On 32-bit targets we can't use the scalar instruction so we
insert the scalar into a vector and use packed conversions.
Previously we used either v4f32->v4i64 or v4f64->v4i64 to avoid
some complexity creating target specific ISD opcodes for
v4f32->v2i64. But this causes extra vzeroupper instructions and
possibly frequency throttling on Intel CPUs.
This patch changes this to create a 128-bit vector and uses a
target specific ISD opcode if needed.
Fix several several additional problems with the int <-> FP conversion
logic both in common code and in the X86 target. In particular:
- The STRICT_FP_TO_UINT expansion emits a floating-point compare. This
compare can raise exceptions and therefore needs to be a strict compare.
I've made it signaling (even though quiet would also be correct) as
signaling is the more usual default for an LT. This code exists both
in common code and in the X86 target.
- The STRICT_UINT_TO_FP expansion algorithm was incorrect for strict mode:
it emitted two STRICT_SINT_TO_FP nodes and then used a select to choose one
of the results. This can cause spurious exceptions by the STRICT_SINT_TO_FP
that ends up not chosen. I've fixed the algorithm to use only a single
STRICT_SINT_TO_FP instead.
- The !isStrictFPEnabled logic in DoInstructionSelection would sometimes do
the wrong thing because it calls getOperationAction using the result VT.
But for some opcodes, incuding [SU]INT_TO_FP, getOperationAction needs to
be called using the operand VT.
- Remove some (obsolete) code in X86DAGToDAGISel::Select that would mutate
STRICT_FP_TO_[SU]INT to non-strict versions unnecessarily.
Reviewed by: craig.topper
Differential Revision: https://reviews.llvm.org/D71840
Summary:
The only useful information the UndefValue conveys is the address space,
which MachinePointerInfo can represent directly without referring to an
IR value.
Reviewers: arsenm, rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, Petar.Avramovic, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D71838
This moves the X86 specific transform from rL364407
into DAGCombiner to generically handle 'little to big' cases
(for example: extract_subvector(v2i64 bitcast(v16i8))). This
allows us to remove both the x86 implementation and the aarch64
bitcast(extract_subvector(bitcast())) combine.
Earlier patches that dealt with regressions initially exposed
by this patch:
rG5e5e99c041e4
rG0b38af89e2c0
Patch by: @RKSimon (Simon Pilgrim)
Differential Revision: https://reviews.llvm.org/D63815
As the extern_weak target might be missing, resolving to the absolute
address zero, we can't use the normal direct PC-relative branch
instructions (as that would result in relocations out of range).
Improve the classifyGlobalFunctionReference method to set
MO_DLLIMPORT/MO_COFFSTUB, and simplify the existing code in
AArch64TargetLowering::LowerCall to use the return value from
classifyGlobalFunctionReference for these cases.
Add code in both AArch64FastISel and GlobalISel/IRTranslator to
bail out for function calls to extern weak functions on windows,
to let SelectionDAG handle them.
This matches what was done for X86 in 6bf108d77a3c.
Differential Revision: https://reviews.llvm.org/D71721
As the extern_weak target might be missing, resolving to the absolute
address zero, we can't use the normal direct PC-relative branch
instructions (as that would result in relocations out of range).
Instead check the shouldAssumeDSOLocal method and load the address
from a COFF stub.
This matches what was done for X86 in 6bf108d77a3c.
Differential Revision: https://reviews.llvm.org/D71720
The custom node PPCISD::XXREVERSE has completely the same semantics of generic node ISD::BSWAP.
We need to clean up it as we have the combine rules for bswap in the base class, while nothing for xxreverse.
Differential Revision: https://reviews.llvm.org/D70657
Summary:
This patch introduces the ROLBRd and RORBRd pseudo-instructions,
which implemenent the "traditional" rotate operations; instead of
the AVR rotate instructions that use the carry bit.
The code is not optimized at all. Especially when dealing with
loops of rotate instructions, this codegen should be improved some
day.
Related bug: 41358 <https://bugs.llvm.org/show_bug.cgi?id=41358>
//Note//: This is my first submitted patch.
Reviewers: dylanmckay, Jim
Reviewed By: dylanmckay
Subscribers: hiraditya, llvm-commits, dylanmckay, dsprenkels
Tags: #llvm
Patched by dsprenkels (Daan Sprenkels)
Differential Revision: https://reviews.llvm.org/D60365
Summary:
Currently, we set legalization action of `ISD::ROTL` vectors as
`Expand` in `PPCISelLowering`. However, we can exploit `vrl(b|h|w|d)`
to lower `ISD::ROTL` directly.
Differential Revision: https://reviews.llvm.org/D71324
Summary:
Without this check unnecessary FMA instructions are generated when the FSUB terms are reused.
This also has the side-effect that the same value is computed to different levels of precision, which can create undesirable effects if the results are used together in subsequent computation.
Reviewers: arsenm, nhaehnle, foad, tpr, dstuttard, spatel
Reviewed By: arsenm
Subscribers: jvesely, wdng, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D71656
Summary:
We noticed in Julia that the sequence below no longer turned into
a sequence of FMA instructions in LLVM 7+, but it did in LLVM 6.
```
%29 = fmul contract <4 x double> %wide.load, %wide.load16
%30 = fmul contract <4 x double> %wide.load13, %wide.load17
%31 = fmul contract <4 x double> %wide.load14, %wide.load18
%32 = fmul contract <4 x double> %wide.load15, %wide.load19
%33 = fadd fast <4 x double> %vec.phi, %29
%34 = fadd fast <4 x double> %vec.phi10, %30
%35 = fadd fast <4 x double> %vec.phi11, %31
%36 = fadd fast <4 x double> %vec.phi12, %32
```
Unlike Clang, Julia doesn't set the `unsafe-fp-math=true` function
attribute, but rather emits more local instruction flags.
This partially undoes https://reviews.llvm.org/D46854 and if required I can try to minimize the test further.
Reviewers: spatel, mcberg2017
Reviewed By: spatel
Subscribers: chriselrod, merge_guards_bot, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D71495
These said test_f32_olt_s for the type of an overloaded intrinsic.
But the parser doesn't use that part of the name and just uses
the types of the arguments.
Summary:
This is documented as the appropriate template modifier for call operands.
Fixes PR44272, and adds a regression test.
Also adds support for operand modifiers in Intel-style inline assembly.
Reviewers: rnk
Reviewed By: rnk
Subscribers: merge_guards_bot, hiraditya, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D71677
This is another potential regression exposed by D63815.
Here we peek through a bitcast to find an extract subvector and
scale the splat offset based on that:
splat (bitcast (extract X, C)), LaneC --> duplane (bitcast X), LaneC'
Differential Revision: https://reviews.llvm.org/D71672
This seems to have been relying on extra spills being inserted in
these blocks to increase the code size to trigger branch
relaxation. This broke when these spills were avoided. Add some asm to
pad the size of the blocks to make it not matter.
Confusingly, the intrinsic operands do not match the
instruction/custom node. The order is shuffled, and the 3rd operand is
an immediate to select operands.
I'm not 100% sure I did this right, but fdiv still doesn't select end
to end and it will be easier to tell when it does. This at least
avoids an assertion in RegBankSelect and allows hitting the fallback
on selection.
This is a purely cosmetic change that is NFC in terms of the binary
output. I bugs me that I called the attribute DW_AT_LLVM_isysroot
since the "i" is an artifact of GCC command line option syntax
(-isysroot is in the category of -i options) and doesn't carry any
useful information otherwise.
This attribute only appears in Clang module debug info.
Differential Revision: https://reviews.llvm.org/D71722
static void *ifunc(void) __attribute__((ifunc("resolver")));
void foo() { ifunc(); }
The relocation produced by the ifunc() call:
1. gcc -msecure-plt -fPIC => R_PPC_PLTREL24 r_addend=0x8000
2. gcc -msecure-plt -PIE => R_PPC_PLTREL24 r_addend=0x8000
3. clang -msecure-plt -fPIC => R_PPC_PLTREL24 r_addend=0x8000
4. clang -msecure-plt -fPIE => R_PPC_REL24
4 is incorrect. The R_PPC_REL24 needs a call stub due to ifunc. If this
relocation is mixed with other R_PPC_PLTREL24(r_addend=0x8000) in a
function, both GNU ld and lld (after D71621 fix) may produce a wrong
result.
This patch fixes 4 to use R_PPC_PLTREL24, which matches GCC.
Both GNU ld and lld (after D71621) will be happy.
Reviewed By: sfertile
Differential Revision: https://reviews.llvm.org/D71649
The setcc operands are copied into LHS and RHS variables at the top of the function. We also capture the condition code.
A later piece of code swaps the operands and changing the CC variable as part of a canonicalization to make some other checks simpler. But we might not make the transform we canonicalized for. So we continue on through the function where we can use the swapped LHS/RHS variables and access the original condition code operand instead of the modified CC variable. This leads to a setcc being created with the original condition code, but with swapped operands.
To mitigate this, this patch does a couple things. The LHS/RHS/CC variables are made const to keep them from being modified like this again. The transform that needs the swap now uses temporary copies of the variables. And the transform that used the original condition code operand has been altered to use the CC variable we cached originally. Either of these changes are enough to fix the issue, but doing both to make this code very safe.
I also considered rewriting the swap code in some way to check both permutations without explicitly swapping or needing temporary variables, but held off on that.
Differential Revision: https://reviews.llvm.org/D71736
Summary: Replace the integer immediate intrisics with splat vector variants so they can be applied as optimizations for the C/C++ intrinsics.
Reviewers: sdesmalen, huntergr, rengolin, efriedma, c-rhodes, mgudim, kmclaughlin
Subscribers: tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, llvm-commits, amehsan
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D71614
The SELR(Mux) instructions can be converted to two-address form as LOCR(Mux)
instructions whenever one of the sources are the same reg as dest. By adding
this mapping in getTwoOperandOpcode(), we get:
- Two-address hints in getRegAllocationHints() for select register
instructions.
- No need anymore for special handling in SystemZShortenInst.cpp -
shortenSelect() removed.
The two-address hints are now added before the GRX32 hints, which should be
preferred.
Review: Ulrich Weigand
https://reviews.llvm.org/D68870
It was recently discovered that the handling of CC values was actually broken
since overflow was not properly handled ('nsw' flag not checked for).
Add and sub instructions now have a new target specific instruction flag
named SystemZII::CCIfNoSignedWrap. It means that the CC result can be used
instead of a compare with 0, but only if the instruction has the 'nsw' flag
set.
This patch also adds the improvements of conversion to logical instructions
and the analyzing of add with immediates, to be able to eliminate more
compares.
Review: Ulrich Weigand
https://reviews.llvm.org/D66868
The back-end currently has special DAGCombine code to detect
cases where two floating-point extend or truncate operations
can be combined into a single vector operation.
This patch extends that support to also handle strict FP operations.
Note that currently only the case where both operations have the
same input chain are supported. This already suffices to cover
the common case where the operations result from scalarizing a
non-legal vector type. More general cases can be supported in
the future.
In general SVE intrinsics are considered predicated and merging
with everything else having suitable decoration. For predicated
zeroing operations (like the predicate logical instructions) we
use the "_z" suffix. After this change all intrinsics use their
expected names (i.e. orr instead of or and eor instead of xor).
I've removed intrinsics and patterns for condition code setting
instructions as that data is not returned as part of the intrinsic.
The expectation is to ask for a cc flag explicitly.
For example:
a = and_z(pg, p1, p2)
cc = ptest_<flag>(pg, a)
With the code generator expected to use "s" variants of instructions
when available.
Differential Revision: https://reviews.llvm.org/D71715
Recommit 23c28c40436143006be740533375c036d11c92cd (reverted in
dcb48f50bdfa0fa47b62d089b6ed999d857fc9f8) with a fix for an assert
"Request for a fixed size on a scalable object" being triggered in
`LowerSVEIntrinsicEXT`. The fix is to call `getKnownMinSize` on the
TypeSize object.
1) Fix an issue with the incorrect value being used for the number of
elements being passed to [d|w]lstp. We were trying to check that
the value was available at LoopStart, but this doesn't consider
that the last instruction in the block could also define the
register. Two helpers have been added to RDA for this.
2) Insert some code to now try to move the element count def or the
insertion point so that we can perform more tail predication.
3) Related to (1), the same off-by-one could prevent us from
generating a low-overhead loop when a mov lr could have been
the last instruction in the block.
4) Fix up some instruction attributes so that not all the
low-overhead loop instructions are labelled as branches and
terminators - as this is not true for dls/dlstp.
Differential Revision: https://reviews.llvm.org/D71609
Record the discovered VPT blocks while checking for validity and, for
now, only handle blocks that begin with VPST and not VPT. We're now
allowing more than one instruction to define vpr, but each block must
somehow be predicated using the vctp. This leaves us with several
scenarios which need fixing up:
1) A VPT block with is only predicated by the vctp and has no
internal vpr defs.
2) A VPT block which is only predicated by the vctp but has an
internal vpr def.
3) A VPT block which is predicated upon the vctp as well as another
vpr def.
4) A VPT block which is not predicated upon a vctp, but contains it
and all instructions within the block are predicated upon in.
The changes needed are, for:
1) The easy one, just remove the vpst and unpredicate the
instructions in the block.
2) Remove the vpst and unpredicate the instructions up to the
internal vpr def. Need insert a new vpst to predicate the
remaining instructions.
3) No nothing.
4) The vctp will be inside a vpt and the instruction will be removed,
so adjust the size of the mask on the vpst.
Differential Revision: https://reviews.llvm.org/D71107
Summary:
When we use undefined symbol with its qualname, we are not able
to generate that symbol because of the logic of early "continue"
that skip the qualname symbol. This patch fixes it.
Differential revision: https://reviews.llvm.org/D71667
