Tests updated to explicitly use fast-isel at -O0 instead of implicitly.
This change also allows an explicit -fast-isel option to override an
implicitly enabled global-isel. Otherwise -fast-isel would have no effect at -O0.
Differential Revision: https://reviews.llvm.org/D41362
llvm-svn: 321655
Our internal testing has revealed has discovered bugs in PPC builds.
I have forward reproduction instructions to the original author (Nirav).
llvm-svn: 321649
We can use zmm move with zero masking for this. We already had patterns for using a masked move, but we didn't check for the zero masking case separately.
llvm-svn: 321612
The CONCAT_VECTORS will be lowered to INSERT_SUBVECTOR later. In the modified cases this seems to be enough to trick a later DAG combine into running in a different order than allows the ANDs to be removed.
I'll admit this is a bit of a hack that happens to work, but using CONCAT_VECTORS is more consistent with other legalization code anyway.
llvm-svn: 321611
Don't combine buildvector(binop(),binop(),binop(),binop()) -> binop(buildvector(), buildvector()) if its a splat - keep the binop scalar and just splat the result to avoid large vector constants.
llvm-svn: 321607
We end up using an i8 load via an isel pattern from v8i1 anyway. This just makes it more explicit. This seems to improve codgen in some cases and I'd like to kill off some of the load patterns.
llvm-svn: 321598
The test needs to be changed; it was exercising UB and that likely wasn't the intent of the test author. I simply removed the checks because I have absolutely no idea what this test was trying to accomplish. With multiple check patterns, no explanation, and no familiarity on my part with the ISA a true fix is going to have to come from someone familiar with the target.
llvm-svn: 321591
The test in question was checking for a particular intepretation of undefined behavior. Relax the test to check that we simply don't crash.
Sorry for the breakage, I don't generally build AMDGPU locally and just saw the failure this morning.
llvm-svn: 321589
If the callee and caller use different calling convensions, we cannot apply TCO if the callee requires arguments on stack; e.g. C calling convention and Fast CC use the same registers for parameter passing, but the stack offset is not necessarily same.
This patch also recommit r319218 "[PowerPC] Allow tail calls of fastcc functions from C CallingConv functions." by @sfertile since the problem reported in r320106 should be fixed.
Differential Revision: https://reviews.llvm.org/D40893
llvm-svn: 321579
Initially, if the `c` constraint applied to the wrong data type that
causes LLVM to assert. This commit replaces the assert by an error
message.
llvm-svn: 321565
As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles.
This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles.
By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks.
We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc.
Differential Revision: https://reviews.llvm.org/D38318
llvm-svn: 321553
Revision 320791 introduced a pass that transforms reg+reg instructions to
reg+imm if they're fed by "load immediate". However, it didn't
handle out-of-range shifts correctly as reported in PR35688.
This patch fixes that and therefore the PR.
Furthermore, there was undefined behaviour in the patch where the RHS of an
initialization expression was 32 bits and constant `1` was shifted left 32
bits. This was fixed by ensuring the RHS is 64 bits just like the LHS.
Differential Revision: https://reviews.llvm.org/D41369
llvm-svn: 321551
Previously we used an extend from v8i1 to v8i32/v8i64. Then extracted to the final width. But if we have VLX we should extract first. This way we don't end up with an overly large extend.
This allows us to use vcmpeq to make all ones for the sign extend when DQI isn't available. Otherwise we get a VPTERNLOG.
If we make v2i1/v4i1 legal like proposed in D41560, we could always do this and rely on the lowering of the extend to widen when necessary.
llvm-svn: 321538
The exception handler thunk needs to reference the LSDA of the parent
function, which won't be emitted if it's available_externally.
Fixes PR35736. ThinLTO ends up producing available_externally functions
that use _CxxFrameHandler3.
llvm-svn: 321532
If there are 17 or more leading zeros to the v4i32 elements, then we can use PMADD for the integer multiply when PMULLD is unavailable or slow.
The 17 bits need to be zero as the PMADDWD performs a v8i16 signed-mul-extend + pairwise-add - the upper 16 so we're adding a zero pair and the 17th bit so we don't incorrectly sign extend.
Differential Revision: https://reviews.llvm.org/D41484
llvm-svn: 321516
My original implementation ran as a DAG combine post type legalization, but it turns out we don't run that DAG combine step if type legalization didn't change anything. Attempts to make the combine run before type legalization as well hit other issues.
So just do it in LowerMUL where we can catch more cases.
llvm-svn: 321496
r319980 added new patterns to the machine combiner for transforming (fsub (fmul
x y) z) into (fmla (fneg z) x y). That is, fsub's where the first source
operand is an fmul are transformed. We previously only matched the case where
the second source operand of an fsub was an fmul, transforming (fsub z (fmul x
y)) into (fmls z x y). Now, if we have an fsub where both source operands are
fmuls, both of the above patterns are applicable.
However, the order in which we add the patterns to the list of candidates
determines the transformation that takes place, since only the first pattern
that matches will be used. This patch changes the order these two patterns are
added to the list of candidates such that we prefer the case where the second
source operand is an fmul (the fmls case), rather than the other one (the
fmla/fneg case). When both source operands are fmuls, this ordering results in
fewer instructions.
Differential Revision: https://reviews.llvm.org/D41587
llvm-svn: 321491
For example, float operations may fail to constant fold under certain circumstances (inf/nan/denormal creation etc.)
Reduced from oss-fuzz #4802 test case
llvm-svn: 321488
Normally we catch this during lowering, but vXi64 mul is considered legal when we have AVX512DQ.
This DAG combine allows us to avoid PMULLQ with AVX512DQ if we can prove its unnecessary. PMULLQ is 3 uops that take 4 cycles each. While pmuldq/pmuludq is only one 4 cycle uop.
llvm-svn: 321437
It's nice that pmullq exists, but it has higher latency and probably lower throughput than pmuldq/pmuludq. We should prefer those if we can.
llvm-svn: 321436
Match regular x87 memory fold instructions with load/sideeffects tags, to prevent the schedulers from re-ordering them across the fnstcw/fldcw sequences for truncating stores while they are still pseudo during the stack conversion pass.
llvm-svn: 321424
Previously we extended v2i1 to v2f64 and then tried to use cvtuqq2pd/cvtqq2pd, but that only works with avx512dq. So we ended up scalarizing it. Now we widen to v4i1 first and extend to v4i32.
llvm-svn: 321420
This moves the combine for turning ANDs into shuffle with zero out of SimplifyVBinOps and places it only in visitAND below the reassociate handling. This fixes the specific case I noticed where we failed to combine two ands with constants.
llvm-svn: 321417
This seems to improve X86's ability to match this into an address computation. Otherwise the other operand gets assigned to the base register and the stack pointer + frame index ends up in the index register. But index registers can't encode ESP/RSP so we end up having to move it into another register to meet the constraint.
I could try to improve the address matcher in X86, but swapping the producer seemed easier. Several other places already have the operands in this order so this is at least consistent.
llvm-svn: 321370
Despite what the comment said there isn't better codegen for 512-bit vectors. The 128/256/512 bit implementation jus stores to memory and loads an element. There's no advantage to doing that with a larger size. In fact in many cases it causes a stack realignment and generates worse code.
llvm-svn: 321369
Pointer constants are pretty rare, since we usually represent them as
integer constants and then cast to pointer. One notable exception is the
null pointer constant, which is represented directly as a G_CONSTANT 0
with pointer type. Mark it as legal and make sure it is selected like
any other integer constant.
llvm-svn: 321354
Previously prefetch was only considered legal if sse was enabled, but it should be supported with 3dnow as well.
The prfchw flag now imply at least some form of prefetch without the write hint is available, either the sse or 3dnow version. This is true even if 3dnow and sse are explicitly disabled.
Similarly prefetchwt1 feature implies availability of prefetchw and the the prefetcht0/1/2/nta instructions. This way we can support _MM_HINT_ET0 using prefetchw and _MM_HINT_ET1 with prefetchwt1. And its assumed that if we have levels for the write hint we would have levels for the non-write hint, thus why we enable the sse prefetch instructions.
I believe this behavior is consistent with gcc. I've updated the prefetch.ll to test all of these combinations.
llvm-svn: 321335
This should only affect what we do for v8i16. Previously we went to v8i64, but if we have VLX we only need v8i32. This prevents an unnecessary zmm usage.
llvm-svn: 321303
We should have equally good shuffle options for v8i32 with VLX. This was spotted during my attempts to remove 512-bit vectors from SKX.
We still use 512-bits for v16i1, v32i1, and v64i1. I'm less sure we can handle those well with narrower vectors. i32 and i64 element sizes get the best shuffle support.
llvm-svn: 321291
The knownbits_mask_or_shuffle_uitofp change is interesting - shuffle combines manage to kick in, removing the AND constant mask load. For targets with fast-variable-shuffle this should reduce further to VPOR+VPSHUFB+VCVTDQ2PS.
llvm-svn: 321279
The build failure was caused by an assertion in pre-legalization DAGCombine:
Combining: t6: ppcf128 = uint_to_fp t5
... into: t20: f32 = PPCISD::FCFIDUS t19
which is clearly wrong since ppcf128 are definitely different type with f32 and
we cannot change the node value type when do DAGCombine. The fix is don't
handle ppc_fp128 or i1 conversions in PPCTargetLowering::combineFPToIntToFP and
leave it to downstream to legalize it and expand it to small legal types.
Differential Revision: https://reviews.llvm.org/D41411
llvm-svn: 321276
If the SRL node is only used by an AND, we may be able to set the
ExtVT to the width of the mask, making the AND redundant. To support
this, another check has been added in isLegalNarrowLoad which queries
whether the load is valid.
Differential Revision: https://reviews.llvm.org/D41350
llvm-svn: 321259
If a block has N predecessors, then the current algorithm will try to
sink common code to this block N times (whenever we visit a
predecessor). Every attempt to sink the common code includes going
through all predecessors, so the complexity of the algorithm becomes
O(N^2).
With this patch we try to sink common code only when we visit the block
itself. With this, the complexity goes down to O(N).
As a side effect, the moment the code is sunk is slightly different than
before (the order of simplifications has been changed), that's why I had
to adjust two tests (note that neither of the tests is supposed to test
SimplifyCFG):
* test/CodeGen/AArch64/arm64-jumptable.ll - changes in this test mimic
the changes that previous implementation of SimplifyCFG would do.
* test/CodeGen/ARM/avoid-cpsr-rmw.ll - in this test I disabled common
code sinking by a command line flag.
llvm-svn: 321236
The AArch64 backend contains code to optimize {s,u}{add,sub}.with.overflow during SelectionDAG. This commit ports that code to the ARM backend.
Differential revision: https://reviews.llvm.org/D35635
llvm-svn: 321224
Add support for 'objdump -print-imm-hex' for imm64, operand imm
and branch target. If user programs encode immediate values
as hex numbers, such an option will make it easy to correlate
asm insns with source code. This option also makes it easy
to correlate imm values with insn encoding.
There is one changed behavior in this patch. In old way, we
print the 64bit imm as u64:
O << (uint64_t)Op.getImm();
and the new way is:
O << formatImm(Op.getImm());
The formatImm is defined in llvm/MC/MCInstPrinter.h as
format_object<int64_t> formatImm(int64_t Value)
So the new way to print 64bit imm is i64 type.
If a 64bit value has the highest bit set, the old way
will print the value as a positive value and the
new way will print as a negative value. The new way
is consistent with x86_64.
For the code (see the test program):
...
if (a == 0xABCDABCDabcdabcdULL)
...
x86_64 objdump, with and without -print-imm-hex, looks like:
48 b8 cd ab cd ab cd ab cd ab movabsq $-6067004223159161907, %rax
48 b8 cd ab cd ab cd ab cd ab movabsq $-0x5432543254325433, %rax
Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 321215
Gather/scatter can implicitly sign extend from i32->i64 on indices. So if we know the sign bit of the input to a zext is 0 we can use the implicit extension.
llvm-svn: 321209
This patch turns shuffles of fadd/fsub with fmul into fmsubadd.
Patch by Dmitry Venikov
Differential Revision: https://reviews.llvm.org/D40335
llvm-svn: 321200
We get an assertion in RegBankSelect for code along the lines of
my_32_bit_int = my_64_bit_int, which tends to translate into a 64-bit
load, followed by a G_TRUNC, followed by a 32-bit store. This appears in
a couple of places in the test-suite.
At the moment, the legalizer doesn't distinguish between integer and
floating point scalars, so a 64-bit load will be marked as legal for
targets with VFP, and so will the rest of the sequence, leading to a
slightly bizarre G_TRUNC reaching RegBankSelect.
Since the current support for 64-bit integers is rather immature, this
patch works around the issue by explicitly handling this case in
RegBankSelect and InstructionSelect. In the future, we may want to
revisit this decision and make sure 64-bit integer loads are narrowed
before reaching RegBankSelect.
llvm-svn: 321165
Summary:
Implement lower of unsigned saturation on an interval [0, k] where k + 1 is a power of two using USAT instruction in a similar way to how [~k, k] is lowered using SSAT on ARM models that supports it.
Patch by Marten Svanfeldt
Reviewers: t.p.northover, pbarrio, eastig, SjoerdMeijer, javed.absar, fhahn
Reviewed By: fhahn
Subscribers: fhahn, aemerson, javed.absar, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D41348
llvm-svn: 321164
The gather instruction will implicitly sign extend to the pointer width, we don't need to further extend it. This can prevent unnecessary splitting in some cases.
There's still an issue that lowering on non-VLX can introduce another sign extend that doesn't get combined with shifts from a lowered sign_extend_inreg.
llvm-svn: 321152
This patch fixes a bug in the redundant compare elimination reported in https://reviews.llvm.org/rL320786 and re-enables the optimization.
The redundant compare elimination assumes that we can replace signed comparison with unsigned comparison for the equality check. But due to the difference in the sign extension behavior we cannot change the opcode if the comparison is against an immediate and the most significant bit of the immediate is one.
Differential Revision: https://reviews.llvm.org/D41385
llvm-svn: 321147
Summary:
Extend overlapping store elision to handle overwrites of stores by
larger stores.
Nontemporal tests have been modified to add memory dependencies to
prevent store elision.
Reviewers: craig.topper, rnk, t.p.northover
Subscribers: javed.absar, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D40969
llvm-svn: 321089
We try to prevent shuffle combining to value types that would stop the folding of masked operations, but by just returning early, we were failing to try different shuffle types.
The TODOs are all still relevant here to improve codegen but we're lacking test examples.
llvm-svn: 321085
As mentioned in D38318 and D40865, modern Intel processors prefer to combine multiple shuffles to a variable shuffle mask (PSHUFB/VPERMPS etc.) instead of having multiple stage 'fixed' shuffles which put more pressure on Port 5 (at the expense of extra shuffle mask loads).
This patch provides a FeatureFastVariableShuffle target flag for Haswell+ CPUs that prefers combining 2 or more fixed shuffles to a single variable shuffle (default is 3 shuffles).
The long term aim is to drive more of this from schedule data (probably via the MC) but we're not close to being ready for that yet.
Differential Revision: https://reviews.llvm.org/D41323
llvm-svn: 321074
Extension to D39729 which performed this for vXi16, with the same bit flipping to handle SMAX/SMIN/UMAX cases, vXi8 UMIN horizontal reductions can be performed.
This makes use of the fact that by performing a pair-wise i8 SHUFFLE/UMIN before PHMINPOSUW, we both get the UMIN of each pair but also zero-extend the upper bits ready for v8i16.
Differential Revision: https://reviews.llvm.org/D41294
llvm-svn: 321070
This instruction is encoded as zero, so we have handle that case when checking
for unimplemented opcodes when producing the encoding for an instruction.
llvm-svn: 321066
BWI supports shifting by word amounts. Even if VLX isn't support we can still widen to v32i16 and extract the lower half. For SKX its preferrable to not use 512-bit vector if we can.
llvm-svn: 321059
We need to handle IR for tests that want to do lowering (or just
-stop-after with IR as input). I've run this on one AArch64 test to
demonstrate what it looks like.
llvm-svn: 321048
I missed some prefixes and the fact that on AArch64 we use "bzero"
instead of "__bzero" as on X86 when doing my refactoring in r321035.
Improve tests for bzero.
llvm-svn: 321046
LR was undefined entering outlined functions that contain calls. This made the
machine verifier unhappy when expensive checks were enabled. This fixes that.
llvm-svn: 321014
LLVM IR function names which disable mangling start with '\01'
(https://www.llvm.org/docs/LangRef.html#identifiers).
When an identifier like "\01@abc@" gets dumped to MIR, it is quoted, but
only with single quotes.
http://www.yaml.org/spec/1.2/spec.html#id2770814:
"The allowed character range explicitly excludes the C0 control block
allowed), the surrogate block #xD800-#xDFFF, #xFFFE, and #xFFFF."
http://www.yaml.org/spec/1.2/spec.html#id2776092:
"All non-printable characters must be escaped.
[...]
Note that escape sequences are only interpreted in double-quoted scalars."
This patch adds support for printing escaped non-printable characters
between double quotes if needed.
Should also fix PR31743.
Differential Revision: https://reviews.llvm.org/D41290
llvm-svn: 320996
Fix an off by one error in the bounds checking for 'dinsu' and update
the ranges in the test comments so that they are accurate.
This version has the correct commit message.
Reviewers: atanasyan
Differential Revision: https://reviews.llvm.org/D41183
llvm-svn: 320991
Fix an off by one error in the bounds checking for 'dinsu' and update
the ranges in the test comments so that they are accurate.
Reviewers: atanasyan
https://reviews.llvm.org/D41183
llvm-svn: 320974
For Cylone, the instruction "movi.2d vD, #0" is executed incorrectly in some rare
circumstances. Work around the issue conservatively by avoiding the instruction entirely.
This patch changes CodeGen so that problematic instructions are never
generated, and the AsmParser so that an equivalent instruction is used (with a
warning).
llvm-svn: 320965
Search from AND nodes to find whether they can be propagated back to
loads, so that the AND and load can be combined into a narrow load.
We search through OR, XOR and other AND nodes and all bar one of the
leaves are required to be loads or constants. The exception node then
needs to be masked off meaning that the 'and' isn't removed, but the
loads(s) are narrowed still.
Differential Revision: https://reviews.llvm.org/D41177
llvm-svn: 320962
This is a follow up to the fmaddsub support added in r320950. Hopefully in the future we can fix lowering to handle this fmsubadd too.
llvm-svn: 320951
Summary:
We had no tests for this and we couldn't do the optimization because of a bad use count check. We need to know how many non-undef pieces of the build vector were filled in and ensure our use count is equal to that. But on the shuffle combine version we need the use count to be 2.
The missing coverage was noticed during the review of D40335.
Reviewers: RKSimon, zvi, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41133
llvm-svn: 320950
Assuming we can safely adjust the broadcast index for the new type to keep it suitably aligned, then peek through BITCASTs when looking for the broadcast source.
Fixes PR32007
llvm-svn: 320933
If the loop operand type is int8 then there will be no residual loop for the
unknown size expansion. Dont create the residual-size and bytes-copied values
when they are not needed.
llvm-svn: 320929
Previously we promoted to v8i64, but we don't need to go all the way to 512-bits. If we have VLX we can use the 256-bit instruction. And even if we don't have VLX we can widen v8i32 to v16i32 and drop the upper half.
llvm-svn: 320926
Summary:
Currently we don't handle v32i1/v64i1 insert_vector_elt correctly as we fail to look at the number of elements closely and assume it can only be v16i1 or v8i1.
We also can't type legalize v64i1 insert_vector_elt correctly on KNL due to the type not being byte addressable as required by the legalizing through memory accesses path requires.
For the first issue, the patch now tries to pick a 512-bit register with the correct number of elements and promotes to that.
For the second issue, we now extend the vector to a byte addressable type, do the stores to memory, load the two halves, and then truncate the halves back to the original type. Technically since we changed the type, we may not need two loads, but actually checking that is more work and for the v64i1 case we do need them.
Reviewers: RKSimon, delena, spatel, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40942
llvm-svn: 320849
The original memcpy expansion inserted the loop basic block inbetween
the 2 new basic blocks created by splitting the original block the memcpy
call was in. This commit makes the new memcpy expansion do the same to keep the
layout of the IR matching between the old and new implementations.
Differential Review: https://reviews.llvm.org/D41197
llvm-svn: 320848
Work towards the unification of MIR and debug output by printing
`%stack.0` instead of `<fi#0>`, and `%fixed-stack.0` instead of
`<fi#-4>` (supposing there are 4 fixed stack objects).
Only debug syntax is affected.
Differential Revision: https://reviews.llvm.org/D41027
llvm-svn: 320827
The following CFI directives are suported by MC but not by MIR:
* .cfi_rel_offset
* .cfi_adjust_cfa_offset
* .cfi_escape
* .cfi_remember_state
* .cfi_restore_state
* .cfi_undefined
* .cfi_register
* .cfi_window_save
Add support for printing, parsing and update tests.
Differential Revision: https://reviews.llvm.org/D41230
llvm-svn: 320819
Amara Emerson