This primarily affects add/fadd/mul/fmul/and/or/xor/pmuludq/pmuldq/max/min/fmaxc/fminc/pmaddwd/pavg.
We already commuted the unmasked and zero masked versions.
I've added 512-bit stack folding tests for most of the instructions
affected. I've tested needing commuting and not commuting across
unmasked, merged masked, and zero masked. The 128/256 bit instructions
should behave similarly.
llvm-svn: 362746
This is intended to enable the use of an immediate blend or
more optimal instruction. But if the passthru is zero we don't
need any additional instructions.
llvm-svn: 362675
One of the sources controls the pass through value for the upper bits
of the result so we can't really commute it.
In practice this problem isn't a functional issue because we would
only try to commute this instruction in order to fold a load. But
we can't do embedded rounding and fold a load at the same time. So
the load fold would never succeed so I don't think we would ever
commute or at least keep the version after commuting.
llvm-svn: 362647
As far as I know these should be freely reassociatable just like
the floating point MAXC/MINC instructions.
The *reduce* test changes are largely regressions and caused by
the "generic" CPU we default to not having a scheduler model.
The machine-combiner-int-vec.ll test shows the positive benefits
of this change.
Differential Revision: https://reviews.llvm.org/D62787
llvm-svn: 362629
As suggested in D62498 - collectConcatOps() matches both
concat_vectors and insert_subvector patterns, and we see
more test improvements by using the more general match.
llvm-svn: 362620
We already handle the case where we combine shuffle(extractsubvector(x),extractsubvector(x)), this relaxes the requirement to permit different sources as long as they have the same value type.
This causes a couple of cases where the VPERMV3 binary shuffles occur at a wider width than before, which I intend to improve in future commits - but as only the subvector's mask indices are defined, these will broadcast so we don't see any increase in constant size.
llvm-svn: 362599
-Use early returns to reduce indentation
-Replace multipe ifs with a switch.
-Replace an assert with an llvm_unreachable default in the switch.
-Check that the FP type we're going to use for the
X86ISD::FAND/FOR/FXOR is legal rather than checking that the
integer type matches the width of a legal scalar fp type. This all
runs after legalization so it shouldn't really matter, but making
sure we're using a valid type in the X86ISD node is really
whats important.
llvm-svn: 362565
We already need to have patterns for X86ISD::RNDSCALE to support software intrinsics. But we currently have 5 sets of patterns for the 5 rounding operations. For of these 6 patterns we have to support 3 vectors widths, 2 element sizes, sse/vex/evex encodings, load folding, and broadcast load folding. This results in a fair amount of bytes in the isel table.
This patch adds code to PreProcessIselDAG to morph the fceil/ffloor/ftrunc/fnearbyint/frint to X86ISD::RNDSCALE. This way we can remove everything, but the intrinsic pattern while still allowing the operations to be considered Legal for DAGCombine and Legalization. This shrinks the DAGISel by somewhere between 9K and 10K.
There is one complication to this, the STRICT versions of these nodes are currently mutated to their none strict equivalents at isel time when the node is visited. This won't be true in the future since that loses the chain ordering information. For now I've also added support for the non-STRICT nodes to Select so we can change the STRICT versions there after they've been mutated to their non-STRICT versions. We'll probably need a STRICT version of RNDSCALE or something to handle this in the future. Which will take us back to needing 2 sets of patterns for strict and non-strict, but that's still better than the 11 or 12 sets of patterns we'd need.
We can probably do something similar for scalar, but I haven't looked at it yet.
Differential Revision: https://reviews.llvm.org/D62757
llvm-svn: 362535
This shows up as a side issue to the main problem for the AVX target example from PR37428:
https://bugs.llvm.org/show_bug.cgi?id=37428 - https://godbolt.org/z/7tpRa3
But as we can see in the pile of existing test diffs, it's actually a widespread problem
that affects any AVX or later target. Apart from a couple of oddballs, I think these are
all improvements for the reasons stated in the code comment: we do not want to enable YMM
unnecessarily (avoid vzeroupper and frequency throttling) and some cores split 256-bit
stores anyway.
We could say that MergeConsecutiveStores() is going overboard on some of these examples,
but that won't solve the problem completely. But that is a reason I'm proposing this as
a lowering rather than a combine: we will infinite loop fighting the merge code if we try
this earlier.
Differential Revision: https://reviews.llvm.org/D62498
llvm-svn: 362524
The proposal in D62498 showed that x86 would benefit from vector
store splitting, but that may conflict with the generic DAG
combiner's store merging transforms.
Add memory type to the existing TLI hook that enables the merging
transforms, so we can limit those changes to scalars only for x86.
llvm-svn: 362507
As discussed on D62777 - we should be able to use this in more SSE41+ cases as well but that requires us to separate it from the OR(AND(),ANDN()) matcher.
llvm-svn: 362504
r362199 fixed it for zero masking, but not zero masking. The load
folding in the peephole pass hid the bug. This patch turns off
the peephole pass on the relevant test to ensure coverage.
llvm-svn: 362440
Move this combine from x86 into generic DAGCombine, which currently only manages cases where the bitcast is between types of the same scalarsize.
Differential Revision: https://reviews.llvm.org/D59188
llvm-svn: 362324
These patterns can incorrectly narrow a volatile load from 128-bits to 64-bits.
Similar to PR42079.
Switch to using (v4i32 (bitcast (v2i64 (scalar_to_vector (loadi64))))) as the
load pattern used in the instructions.
This probably still has issues in 32-bit mode where loadi64 isn't legal. Maybe
we should use VZMOVL for widened loads even when we don't need the upper bits
as zeroes?
llvm-svn: 362203
DAG combine will usually fold fpextend+load to an fp extload anyway. So the
256 and 512 patterns were probably unnecessary. The 128 bit pattern was special
in that it looked for a v4f32 load, but then used it in an instruction that
only loads 64-bits. This is bad if the load happens to be volatile. We could
probably make the patterns volatile aware, but that's more work for something
that's probably rare. The peephole pass might kick in and save us anyway. We
might also be able to fix this with some additional DAG combines.
This also adds patterns for vselect+extload to enabled masked vcvtps2pd to be
used. Previously we looked for the unlikely vselect+fpextend+load.
llvm-svn: 362199
This makes the 5 address operands come first. And the data operand comes last.
This matches the operand order the instruction is created with. It's also the
expected order in X86MCInstLower. So everything appeared to work, but the
operands didn't match their declared type.
Fixes a -verify-machineinstrs failure.
Also remove the isel patterns from these instructions since they should only
be used for stack spills and reloads. I'm not even sure what types the patterns
were looking for to match.
llvm-svn: 362193
The result types aren't mentioned in the pattern name so really shouldn't be in the PatFrags.
The users of these either have their own type constraint or rely on the type constranit system to realize the only legal extend would be to f64.
llvm-svn: 362175
The LoadExt table defaults to all combinations being Legal. For
vector types, only src VTs with an i1 element type were ever changed.
So we don't need to mark them legal manually.
llvm-svn: 362170
We already have good codegen for (vXiY *ext(vXi1 bitcast(iX))) cases, this patch uses it for loads of vXi1 types as well - changing the load into a iX integer load, and bitcasting so that combineToExtendBoolVectorInReg can then use it.
Differential Revision: https://reviews.llvm.org/D62449
llvm-svn: 362081
avoid static check fail
RegClassOrBank is an object of RegClassOrRegBank, which is defined as
using llvm::RegClassOrRegBank = typedef PointerUnion<const
TargetRegisterClass *, const RegisterBank *>
so control flow can not get here. Use ""llvm_unreachable" here to avoid
"null pointer" confusion.
Patch by Shengchen Kan (skan)
Differential Revision: https://reviews.llvm.org/D62006
Signed-off-by: pengfei <pengfei.wang@intel.com>
llvm-svn: 361912
D18885 emitted 5 bytes for call *foo@tlsdesc(%rax). It should use the
2-byte form instead and let R_X86_64_TLSDESC_CALL apply to the beginning
of the call instruction.
The 2-byte form was deliberately chosen to make ->LE and ->IE relaxation work:
0: 48 8d 05 00 00 00 00 lea 0x0(%rip),%rax # 7 <.text+0x7>
3: R_X86_64_GOTPC32_TLSDESC a-0x4
7: ff 10 callq *(%rax)
7: R_X86_64_TLSDESC_CALL a
=>
0: 48 c7 c0 fc ff ff ff mov $0xfffffffffffffffc,%rax
7: 66 90 xchg %ax,%ax
Also change the symbol type to STT_TLS when VK_TLSCALL or VK_TLSDESC is
seen.
Reviewed By: compnerd
Differential Revision: https://reviews.llvm.org/D62512
llvm-svn: 361910
This patch add the ISD::LRINT and ISD::LLRINT along with new
intrinsics. The changes are straightforward as for other
floating-point rounding functions, with just some adjustments
required to handle the return value being an interger.
The idea is to optimize lrint/llrint generation for AArch64
in a subsequent patch. Current semantic is just route it to libm
symbol.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D62017
llvm-svn: 361875
If we don't have VLX then 256-bit SET0 should be lowered
to VPXOR with ZMM registers. This restores functionality
accidentally removed by r309926.
Differential Revision: https://reviews.llvm.org/D62415
llvm-svn: 361843
This shows up as a side issue to the main problem for the AVX target example from PR37428:
https://bugs.llvm.org/show_bug.cgi?id=37428 - https://godbolt.org/z/7tpRa3
But as we can see in the pile of existing test diffs, it's actually a widespread problem
that affects any AVX or later target. Apart from a couple of oddballs, I think these are
all improvements for the reasons stated in the code comment: we do not want to enable YMM
unnecessarily (avoid vzeroupper and frequency throttling) and some cores split 256-bit
stores anyway.
We could say that MergeConsecutiveStores() is going overboard on some of these examples,
but that won't solve the problem completely. But that is the reason I'm proposing this as
a lowering rather than a combine: we will infinite loop fighting the merge code if we try
this earlier.
Differential Revision: https://reviews.llvm.org/D62498
llvm-svn: 361822
Forking this out of the discussion in D62498
(and assuming that will be committed later, so adding the helper function here).
The LangRef says:
"the backend should never split or merge target-legal volatile load/store instructions."
Differential Revision: https://reviews.llvm.org/D62506
llvm-svn: 361815
We were only testing for direct SETCC results - this allows us to peek through AND/OR/XOR combinations of the comparison results as well.
There's a missing SEXT(PACKSS) fold that I need to investigate for v8i1 cases before I can enable it there as well.
llvm-svn: 361716
If we have a known non-nan operand, place it in the second operand
of fmin/fmax that is returned if either operand is nan.
Differential Revision: https://reviews.llvm.org/D62448
llvm-svn: 361704
INC/DEC is really a special case of a more generic issue. We should also turn leas into add reg/reg or add reg/imm regardless of the slow lea flags.
This also supports LEA64_32 which has 64 bit input registers and 32 bit output registers. So we need to convert the 64 bit inputs to their 32 bit equivalents to check if they are equal to base reg.
One thing to note, the original code preserved the kill flags by adding operands to the new instruction instead of using addReg. But I think tied operands aren't supposed to have the kill flag set. I dropped the kill flags, but I could probably try to preserve it in the add reg/reg case if we think its important. Not sure which operand its supposed to go on for the LEA64_32r instruction due to the super reg implicit uses. Though I'm also not sure those are needed since they were probably just created by an INSERT_SUBREG from a 32-bit input.
Differential Revision: https://reviews.llvm.org/D61472
llvm-svn: 361691
This copies the Sandy Bridge zero idiom support to later CPUs. Adding the AVX2 and AVX512F/VL instructions as appropriate.
Differential Revision: https://reviews.llvm.org/D62360
llvm-svn: 361690
This patch adds the overridable TargetLowering::getTargetConstantFromLoad function which allows targets to return any constant value loaded by a LoadSDNode node - only X86 makes use of this so far but everything should be in place for other targets.
computeKnownBits then uses this function to improve codegen, notably vector code after legalization.
A future commit will do the same for ComputeNumSignBits but computeKnownBits sees the bigger benefit.
This required a couple of fixes:
* SimplifyDemandedBits must early-out for getTargetConstantFromLoad cases to prevent infinite loops of constant regeneration (similar to what we already do for BUILD_VECTOR).
* Fix a DAGCombiner::visitTRUNCATE issue as we had trunc(shl(v8i32),v8i16) <-> shl(trunc(v8i16),v8i32) infinite loops after legalization on AVX512 targets.
Differential Revision: https://reviews.llvm.org/D61887
llvm-svn: 361620
Fixes https://bugs.llvm.org/show_bug.cgi?id=40969
The functions findPotentiallyBlockedCopies and buildCopy are currently not
accounting for the presence of debug instructions. In the former this results
in the optimization not being trigerred, and in the latter results in
inconsistent codegen.
This patch enables the optimization to be performed in a debug build and
ensures the codegen is consistent with non-debug builds.
Patch by Chris Dawson.
Differential Revision: https://reviews.llvm.org/D61680
llvm-svn: 361527
In general dynamic/local dynamic TLS models, with -fno-plt,
* x86: emit `calll *___tls_get_addr@GOT(%ebx)` instead of `calll ___tls_get_addr@PLT`
Note, on x86, if we can get rid of %ebx as the PIC register,
it may be better to use a register not preserved across function calls.
* x86_64: emit `callq *__tls_get_addr@GOTPCREL(%rip)` instead of `callq __tls_get_addr@PLT`
Reorganize the code by separating 32-bit and 64-bit.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D62106
llvm-svn: 361453
We effectively had a second set of isel patterns that tried to use a
regular store instruction and an extract_subreg instruction. Or a masked move
and an extract_subreg. These patterns were intended to override the
matching of VEXTRACT instructions by taking advantage of the priority
of the explicit immediate 0 for the index.
This patch instaed just disables the immediate 0 matchin the VEXTRACT
patterns. This each of the component pieces of the larger patterns will
match by themselves.
This found a bug of sorts were we didn't use 128-bit store for 512->128
extract on KNL. Its unclear what the right thing here should be.
Using the vextract avoids constraining the register allocator to use
xmm0-15. But it always results in a longer encoding if the register
allocator ends up choosing xmm0-15 anyway.
llvm-svn: 361431
We were turning roundss/sd/ps/pd intrinsics with immediates of 1 or 2 into
llvm.floor/ceil. The llvm.ceil/floor intrinsics are supposed to correspond
to the libm functions. For the libm functions we need to disable the
precision exception so the llvm.floor/ceil functions should always map to
encodings 0x9 and 0xA.
We had a mix of isel patterns where some used 0x9 and 0xA and others used
0x1 and 0x2. We need to be consistent and always use 0x9 and 0xA.
Since we have no way in isel of knowing where the llvm.ceil/floor came
from, we can't map X86 specific intrinsics with encodings 1 or 2 to it.
We could map 0x9 and 0xA to llvm.ceil/floor instead, but I'd really like
to see a use case and optimization advantage first.
I've left the backend test cases to show the blend we now emit without
the extra isel patterns. But I've removed the InstCombine tests completely.
llvm-svn: 361425
When the tiny code model is requested for a target machine that does not
support this, we get an error message (which is nice) but also this diagnostic
and request to submit a bug report:
fatal error: error in backend: Target does not support the tiny CodeModel
[Inferior 2 (process 31509) exited with code 0106]
clang-9: error: clang frontend command failed with exit code 70 (use -v to see invocation)
(gdb) clang version 9.0.0 (http://llvm.org/git/clang.git 29994b0c63a40f9c97c664170244a7bba5ecc15e) (http://llvm.org/git/llvm.git 95606fdf91c2d63a931e865f4b78b2e9828ddc74)
Target: arm-arm-none-eabi
Thread model: posix
clang-9: note: diagnostic msg: PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace, preprocessed source, and associated run script.
clang-9: note: diagnostic msg:
********************
PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:
Preprocessed source(s) and associated run script(s) are located at:
clang-9: note: diagnostic msg: /tmp/tiny-dfe1a2.c
clang-9: note: diagnostic msg: /tmp/tiny-dfe1a2.sh
clang-9: note: diagnostic msg:
But this is not a bug, this is a feature. :-) Not only is this not a bug, this
is also pretty confusing. This patch causes just to print the fatal error and
not the diagnostic:
fatal error: error in backend: Target does not support the tiny CodeModel
Differential Revision: https://reviews.llvm.org/D62236
llvm-svn: 361370
Fix for https://bugs.llvm.org/show_bug.cgi?id=41971. Make the
combineVectorSizedSetCCEquality() transform more conservative by
checking that the bitcast to the vector type will be cheap/free
for both operands. I'm considering it cheap if it's a constant,
a load or already a vector. I've dropped the explicit check for
f128 because it should fall out naturally (in the cases where
it'd be detrimental).
Differential Revision: https://reviews.llvm.org/D62220
llvm-svn: 361352
CET-IBT enabled
Return-twice functions will indirectly jump after the caller's position.
So when CET-IBT is enable, we should make sure these is endbr*
instructions follow these Return-twice function caller. Like GCC does.
Patch by Xiang Zhang (xiangzhangllvm)
Differential Revision: https://reviews.llvm.org/D61881
llvm-svn: 361342
Refactor DIExpression::With* into a flag enum in order to be less
error-prone to use (as discussed on D60866).
Patch by Djordje Todorovic.
Differential Revision: https://reviews.llvm.org/D61943
llvm-svn: 361137
Same as what we do for vector reductions in combineHorizontalPredicateResult, use movmsk+cmp for scalar (and(extract(x,0),extract(x,1)) reduction patterns.
llvm-svn: 361052
This can be used to create references among sections. When --gc-sections
is used, the referenced section will be retained if the origin section
is retained.
See R_MIPS_NONE (D13659), R_ARM_NONE (D61992), R_AARCH64_NONE (D61973) for similar changes.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D62014
llvm-svn: 360983
These are valid Jcc, but aren't based on the EFLAGS condition codes (Intel 64
and IA-32 Architetcures Software Developer's Manual Vol. 1, Appendix B). These
are covered in clang/test, but not llvm/test.
llvm-svn: 360960
In Intel syntax, it's not uncommon to see a "short" modifier on Jcc conditional
jumps, which indicates the offset should be a "short jump" (8-bit immediate
offset from EIP, -128 to +127). This patch expands to all recognized Jcc
condition codes, and removes the inline restriction.
Clang already ignores "jmp short" in inline assembly. However, only "jmp" and a
couple of Jcc are actually checked, and only inline (i.e., not when using the
integrated assembler for asm sources). A quick search through asm-containing
libraries at hand shows a pretty broad range of Jcc conditions spelled with
"short."
GAS ignores the "short" modifier, and instead uses an encoding based on the
given immediate. MS inline seems to do the same, and I suspect MASM does, too.
NASM will yield an error if presented with an out-of-range immediate value.
Example of GCC 9.1 and MSVC v19.20, "jmp short" with offsets that do and do not
fit within 8 bits: https://gcc.godbolt.org/z/aFZmjY
Differential Revision: https://reviews.llvm.org/D61990
llvm-svn: 360954
This better matches the verbiage in Intel documentation, and should help avoid
confusion between these two different kinds of values, both of which are parsed
from mnemonics.
llvm-svn: 360953
Summary:
This refactors four pieces of code that create SDNodes for references to
symbols:
- normal global address lowering (LEA, MOV, etc)
- callee global address lowering (CALL)
- external symbol address lowering (LEA, MOV, etc)
- external symbol address lowering (CALL)
Each of these pieces of code need to:
- classify the reference
- lower the symbol
- emit a RIP wrapper if needed
- emit a load if needed
- add offsets if needed
I think handling them all in one place will make the code easier to
maintain in the future.
Reviewers: craig.topper, RKSimon
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61690
llvm-svn: 360952
This suppresses exceptions which is what we should be doing for ceil and floor. We already use the correct immediate
in patterns without masking.
llvm-svn: 360915
This patch add the ISD::LROUND and ISD::LLROUND along with new
intrinsics. The changes are straightforward as for other
floating-point rounding functions, with just some adjustments
required to handle the return value being an interger.
The idea is to optimize lround/llround generation for AArch64
in a subsequent patch. Current semantic is just route it to libm
symbol.
llvm-svn: 360889
If we're trying to match an LEA, its possible the LEA match will be deemed unprofitable. In which case the negation we created in matchAddress would be left dangling in the SelectionDAG. This could artificially increase use counts for other nodes in the DAG. Though I don't have an example of that. But it just seems like bad form to have dangling nodes in isel.
Differential Revision: https://reviews.llvm.org/D61047
llvm-svn: 360823
These particular instructions only operate on 128-bit vectors and have no wider equivalents. And the
element size is always known.
One could argue that MOVSS/MOVSD could be merged, but that's probably disruptive to code in
X86ISelLowering and probably low value.
llvm-svn: 360815
They encode the same way, but OR32mi8Locked sets hasUnmodeledSideEffects set
which should be stronger than the mayLoad/mayStore on LOCK_OR32mi8. I think
this makes sense since we are using it as a fence.
This also seems to hide the operation from the speculative load hardening pass
so I've reverted r360511.
llvm-svn: 360747
Move the declarations of getThe<Name>Target() functions into a new header in
TargetInfo and make users of these functions include this new header.
This fixes a layering problem.
llvm-svn: 360736
This was the portion split off D58632 so that it could follow the redzone API cleanup. Note that I changed the offset preferred from -8 to -64. The difference should be very minor, but I thought it might help address one concern which had been previously raised.
Differential Revision: https://reviews.llvm.org/D61862
llvm-svn: 360719
D61068 handled vector shifts, this patch does the same for scalars where there are similar number of pipes for shifts as bit ops - this is true almost entirely for AMD targets where the scalar ALUs are well balanced.
This combine avoids AND immediate mask which usually means we reduce encoding size.
Some tests show use of (slow, scaled) LEA instead of SHL in some cases, but thats due to particular shift immediates - shift+mask generate these just as easily.
Differential Revision: https://reviews.llvm.org/D61830
llvm-svn: 360684
This is a follow on to D58632, with the same logic. Given a memory operation which needs ordering, but doesn't need to modify any particular address, prefer to use a locked stack op over an mfence.
Differential Revision: https://reviews.llvm.org/D61863
llvm-svn: 360649
Returning SDValue() makes the caller think that nothing happened and it will
end up executing the Expand path. This generates extra nodes that will need to
be pruned as dead code.
Returning an ISD::MERGE_VALUES will tell the caller that we'd like to make a
change and it will take care of replacing uses. This will prevent falling into
the Expand path.
llvm-svn: 360627
These are updates to match how isel table would emit a LOCK_OR32mi8 node.
-Use i32 for the immediate zero even though only 8 bits are encoded.
-Use i16 for segment register.
-Use LOCK_OR32mi8 for idempotent atomic operations in 32-bit mode to match
64-bit mode. I'm not sure why OR32mi8Locked and LOCK_OR32mi8 both exist. The
only difference seems to be that OR32mi8Locked is marked as UnmodeledSideEffects=1.
-Emit an extra i32 result for the flags output.
I don't know if the types here really matter just noticed it was inconsistent
with normal behavior.
llvm-svn: 360619
Summary:
X86TargetLowering::LowerAsmOperandForConstraint had better support than
TargetLowering::LowerAsmOperandForConstraint for arbitrary depth
getelementpointers for "i", "n", and "s" extended inline assembly
constraints. Hoist its support from the derived class into the base
class.
Link: https://github.com/ClangBuiltLinux/linux/issues/469
Reviewers: echristo, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, E5ten, kees, jyknight, nemanjai, javed.absar, eraman, hiraditya, jsji, llvm-commits, void, craig.topper, nathanchance, srhines
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61560
llvm-svn: 360604
Fixes the regression noted in D61782 where a VZEXT_MOVL was being inserted because we weren't discriminating between 'zeroable' and 'all undef' for the upper elts.
Differential Revision: https://reviews.llvm.org/D61782
llvm-svn: 360596
Now that we can use HADD/SUB for scalar additions from any pair of extracted elements (D61263), we can relax the one use limit as we will be able to merge multiple uses into using the same HADD/SUB op.
This exposes a couple of missed opportunities in LowerBuildVectorv4x32 which will be committed separately.
Differential Revision: https://reviews.llvm.org/D61782
llvm-svn: 360594
I've included a new fix in X86RegisterInfo to prevent PR41619 without
reintroducing r359392. We might be able to improve that in the base class
implementation of shouldRewriteCopySrc somehow. But this hopefully enables
forward progress on SimplifyDemandedBits improvements for now.
Original commit message:
This patch adds support for BigBitWidth -> SmallBitWidth bitcasts, splitting the DemandedBits/Elts accordingly.
The AMDGPU backend needed an extra (srl (and x, c1 << c2), c2) -> (and (srl(x, c2), c1) combine to encourage BFE creation, I investigated putting this in DAGComb
but it caused a lot of noise on other targets - some improvements, some regressions.
The X86 changes are all definite wins.
llvm-svn: 360552
See if we can simplify the demanded vector elts from the extraction before trying to simplify the demanded bits.
This helps us with target shuffles and hops in particular.
llvm-svn: 360535
The original costs stopped at SSE42, I've added conservative estimates for everything down to SSE1/SSE2 and moved some of the SSE42 costs to SSE41 (really only the addition of PCMPGT makes any difference).
I've also added missing vXi8 costs (we use PHMINPOSUW for i8/i16 for scarily quick results) and 256-bit vector costs for AVX1.
llvm-svn: 360528
For some targets, there is a circular dependency between InstPrinter and
MCTargetDesc. Merging them together will fix this. For the other targets,
the merging is to maintain consistency so all targets will have the same
structure.
llvm-svn: 360484
As requested in D58632, cleanup our red zone detection logic in the X86 backend. The existing X86MachineFunctionInfo flag is used to track whether we *use* the redzone (via a particularly optimization?), but there's no common way to check whether the function *has* a red zone.
I'd appreciate careful review of the uses being updated. I think they are NFC, but a careful eye from someone else would be appreciated.
Differential Revision: https://reviews.llvm.org/D61799
llvm-svn: 360479
After D58632, we can create idempotent atomic operations to the top of stack.
This confused speculative load hardening because it thinks accesses should have
virtual register base except for the cases it already excluded.
This commit adds a new exclusion for this case. I'll try to reduce a test case
for this, but this fix was verified to work by the reporter. This should avoid
needing to revert D58632.
llvm-svn: 360475
Summary: Skip over prefetches when assigning debug info to instructions with memory operands. This way, the debug info is stable after instrumenting a binary with prefetches, allowing for iterative profiling and instrumentation.
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: aprantl, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61789
llvm-svn: 360471
Fixes https://bugs.llvm.org/show_bug.cgi?id=40969
The functions findPotentiallyBlockedCopies and buildCopy are currently not
accounting for the presence of debug instructions. In the former this results
in the optimization not being trigerred, and in the latter results in
inconsistent codegen.
This patch enables the optimization to be performed in a debug build and
ensures the codegen is consistent with non-debug builds.
Patch by Chris Dawson.
Differential Revision: https://reviews.llvm.org/D61680
llvm-svn: 360436
If we only use the lower xmm of a ymm hop, then extract the xmm's (for free), perform the xmm hop and then insert back into a ymm (for free).
Fixes some of the regressions noted in D61782
llvm-svn: 360435
The current lowering uses an mfence. mfences are substaintially higher latency than the locked operations originally requested, but we do want to avoid contention on the original cache line. As such, use a locked instruction on a cache line assumed to be thread local.
Differential Revision: https://reviews.llvm.org/D58632
llvm-svn: 360393
As reported on PR39920, "slow horizontal ops" targets tend to internally expand to 2*shuffle+add/sub - so if we can reduce 2*shuffle+add/sub to a hadd/sub then we should do it - similar port usage but reduced instruction count.
This works out in most cases, although the "PR22377" regression in vector-shuffle-combining.ll is annoying - going from 2*shuffle+add+shuffle to hadd+2*shuffle - I've opened PR41813 to cover this.
Differential Revision: https://reviews.llvm.org/D61308
llvm-svn: 360360
I've started this cleanup more several times now, but got sidetracked
elsewhere, e.g. by llvm-exegesis problems. Not this time, finally!
This is mainly cleaning up the inverse throughput values,
and a few latencies/uops, based on the llvm-exegesis measured values.
Though this is not complete by any means,
there's certainly more cleanup to be done.
The performance numbers (i've only checked by RawSpeed benchmark) aren't
really surprising - overall this *slightly* (< -1%) improves perf.
llvm-svn: 360341
This code was never covered by tests, in PR41786 it was pointed out that
the deletion part doesn't work, and in a full Chrome build I was never
able to hit the code path that looks through copies. It seems the
situation it's supposed to handle doesn't actually come up in practice.
Delete it to simplify the code.
Differential revision: https://reviews.llvm.org/D61671
llvm-svn: 360320
Summary:
A COFF stub indirects the reference to a symbol through memory. A
.refptr.$sym global variable pointer is created to refer to $sym.
Typically mingw uses these for external global variable declarations,
but we can use them for weak function declarations as well.
Updates the dso_local classification to add a special case for
extern_weak symbols on COFF in both clang and LLVM.
Fixes PR37598
Reviewers: smeenai, mstorsjo
Subscribers: hiraditya, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D61615
llvm-svn: 360207
Basic "revectorization" combine, we can probably do more opcodes here but it can be a tricky cost-benefit depending on where the subvectors came from - but this case helps shuffle combining.
llvm-svn: 360134
The FR32/FR64/VR128/VR256 register classes don't contain the upper 16 registers. For most cases we use the default implementation which will find any register class that contains the register in question if the VT is legal for the register class. But if the VT is i32 or i64, we won't find a matching register class and will instead up in the code modified in this patch.
If the requested register is x/y/zmm16-31 we weren't returning a register class that contains those registers and will hit an assertion in the caller.
To fix this, I've changed to use the extended register class instead. I don't believe we need a subtarget check to see if avx512 is enabled. The default implementation just pick whatever register class it finds first. I checked and we currently pick FR32X for XMM0 with an f32 type using the default implementation regardless of whether avx512 is enabled. So I assume its it is ok to do the same for i32.
Differential Revision: https://reviews.llvm.org/D61457
llvm-svn: 360102
We require d/q suffixes on the memory form of these instructions to disambiguate the memory size.
We don't require it on the register forms, but need to support parsing both with and without it.
Previously we always printed the d/q suffix on the register forms, but it's redundant and
inconsistent with gcc and objdump.
After this patch we should support the d/q for parsing, but not print it when its unneeded.
llvm-svn: 360085
Reverts "[X86] Remove (V)MOV64toSDrr/m and (V)MOVDI2SSrr/m. Use 128-bit result MOVD/MOVQ and COPY_TO_REGCLASS instead"
Reverts "[TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling"
Eric Christopher and Jorge Gorbe Moya reported some issues with these patches to me off list.
Removing the CodeGenOnly instructions has changed how fneg is handled during fast-isel with sse/sse2. We're now emitting fsub -0.0, x instead
moving to the integer domain(in a GPR), xoring the sign bit, and then moving back to xmm. This is because the fast isel table no longer
contains an entry for (f32/f64 bitcast (i32/i64)) so the target independent fneg code fails. The use of fsub changes the behavior of nan with
respect to -O2 codegen which will always use a pxor. NOTE: We still have a difference with double with -m32 since the move to GPR doesn't work
there. I'll file a separate PR for that and add test cases.
Since removing the CodeGenOnly instructions was fixing PR41619, I'm reverting r358887 which exposed that PR. Though I wouldn't be surprised
if that bug can still be hit independent of that.
This should hopefully get Google back to green. I'll work with Simon and other X86 folks to figure out how to move forward again.
llvm-svn: 360066
scan-build was reporting that CommutableOpIdx1 never used its original initialized value - move it down to where its first used to make the real initialization more obvious (and matches the comment that's there).
llvm-svn: 360028
Summary:
1. Enable infrastructure of AVX512_BF16, which is supported for BFLOAT16 in Cooper Lake;
2. Enable VCVTNE2PS2BF16, VCVTNEPS2BF16 and DPBF16PS instructions, which are Vector Neural Network Instructions supporting BFLOAT16 inputs and conversion instructions from IEEE single precision.
VCVTNE2PS2BF16: Convert Two Packed Single Data to One Packed BF16 Data.
VCVTNEPS2BF16: Convert Packed Single Data to Packed BF16 Data.
VDPBF16PS: Dot Product of BF16 Pairs Accumulated into Packed Single Precision.
For more details about BF16 isa, please refer to the latest ISE document: https://software.intel.com/en-us/download/intel-architecture-instruction-set-extensions-programming-reference
Author: LiuTianle
Reviewers: craig.topper, smaslov, LuoYuanke, wxiao3, annita.zhang, RKSimon, spatel
Reviewed By: craig.topper
Subscribers: kristina, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60550
llvm-svn: 360017
The x/y/z suffix is needed to disambiguate the memory form in at&t syntax since no xmm/ymm/zmm register is mentioned.
But we should also allow it for the register and broadcast forms where its not needed for consistency. This matches gas.
The printing code will still only use the suffix for the memory form where it is needed.
llvm-svn: 359903
The default impementation in the base class for TargetLowering::getRegForInlineAsmConstraint doesn't work for mask registers when the VT is a scalar type integer types since the only legal mask types are vXi1. So we end up just getting whatever the first register class that contains the register. Currently this appears to be VK1, but its really dependent on the order tablegen outputs the register classes.
Some code in the caller ends up looking up the type for this register class and find v1i1 then generates a copyfromreg from the physical k-register with the v1i1 type. Then it generates an any_extend from v1i1 to the scalar VT which isn't legal. This bad any_extend sticks around until isel where it selects a MOVZX32rr8 with a v1i1 input or maybe a i8 input. Not sure but eventually we pick up a copy from VK1 to GR8 in MachineIR which isn't supported. This leads to a failure in physical register copying.
This patch uses the scalar type to find a VK class of the right size. In the attached test case this will be VK16. This causes a bitcast from vk16 to i16 to be generated instead of an any_extend. This will be properly iseled to a VK16 to GR32 copy and a GR32->GR16 extract_subreg.
Fixes PR41678
Differential Revision: https://reviews.llvm.org/D61453
llvm-svn: 359837
Limiting scalar hadd/hsub generation to the lowest xmm looks to be unnecessary - we will be extracting one upper xmm whatever, and we can remove a shuffle by using the hop which is inline with what shouldUseHorizontalOp expects to happen anyway.
Testing on btver2 (the main target for fast-hops) shows this is beneficial even for float ops where we have a 'shuffle' to extract the float result:
https://godbolt.org/z/0R-U-K
Differential Revision: https://reviews.llvm.org/D61426
llvm-svn: 359786
The broadcasting variant for instruction vfpclassp[d,s] shouldn't use suffix q/l. So remove them from the template.
Patch by Pengfei Wang
Differential Revision: https://reviews.llvm.org/D61295
llvm-svn: 359753
We already perform horizontal add/sub if we extract from elements 0 and 1, this patch extends it to non-0/1 element extraction indices (as long as they are from the lowest 128-bit vector).
Differential Revision: https://reviews.llvm.org/D61263
llvm-svn: 359707
Pengfei Wang