfo &STI,
std::string &Info)
Which returns 'false' for not deprecated, and 'true' for deprecated
and store the warning message in 'Info'.
The MCTargetAsmParser constructor was chaned to take an extra argument of
the MCInstrInfo class, so out-of-tree targets will need to be changed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190598 91177308-0d34-0410-b5e6-96231b3b80d8
Aggressive anti-dependency breaking is enabled by default for all PPC cores.
This provides a general speedup on the P7 and other platforms (among other
factors, the instruction group formation for the non-embedded PPC cores is done
during post-RA scheduling). In order to do this safely, the incompatibility
between uses of the MFOCRF instruction and anti-dependency breaking are
resolved by marking MFOCRF with hasExtraSrcRegAllocReq. As noted in the removed
FIXME, the problem was that MFOCRF's output is sensitive to the identify of the
source register, and always paired with a shift to undo this effect. Because
anti-dependency breaking is unaware of this hidden dependency of the shift
amount on the source register of the MFOCRF instruction, changing that register
must be inhibited.
Two test cases were adjusted: The SjLj test was made more insensitive to
register choices and scheduling; the saveCR test disabled anti-dependency
breaking because part of what it is testing is proper register reuse.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190587 91177308-0d34-0410-b5e6-96231b3b80d8
For _XYZ, the type of VDATA is v4i32, because v3i32 doesn't exist.
The ADDR64 bit is not exposed. A simpler intrinsic that doesn't take
a resource descriptor might be nicer.
The maximum number of input SGPRs is bumped to 17.
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190575 91177308-0d34-0410-b5e6-96231b3b80d8
This fixes some regressions in the piglit local memory store tests
introduced by recent commits which made the scheduler aware of the trans
slot.
It's not possible to test this using lit, because there is no way to
determine from the assembly dumps whether or not an instruction is in
the trans slot.
Even if this were possible, the test would be highly sensitive to
changes in the scheduler and might generate confusing false negatives.
Reviewed-by: Vincent Lejeune<vljn at ovi.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190574 91177308-0d34-0410-b5e6-96231b3b80d8
As Andy pointed out to me a long time ago, there are no structural hazards in
the later pipeline stages of the A2, and so modeling them is useless. Also,
modeling the top pre-dispatch stages is deceiving because, when multiple
hardware threads are active, those resources are shared among the threads. The
bypass definitions were mostly wrong, and so those have been removed. The
resulting itinerary is much simpler, and more accurate.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190562 91177308-0d34-0410-b5e6-96231b3b80d8
For embedded PPC cores (especially the A2 core), using the MI scheduler with AA
is far superior to the other scheduling options.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190558 91177308-0d34-0410-b5e6-96231b3b80d8
The PowerPC A2 core greatly benefits from aggressive concatenation unrolling;
use the new getUnrollingPreferences to enable this by default when targeting
the PPC A2 core.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190549 91177308-0d34-0410-b5e6-96231b3b80d8
We were figuring out whether to use tPICADD or PICADD, then just using
tPICADD unconditionally anyway. Oops.
A testcase from someone familiar enough with ELF to produce one would
be appreciated. The existing PIC testcase correctly verifies the .s
generated, but that doesn't catch this bug, which only showed up in
direct-to-object mode.
http://llvm.org/bugs/show_bug.cgi?id=17180
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190417 91177308-0d34-0410-b5e6-96231b3b80d8
The main complication here is that TM and TMY (the memory forms) set
CC differently from the register forms. When the tested bits contain
some 0s and some 1s, the register forms set CC to 1 or 2 based on the
value the uppermost bit. The memory forms instead set CC to 1
regardless of the uppermost bit.
Until now, I've tried to make it so that a branch never tests for an
impossible CC value. E.g. NR only sets CC to 0 or 1, so branches on the
result will only test for 0 or 1. Originally I'd tried to do the same
thing for TM and TMY by using custom matching code in ISelDAGToDAG.
That ended up being very ugly though, and would have meant duplicating
some of the chain checks that the common isel code does.
I've therefore gone for the simpler alternative of adding an extra
operand to the TM DAG opcode to say whether a memory form would be OK.
This means that the inverse of a "TM;JE" is "TM;JNE" rather than the
more precise "TM;JNLE", just like the inverse of "TMLL;JE" is "TMLL;JNE".
I suppose that's arguably less confusing though...
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190400 91177308-0d34-0410-b5e6-96231b3b80d8
The work on this project was left in an unfinished and inconsistent state.
Hopefully someone will eventually get a chance to implement this feature, but
in the meantime, it is better to put things back the way the were. I have
left support in the bitcode reader to handle the case-range bitcode format,
so that we do not lose bitcode compatibility with the llvm 3.3 release.
This reverts the following commits: 155464, 156374, 156377, 156613, 156704,
156757, 156804 156808, 156985, 157046, 157112, 157183, 157315, 157384, 157575,
157576, 157586, 157612, 157810, 157814, 157815, 157880, 157881, 157882, 157884,
157887, 157901, 158979, 157987, 157989, 158986, 158997, 159076, 159101, 159100,
159200, 159201, 159207, 159527, 159532, 159540, 159583, 159618, 159658, 159659,
159660, 159661, 159703, 159704, 160076, 167356, 172025, 186736
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190328 91177308-0d34-0410-b5e6-96231b3b80d8
stores, make sure the load or store that accesses the higher half does not have
an alignment that is larger than the offset from the original address.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190318 91177308-0d34-0410-b5e6-96231b3b80d8
IT blocks can only be one instruction lonf, and can only contain a subset of
the 16 instructions.
Patch by Artyom Skrobov!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190309 91177308-0d34-0410-b5e6-96231b3b80d8
Fix XCoreLowerThreadLocal trying to initialise globals
which have no initializer.
Add handling of const expressions containing thread local variables.
These need to be replaced with instructions, as the thread ID is
used to access the thread local variable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190300 91177308-0d34-0410-b5e6-96231b3b80d8
This sidesteps a bug in PrescheduleNodesWithMultipleUses() which
does not check if callResources will be affected by the transformation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190299 91177308-0d34-0410-b5e6-96231b3b80d8
We used to generate the compact unwind encoding from the machine
instructions. However, this had the problem that if the user used `-save-temps'
or compiled their hand-written `.s' file (with CFI directives), we wouldn't
generate the compact unwind encoding.
Move the algorithm that generates the compact unwind encoding into the
MCAsmBackend. This way we can generate the encoding whether the code is from a
`.ll' or `.s' file.
<rdar://problem/13623355>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190290 91177308-0d34-0410-b5e6-96231b3b80d8
precision loads and stores as well as reg+imm double precision loads and stores.
Previously, expansion of loads and stores was done after register allocation,
but now it takes place during legalization. As a result, users will see double
precision stores and loads being emitted to spill and restore 64-bit FP registers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190235 91177308-0d34-0410-b5e6-96231b3b80d8
The architecture has many comparison instructions, including some that
extend one of the operands. The signed comparison instructions use sign
extensions and the unsigned comparison instructions use zero extensions.
In cases where we had a free choice between signed or unsigned comparisons,
we were trying to decide at lowering time which would best fit the available
instructions, taking things like extension type into account. The code
to do that was getting increasingly hairy and was also making some bad
decisions. E.g. when comparing the result of two LLCs, it is better to use
CR rather than CLR, since CR can be fused with a branch while CLR can't.
This patch removes the lowering code and instead adds an operand to
integer comparisons to say whether signed comparison is required,
whether unsigned comparison is required, or whether either is OK.
We can then leave the choice of instruction up to the normal isel code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190138 91177308-0d34-0410-b5e6-96231b3b80d8
If the DAG already has only legal types, then the second round of DAG combines
is skipped. In this case VSELECT+SETCC patterns that match a more efficient
instruction (e.g. min/max) are never recognized.
This fix allows VSELECT+SETCC combines if the types are already legal before DAG
type legalization.
Reviewer: Nadav
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190105 91177308-0d34-0410-b5e6-96231b3b80d8
expression uses an assembler temporary symbol from an assignment. In this case
the symbol does not have a fragment so the use of getFragment() would be NULL
and caused a crash. In the case of an assembler temporary symbol we want to use
the AliasedSymbol (if any) which will create a local relocation entry, but if
it is not an assembler temporary symbol then let it use that symbol with an
external relocation entry.
rdar://9356266
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190096 91177308-0d34-0410-b5e6-96231b3b80d8