The VINSERTF128 instruction is often much quicker, and never slower, than the more general VPERM2F128 instruction, so we should try to use that in more circumstances.
This requires a fallback to a commuted VPERM2F128 for the case where we need to fold the 256-bit vector source instead of the 128-bit subvector source.
There is one interesting side effect - DAGCombine's narrowExtractedVectorLoad combine gets called in a number of locations, this often creates an extracted subvector load without regard to other uses of the original wider load. I'm expecting AVX cpus to be capable of merging such aliased loads, but I do wonder whether narrowExtractedVectorLoad's call to X86TargetLowering::shouldReduceLoadWidth needs to be altered to check for more partial uses?
Noticed while investigating the quality of interleaved load/store codegen.
Differential Revision: https://reviews.llvm.org/D111960
Ooops. It constructs codegen infra and provide only basic code to generate first add instruction successfully.
Differential Revision: https://reviews.llvm.org/D112206
Register operands with superclasses can possibly have multiple regBanks
if they have different register types. The regBank ambiguity resolved
during regbankselect should be used to constrain the operand regclass
instead of obtaining one from the MCInstrDesc.
This is a prerequisite patch for D109300 that introduces allocatable AV_*
Superclasses for AMDGPU by combining both VGPRs and AGPRs and we want to
restrain the regclass to either A or V based on the incoming regbank.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D112323
With Global ISel getReservedRegs() is called before function is
regbank selected for the first time. Defer caching of usesAGPRs()
in this case.
Differential Revision: https://reviews.llvm.org/D112644
These compiler-rt-only symbols aren't available in libgcc. Similar to
D108842, D108844, and D108926.
Fixes: pr/52043
Reviewed By: craig.topper, rengolin
Differential Revision: https://reviews.llvm.org/D112750
The tests are based on the example from:
https://llvm.org/PR52032
I suspect that it looks worse than it actually is. :)
That is, llvm-mca says there's no uop/timing difference with the
load folding and pcmpeq vs. broadcast on Haswell (and probably
other targets).
The load-folding definitely makes the code smaller, so it's good
for that at least. So this requires carving a narrow hole in the
transform to get just this case without changing others that look
good as-is (in other words, the transform still seems good for
most examples).
Differential Revision: https://reviews.llvm.org/D112464
`classifyLocalReference(nullptr)` is called to get the appropriate
relocation type for jump tables. We should not use @GOTPCREL for this
case.
The new test cases trigger assertions without this patch.
Reviewed By: eugenis
Differential Revision: https://reviews.llvm.org/D112832
If the VL operand of a mask register instruction comes from an
explicit vsetvli with a different VTYPE, we can still avoid needing
a vsetvli as long as the SEW/LMUL ratio is the same and policy bits
match.
Differential Revision: https://reviews.llvm.org/D112762
Add support for generating TargetFrameIndex in complex patterns for
indexed addressing modes in SVE. Additionally, add missing load/stores
to getMemOpInfo and getLoadStoreImmIdx.
Differential Revision: https://reviews.llvm.org/D112617
createReplacementInstr was a trivial wrapper around
ConstantExpr::getAsInstruction, which also inserted the new instruction
into a basic block. Implement this directly in getAsInstruction by
adding an InsertBefore parameter and change all callers to use it. NFC.
A follow-up patch will remove createReplacementInstr.
Differential Revision: https://reviews.llvm.org/D112791
Change numBitsSigned to return the minimum size of a signed integer that
can hold the value. This is different by one from the previous result
but is more consistent with numBitsUnsigned. Update all callers. All
callers are now more consistent between the signed and unsigned cases,
and some callers get simpler, especially the ones that deal with
quantities like numBitsSigned(LHS) + numBitsSigned(RHS).
Differential Revision: https://reviews.llvm.org/D112813
As mentioned on D108539, when the gather indices are smaller than the pointer size, they are sign-extended BEFORE scale is applied, making the general fold unsafe.
If the index have sufficient sign-bits then folding the scale could be safe - I'll investigate this.
This adds support for SVE structured loads/stores to the relevant target
hooks, such that we can support these instructions in the InterleavedAccess
pass.
Depends on D112078
Differential Revision: https://reviews.llvm.org/D112303
Shift node is still needed to check if the shift is shr or shl to increment/decrement offset. Do not override the node.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D112733
Add i32x4.relaxed_trunc_f32x4_s, i32x4.relaxed_trunc_f32x4_u,
i32x4.relaxed_trunc_f64x2_s_zero, i32x4.relaxed_trunc_f64x2_u_zero.
These are only exposed as builtins, and require user opt-in.
Differential Revision: https://reviews.llvm.org/D112186
This is relanding commit da1d1a0869 .
This patch additionally addresses failures found in buildbots & post review comments.
ARM EHABI[1] specifies the __cxa_end_cleanup to be called after cleanup.
It will call the UnwindResume.
__cxa_begin_cleanup will be called from libcxxabi while __cxa_end_cleanup is never called.
This will trigger a termination when a foreign exception is processed while UnwindResume is called
because the global state will be wrong due to the missing __cxa_end_cleanup call.
Additional test here: D109856
[1] https://github.com/ARM-software/abi-aa/blob/main/ehabi32/ehabi32.rst#941compiler-helper-functions
Reviewed By: logan
Differential Revision: https://reviews.llvm.org/D111703
timm isn't the common case, and TImmLeafs should make it clear what
they are. We're adding a plain ImmLeaf for 0_65535, so rename
i64_imm0_65535 to timm64_0_65535, and imm32_0_7 to timm32_0_7.
This extends `optimizeCompareInstr` to re-use previous comparison
results if the previous comparison was with an immediate that was 1
bigger or smaller. Example:
CMP x, 13
...
CMP x, 12 ; can be removed if we change the SETg
SETg ... ; x > 12 changed to `SETge` (x >= 13) removing CMP
Motivation: This often happens because SelectionDAG canonicalization
tends to add/subtract 1 often when optimizing for fallthrough blocks.
Example for `x > C` the fallthrough optimization switches true/false
blocks with `!(x > C)` --> `x <= C` and canonicalization turns this into
`x < C + 1`.
Differential Revision: https://reviews.llvm.org/D110867
`X86InstrInfo::optimizeCompareInstr` would only optimize a `SUB`
followed by a `CMP` in `isRedundantFlagInstr`. This extends the code to
also look for other combinations like `CMP`+`CMP`, `TEST`+`TEST`, `SUB
x,0`+`TEST`.
- Change `isRedundantFlagInstr` to run `analyzeCompareInstr` on the
candidate instruction and compare the results. This normalizes things
and gives consistent results for various comparisons (`CMP x, y`,
`SUB x, y`) and immediate cases (`TEST x, x`, `SUB x, 0`,
`CMP x, 0`...).
- Turn `isRedundantFlagInstr` into a member function so it can call
`analyzeCompare`. - We now also check `isRedundantFlagInstr` even if
`IsCmpZero` is true, since we now have cases like `TEST`+`TEST`.
Differential Revision: https://reviews.llvm.org/D110865
This is relanding commit da1d1a0869 .
This patch additionally addresses failures found in buildbots & post review comments.
ARM EHABI[1] specifies the __cxa_end_cleanup to be called after cleanup.
It will call the UnwindResume.
__cxa_begin_cleanup will be called from libcxxabi while __cxa_end_cleanup is never called.
This will trigger a termination when a foreign exception is processed while UnwindResume is called
because the global state will be wrong due to the missing __cxa_end_cleanup call.
Additional test here: D109856
[1] https://github.com/ARM-software/abi-aa/blob/main/ehabi32/ehabi32.rst#941compiler-helper-functions
Reviewed By: logan
Differential Revision: https://reviews.llvm.org/D111703
If the index operand for a gather/scatter intrinsic is being scaled (self-addition or a shl-by-immediate) then we may be able to fold that scaling into the intrinsic scale immediate value instead.
Fixes PR13310.
Differential Revision: https://reviews.llvm.org/D108539
mul24 intrinsic's operands are simplified by
AMDGPUTargetLowering::performIntrinsicWOChainCombine(). This change adds
the mul24hi intrinsics in the combine since its operands can be
simplified like that of the mul24 intrinsics.
Differential Revision: https://reviews.llvm.org/D112702
- Move the `s_and exec` to its correct position before the content of
the waterfall loop
- Use the SI_WATERFALL pseudo instruction, like for sdag, to benefit
from optimizations
- Add support for indirect function calls
To support indirect calls, add a G_SI_CALL instruction without register
class restrictions and insert a waterfall loop when applying register
banks.
Differential Revision: https://reviews.llvm.org/D109052
The support for neoverse-512tvb mirrors the same option available in GCC[1].
There is no functional effect for this option yet.
This patch ensures the driver accepts "-mcpu=neoverse-512tvb", and enough
plumbing is in place to allow the new option to be used in the future.
[1]https://gcc.gnu.org/onlinedocs/gcc/AArch64-Options.html
Differential Revision: https://reviews.llvm.org/D112406
Sync the order of Zvlsseg registers with vector registers to avoid
unnecessary register copies between vector instructions and zvlsseg
instructions.
Differential Revision: https://reviews.llvm.org/D110250
In the instruction encoding, the passthru register is always
tied to the destination register. The CPU scheduler has to wait
for the last writer of this register to finish executing before
the gather can start. This is true even if the initial mask is
all ones so that the passthru will never be used.
By explicitly zeroing the register we can break the false
dependency. The zero idiom is executed completing by the
register renamer and so is immedately considered ready.
Authored by Craig.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D112505
If we know the source operand of COPY is defined by a vector instruction
with tail agnostic and the same LMUL and there is no vsetvli between
COPY and the define instruction to change the vl and vtype, we could use
vmv.v.v or vmv.v.i to copy vector registers to get better performance than
the whole vector register move instructions.
If the source of COPY is from vmv.v.i, we could use vmv.v.i for the
COPY.
This patch only considers all these instructions within one basic block.
Case 1:
```
bb.0:
...
VSETVLI # The first VSETVLI before COPY and VOP.
... # Use this VSETVLI to check LMUL and tail agnostic.
...
vy = VOP va, vb # Define vy.
... # There is no vsetvli between VOP and COPY.
vx = COPY vy
```
Case 2:
```
bb.0:
...
VSETVLI # The first VSETVLI before VOP.
... # Use this VSETVLI to check LMUL and tail agnostic.
...
vy = VOP va, vb # Define vy.
... # There is no vsetvli to change vl between VOP and COPY.
...
VSETVLI # The first VSETVLI before COPY.
... # This VSETVLI does not change vl and vtype.
...
vx = COPY vy
```
Co-Authored-by: Zakk Chen <zakk.chen@sifive.com>
Co-Authored-by: Kito Cheng <kito.cheng@sifive.com>
Differential Revision: https://reviews.llvm.org/D103510
Add new hasVInstructions() which is currently equivalent.
Replace vector uses of hasStdExtZfh/F/D with new vector specific
versions. The vector spec no longer requires that the vectors implement the
same types as scalar. It only requires that the scalar type is
the maximum size the vectors can support. This is currently
implemented using the scalar rule we were using before.
Add new hasVInstructionsI64() begin using to qualify code that
requires i64 vector elements.
This is all NFC for now, but we can start using this to better
implement D112408 which introduces the Zve extensions.
Reviewed By: frasercrmck, eopXD
Differential Revision: https://reviews.llvm.org/D112496
In ARM mode, passing -mtp=cp15 forces the use of an inline MRC system register read to move the thread pointer value into a register.
Currently, in Thumb2 mode, -mtp=cp15 is ignored, and a call to the __aeabi_read_tp helper is emitted instead.
This is inconsistent, and breaks the Linux/ARM build for Thumb2 targets, as the Linux kernel does not provide an implementation of __aeabi_read_tp,.
Reviewed By: nickdesaulniers, peter.smith
Differential Revision: https://reviews.llvm.org/D112600
- When an unconditional branch is expanded into an indirect branch, if
there is no scavenged register, an SGPR pair needs spilling to enable
the destination PC calculation. In addition, before jumping into the
destination, that clobbered SGPR pair need restoring.
- As SGPR cannot be spilled to or restored from memory directly, the
spilling/restoring of that SGPR pair reuses the regular SGPR spilling
support but without spilling it into memory. As that spilling and
restoring points are fully controlled, we only need to spill that SGPR
into the temporary VGPR, which needs spilling into its emergency slot.
- The target-specific hook is revised to take additional restore block,
where the restoring code is filled. After that, the relaxation will
place that restore block directly before the destination block and
insert an unconditional branch in any fall-through block into the
destination block.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D106449
As noted in D112464, a pre-AVX target may not be able to fold an
under-aligned vector load into another op, so we shouldn't report
that as a load folding candidate. I only found one caller where
this would make a difference -- combineCommutableSHUFP() -- so
that's where I added a test to show the (minor) regression.
Differential Revision: https://reviews.llvm.org/D112545
Combine FADD and FMUL intrinsics into FMA when the result of the FMUL is an FADD operand
with one only use and both use the same predicate.
Differential Revision: https://reviews.llvm.org/D111638
This change introduces subtarget features to predicate certain
instructions and system registers that are available only on
'A' profile targets. Those features are not present when
targeting a generic CPU, which is the default processor.
In other words the generic CPU now means the intersection of
'A' and 'R' profiles. To maintain backwards compatibility we
enable the features that correspond to -march=armv8-a when the
architecture is not explicitly specified on the command line.
References: https://developer.arm.com/documentation/ddi0600/latest
Differential Revision: https://reviews.llvm.org/D110065
ARM EHABI[1] specifies the __cxa_end_cleanup to be called after cleanup.
It will call the UnwindResume.
__cxa_begin_cleanup will be called from libcxxabi while __cxa_end_cleanup is never called.
This will trigger a termination when a foreign exception is processed while UnwindResume is called
because the global state will be wrong due to the missing __cxa_end_cleanup call.
Additional test here: D109856
[1] https://github.com/ARM-software/abi-aa/blob/main/ehabi32/ehabi32.rst#941compiler-helper-functions
Reviewed By: logan
Differential Revision: https://reviews.llvm.org/D111703
Simplify "LUI+SLLI+ADDI+SLLI" and "LUI+ADDIW+SLLI+ADDI+SLLI" to
"LUI+ADDIW+SLLIUW" to reduce total instruction amount.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D111933
The scheduler should set critical/excess register usage thresholds that
are guided by the maximum possible occupancy for the function. This
change is focused on setting proper lower bounds on register usage which
we would typically only see when a specific number of maximum waves is
requested with the "waves-per-eu" attribute, or by setting
"amdgpu-num-vgpr|sgpr" directly. This was broken previously. I have a
follow-on patch that will address issues with the scheduler not
targeting correct upper bounds on register usage which is typical with
launch bounds and min "waves-per-eu".
Changes by this patch:
Set the initial critical register usage thresholds to minimum values
that are determined by the maximum possible occupancy for the function,
or the number of allocatable registers, whichever is lower.
Avoid unisgned overflow if register limits are lower than the register
tracking "ErrorMargin", I.e. when using stress-regalloc=2.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D112373
It was discovered that an extra register COPY remained when expanding a
(variable length) memory operation with a loop and there was another use of
the involved address register(s) afterwards.
A simple fix for this is to COPY the address registers before the loop and
use that new vreg instead.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D112065
We were bailing out of creating 24-bit muls for results wider than 32
bits in AMDGPUCodeGenPrepare. With the 24-bit mulhi intrinsic, this
change teaches AMDGPUCodeGenPrepare to generate the 48-bit mul
correctly.
Differential Revision: https://reviews.llvm.org/D112395
These intrinsics maps to the 24-bit v_mul_hi instructions.
This change also fixes an incorrect assumption on the associativity of
24-bit mulhi in its SDNode record in tblgen.
Differential Revision: https://reviews.llvm.org/D112394
The combine asserted if constants could not be represented as uint64_t.
Use APInts to fix this.
Differential Revision: https://reviews.llvm.org/D112416
All instructions must have a correct size value close to emission when
SystemZLongBranch runs, or a necessary branch relaxation may be missed.
This patch also adds an assert for instruction sizes in SystemZLongBranch.
Review: Ulrich Weigand
This can avoid a vector add and a constant pool load. Or an explicit broadcast in case of non-constant.
Also reverse the transform any time we encounter a constant index addend that can't be moved to base. In that case pull the constant from base into the index. This reduces code size needed for the displacement since we needed the index add anyway. Limit this to scale of 1 to avoid divisibility and wrap issues.
Authored by Craig.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D111595
Removed references to `sanity check` in `PPCBranchCoalescing.cpp` code comments.
No word substitution made in this case, as the comments and code following illustrated are
sufficient IMO.
Reviewed By: quinnp
Differential Revision: https://reviews.llvm.org/D112452
All but 2 of the vector builtins are only used by clang_builtin_alias.
When using clang_builtin_alias, the type string of the builtin is never
checked. Only the types in the function definition used for the alias
are checked.
This patch takes advantage of this to share a single builtin for
many different types. We already used type overloads on the IR intrinsic
so the codegen for the builtins that are being merge were already
the same. This extends the type overloading to the builtins.
I had to make a few tweaks to make this work.
-Floating point vector-vector vmerge now uses the vmerge intrinsic
instead of the vfmerge intrinsic. New isel patterns and tests are
added to support this.
-The SemaChecking for the immediate of vset_v/vget_v has been removed.
Determining the valid range is harder now. I've added masking to
ManualCodegen to ensure valid IR for invalid input.
This reduces the number of builtins from ~25000 to ~1100.
Reviewed By: HsiangKai
Differential Revision: https://reviews.llvm.org/D112102
Before the code would crash with "unhandled opcode in
isAArch64FrameOffsetLegal" when there was a spill from extractelement.
Fixes pr52249
Differential Revision: https://reviews.llvm.org/D112311
This patch enables the use of reciprocal estimates for SVE
when both the -Ofast and -mrecip flags are used.
Reviewed By: david-arm, paulwalker-arm
Differential Revision: https://reviews.llvm.org/D111657
%3:gpr32 = ORRWrs $wzr, %2, 0
%4:gpr64 = SUBREG_TO_REG 0, %3, %subreg.sub_32
If AArch64's 32-bit form of instruction defines the source operand of ORRWrs,
we can remove the ORRWrs because the upper 32 bits of the source operand are
set to zero.
Differential Revision: https://reviews.llvm.org/D110841
Add a new preparation pattern in PPCLoopInstFormPrep pass to reduce register
pressure.
Reviewed By: jsji
Differential Revision: https://reviews.llvm.org/D108750
This extends `optimizeCompareInstr` to continue the backwards search
when it reached the beginning of a basic block. If there is a single
predecessor block then we can just continue the search in that block and
mark the EFLAGS register as live-in.
Differential Revision: https://reviews.llvm.org/D110862
This changes the first part of `optimizeCompareInstr` being split into a
loop with a forward scan for cases that re-use zero flags from a
producer in case of compare with zero and a backward scan for finding an
instruction equivalent to a compare.
The code now uses a single backward scan searching for the next
instructions that reads or writes EFLAGS.
Also:
- Add comments giving examples for the 3 cases handled.
- Check `MI` which contains the result of the zero-compare cases,
instead of re-checking `IsCmpZero`.
- Tweak coding style in some loops.
- Add new MIR based tests that test the optimization in isolation.
This also removes a check for flag readers in situations like this:
```
= SUB32rr %0, %1, implicit-def $eflags
... we no longer stop when there are $eflag users here
CMP32rr %0, %1 ; will be removed
...
```
Differential Revision: https://reviews.llvm.org/D110857
Commit 8fa3e8fa14 added an implicit REP prefix to all VIA PadLock
instructions, but GNU as doesn't add one to xstore, only all the others.
This resulted in a kernel panic regression in FreeBSD upon updating to
LLVM 11 (https://bugs.freebsd.org/259218) which includes the commit in
question. This partially reverts that commit.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D112355
This can merge the acceptable ranges based on the call graph, rather
than the simple application of the attribute. Remove the handling from
the old pass.
Run post-RA SIShrinkInstructions just before post-RA scheduling, instead
of afterwards. After the fixes in D112305 and D112317 this seems to make
no difference, but it paves the way for scheduler tweaks that are
sensitive to the e32 vs e64 encoding of VALU instructions.
Differential Revision: https://reviews.llvm.org/D112341
As described in the comment, the way we change vcc to vcc_lo in these
operands confuses addPhysRegDataDeps into treating them as implicit
pseudo operands. Fix this by setting the correct latency from the
SchedModel after addPhysRegDataDeps wrongly set it to 0.
Differential Revision: https://reviews.llvm.org/D112317
Instead of returning a bool to indicate success and a separate
SDValue, return the SDValue and have the callers check if it is
null.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D112331
This pseudo is expanded very late (AsmPrinter) and therefore has to have a
correct size value, or the branch relaxation pass may make a wrong decision.
Review: Ulrich Weigand
This will allow us to reuse existing interleaved load logic in
lowerInterleavedLoad that exists for neon types, but for SVE fixed
types.
The goal eventually will be to replace the existing ld<n> intriniscs
with these, once a migration path has been sorted out.
Differential Revision: https://reviews.llvm.org/D112078
By definition, interleaving load of stride N means:
load N*VF elements, and shuffle them into N VF-sized vectors,
with 0'th vector containing elements `[0, VF)*stride + 0`,
and 1'th vector containing elements `[0, VF)*stride + 1`.
Example: https://godbolt.org/z/df561Me5E (i64 stride 4 vf 2 => cost 6)
Now, not fully interleaved load, is when not all of these vectors is demanded.
So at worst, we could just pretend that everything is demanded,
and discard the non-demanded vectors. What this means is that the cost
for not-fully-interleaved group should be not greater than the cost
for the same fully-interleaved group, but perhaps somewhat less.
Examples:
https://godbolt.org/z/a78dK5Geq (i64 stride 4 (indices 012u) vf 2 => cost 4)
https://godbolt.org/z/G91ceo8dM (i64 stride 4 (indices 01uu) vf 2 => cost 2)
https://godbolt.org/z/5joYob9rx (i64 stride 4 (indices 0uuu) vf 2 => cost 1)
Right now, for such not-fully-interleaved loads we just use the costs
for fully-interleaved loads. But at least **in general**,
that is obviously overly pessimistic, because **in general**,
not all the shuffles needed to perform the full interleaving
will end up being live.
So what this does, is naively scales the interleaving cost
by the fraction of the live members. I believe this should still result
in the right ballpark cost estimate, although it may be over/under -estimate.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D112307
This doesn't have any effect on codegen now, but it might do in the
future if we shrink instructions before post-RA scheduling, which is
sensitive to live vs dead defs.
Differential Revision: https://reviews.llvm.org/D112305
parseFunctionName allowed a default null pointer, despite it being dereferenced immediately to be used as a reference and that all callers were taking the address of an existing reference.
Fixes static analyzer warning about potential dereferenced nulls
This patch fixes a codegen bug, the test for which was introduced in
D112223.
When merging VSETVLIInfo across blocks, if the 'exit' VSETVLIInfo
produced by a block is found to be compatible with the VSETVLIInfo
computed as the intersection of the 'exit' VSETVLIInfo produced by the
block's predecessors, that blocks' 'exit' info is discarded and the
intersected value is taken in its place.
However, we have one authority on what constitutes VSETVLIInfo
compatibility and we are using it in two different contexts.
Compatibility is used in one context to elide VSETVLIs between
straight-line vector instructions. But compatibility when evaluated
between two blocks' exit infos ignores any info produced *inside* each
respective block before the exit points. As such it does not guarantee
that a block will not produce a VSETVLI which is incompatible with the
'previous' block.
As such, we must ensure that any merging of VSETVLIInfo is performed
using some notion of "strict" compatibility. I've defined this as a full
vtype match, but this is perhaps too pessimistic. Given that test
coverage in this regard is lacking -- the only change is in the failing
test -- I think this is a good starting point.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D112228
In a kernel which does not have calls or AGPR usage we can allocate
the whole vector register budget for VGPRs and have no AGPRs as
long as VGPRs stay addressable (i.e. below 256).
Differential Revision: https://reviews.llvm.org/D111764
G_ICMP is selected to an arithmetic overflow op (ADDS/SUBS/etc) with a dead
destination + a CSINC instruction.
We have a fold which allows us to combine 32-bit adds with G_ICMP.
The problem with G_ICMP is that we model it as always having a 32-bit
destination even though it can be a 64-bit operation. So, we were missing some
opportunities for 64-bit folds.
This patch teaches the fold to recognize 64-bit G_ICMPs + refactors some of
the code surrounding CSINC accordingly.
(Later down the line, I think we should probably change the way we handle G_ICMP
in general.)
Differential Revision: https://reviews.llvm.org/D111088
If a typedef type has __attribute__((btf_decl_tag("str"))) with
bpf target, emit BTF_KIND_DECL_TAG for that type in the BTF.
Differential Revision: https://reviews.llvm.org/D112259
- This patch provides the initial implementation for lowering a call on z/OS according to the XPLINK64 calling convention
- A series of changes have been made to SystemZCallingConv.td to account for these additional XPLINK64 changes including adding a new helper function to shadow the stack along with allocation of a register wherever appropriate
- For the cases of copying a f64 to a gr64 and a f128 / 128-bit vector type to a gr64, a `CCBitConvertToType` has been added and has been bitcasted appropriately in the lowering phase
- Support for the ADA register (R5) will be provided in a later patch.
Reviewed By: uweigand
Differential Revision: https://reviews.llvm.org/D111662
This patch introduces a new function:
AArch64Subtarget::getVScaleForTuning
that returns a value for vscale that can be used for tuning the cost
model when using scalable vectors. The VScaleForTuning option in
AArch64Subtarget is initialised according to the following rules:
1. If the user has specified the CPU to tune for we use that, else
2. If the target CPU was specified we use that, else
3. The tuning is set to "generic".
For CPUs of type "generic" I have assumed that vscale=2.
New tests added here:
Analysis/CostModel/AArch64/sve-gather.ll
Analysis/CostModel/AArch64/sve-scatter.ll
Transforms/LoopVectorize/AArch64/sve-strict-fadd-cost.ll
Differential Revision: https://reviews.llvm.org/D110259
This is a small extension of D112095 to avoid another regression
seen with D112085.
In this case, we allow the same conversion from usubsat to ALU
ops if the target supports vpternlog.
That pattern will get converted later in X86DAGToDAGISel::tryVPTERNLOG().
This seems better than putting a magic immediate constant directly in
this code to create the exact vpternlog that we need. It's possible that
there are other special-cases along these lines, so we should try to
keep all of the vpternlog magic in one place.
Differential Revision: https://reviews.llvm.org/D112138
D106408 was doing this for all targets although it was
reverted due to couple performance regressions on some targets.
The difference for AMDGPU is the ability to rematerialize SOP
instructions with virtual register uses like we already do for VOP.
Differential Revision: https://reviews.llvm.org/D110743
Add relaxed. f32x4.min, f32x4.max, f64x2.min, f64x2.max. These are only
exposed as builtins, and require user opt-in.
Differential Revision: https://reviews.llvm.org/D112146
Our fallback expansion for CTLZ/CTTZ relies on CTPOP. If CTPOP
isn't legal or custom for a vector type we would scalarize the
CTLZ/CTTZ. This is different than CTPOP itself which would use a
vector expansion.
This patch teaches expandCTLZ/CTTZ to rely on the vector CTPOP
expansion instead of scalarizing. To do this I had to add additional
checks to make sure the operations used by CTPOP expansions are all
supported. Some of the operations were already needed for the CTLZ/CTTZ
expansion.
This is a huge improvement to the RISCV which doesn't have a scalar
ctlz or cttz in the base ISA.
For WebAssembly, I've added Custom lowering to keep the scalarizing
behavior. I've also extended the scalarizing to CTPOP.
Differential Revision: https://reviews.llvm.org/D111919
Replace X86ProcFamilyEnum::IntelSLM enum with a TuningUseSLMArithCosts flag instead, matching what we already do for Goldmont.
This just leaves X86ProcFamilyEnum::IntelAtom to replace with general Tuning/Feature flags and we can finally get rid of the old X86ProcFamilyEnum enum.
Differential Revision: https://reviews.llvm.org/D112079
autiasp, autibsp instructions are the counterpart of paciasp/pacibsp instructions
therefore let's emit .cfi_negate_ra_state for these too.
In case of Armv8.3 instruction set the retaa/retbb will do the return and authentication
in one step here we can't emit the . cfi_negate_ra_state because that would be point after
the ret* instruction.
Reviewed By: nickdesaulniers, MaskRay
Differential Revision: https://reviews.llvm.org/D111780
This change implements new DAG nodes TABLE_GET/TABLE_SET, and lowering
methods for load and stores of reference types from IR arrays. These
global LLVM IR arrays represent tables at the Wasm level.
Differential Revision: https://reviews.llvm.org/D111154
Complete the basic integer instruction set and add related predictor in CSKY.td.
And it includes the instruction definition and asm parser support.
Differential Revision: https://reviews.llvm.org/D111701
Add i8x16 relaxed_swizzle instructions. These are only
exposed as builtins, and require user opt-in.
Differential Revision: https://reviews.llvm.org/D112022
Currently, .BTF and .BTF.ext has default alignment of 1.
For example,
$ cat t.c
int foo() { return 0; }
$ clang -target bpf -O2 -c -g t.c
$ llvm-readelf -S t.o
...
Section Headers:
[Nr] Name Type Address Off Size ES Flg Lk Inf Al
...
[ 7] .BTF PROGBITS 0000000000000000 000167 00008b 00 0 0 1
[ 8] .BTF.ext PROGBITS 0000000000000000 0001f2 000050 00 0 0 1
But to have no misaligned data access, .BTF and .BTF.ext
actually requires alignment of 4. Misalignment is not an issue
for architecture like x64/arm64 as it can handle it well. But
some architectures like mips may incur a trap if .BTF/.BTF.ext
is not properly aligned.
This patch explicitly forced .BTF and .BTF.ext alignment to be 4.
For the above example, we will have
[ 7] .BTF PROGBITS 0000000000000000 000168 00008b 00 0 0 4
[ 8] .BTF.ext PROGBITS 0000000000000000 0001f4 000050 00 0 0 4
Differential Revision: https://reviews.llvm.org/D112106
usubsat X, SMIN --> (X ^ SMIN) & (X s>> BW-1)
This would be a regression with D112085 where we combine to
usubsat more aggressively, so avoid that by matching the
special-case where we are subtracting SMIN (signmask):
https://alive2.llvm.org/ce/z/4_3gBD
Differential Revision: https://reviews.llvm.org/D112095
Following on from an earlier patch that introduced support for -mtune
for AArch64 backends, this patch splits out the tuning features
from the processor features. This gives us the ability to enable
architectural feature set A for a given processor with "-mcpu=A"
and define the set of tuning features B with "-mtune=B".
It's quite difficult to write a test that proves we select the
right features according to the tuning attribute because most
of these relate to scheduling. I have created a test here:
CodeGen/AArch64/misched-fusion-addr-tune.ll
that demonstrates the different scheduling choices based upon
the tuning.
Differential Revision: https://reviews.llvm.org/D111551
This patch ensures that we always tune for a given CPU on AArch64
targets when the user specifies the "-mtune=xyz" flag. In the
AArch64Subtarget if the tune flag is unset we use the CPU value
instead.
I've updated the release notes here:
llvm/docs/ReleaseNotes.rst
and added tests here:
clang/test/Driver/aarch64-mtune.c
Differential Revision: https://reviews.llvm.org/D110258
Inspired by D111968, provide a isNegatedPowerOf2() wrapper instead of obfuscating code with (-Value).isPowerOf2() patterns, which I'm sure are likely avenues for typos.....
Differential Revision: https://reviews.llvm.org/D111998
GPR uses argument registers as the first group of registers to allocate.
This patch uses vector argument registers, v8 to v23, as the first group
to allocate.
Differential Revision: https://reviews.llvm.org/D111304
By default clang emits complete contructors as alias of base constructors if they are the same.
The backend is supposed to emit symbols for the alias, otherwise it causes undefined symbols.
@yaxunl observed that this issue is related to the llvm options `-amdgpu-early-inline-all=true`
and `-amdgpu-function-calls=false`. This issue is resolved by only inlining global values
with internal linkage. The `getCalleeFunction()` in AMDGPUResourceUsageAnalysis also had
to be extended to support aliases to functions. inline-calls.ll was corrected appropriately.
Reviewed By: yaxunl, #amdgpu
Differential Revision: https://reviews.llvm.org/D109707
If we're using an ashr to sign-extend the entire upper 16 bits of the i32 element, then we can replace with a lshr. The sign bit will be correctly shifted for PMADDWD's implicit sign-extension and the upper 16 bits are zero so the upper i16 sext-multiply is guaranteed to be zero.
The lshr also has a better chance of folding with shuffles etc.
The feature tells the backend to allow tags in the upper bits of global
variable addresses. These tags will be ignored by upcoming CPUs with
the Intel LAM feature but may be used in instrumentation passes (e.g.,
HWASan).
This patch implements the feature by using @GOTPCREL relocations instead
of direct references to the locally defined global. Thus the full
tagged address can be loaded by a single instruction:
movq global@GOTPCREL(%rip), %rax
Reviewed By: eugenis
Differential Revision: https://reviews.llvm.org/D111343
When compiling for the RWPI relocation model the debug information is wrong:
* the debug location is described as { DW_OP_addr Var }
instead of { DW_OP_constNu Var DW_OP_bregX 0 DW_OP_plus }
* the relocation type is R_ARM_ABS32 instead of R_ARM_SBREL32
Differential Revision: https://reviews.llvm.org/D111404
There are no functionality change.
Fix some comments and rename processAnnotations() to
processDeclAnnotations() to avoid confusion when later
BTF_KIND_TYPE_TAG is introduced (https://reviews.llvm.org/D111199).
Commit 009f3a89d8 ("BPF: remove intrindics @llvm.stacksave()
and @llvm.stackrestore()") implemented IRPeephole pass to remove
llvm.stacksave()/stackrestore() instrinsics.
Buildbot reported a failure:
UNREACHABLE executed at ../lib/IR/LegacyPassManager.cpp:1445!
which is:
llvm_unreachable("Pass modifies its input and doesn't report it");
The code has changed but the implementation didn't return true
for changing. This patch fixed this problem.
This function was copied from ARM where register pairs/triples/quads can wrap around the 32 encoding space. So register 31 can pair with register 0. This is not true for RISCV vectors. The spec specifically mentions the possibility of a future encoding that has more than 32 registers.
This patch removes the modulo from the code and directly checks that destination register is in the source register range and not the beginning of the range. Though I don't expect an identity copy will occur.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D111467
Paul Chaignon reported a bpf verifier failure ([1]) due to using
non-ABI register R11. For the test case, llvm11 is okay while
llvm12 and later generates verifier unfriendly code.
The failure is related to variable length array size.
The following mimics the variable length array definition
in the test case:
struct t { char a[20]; };
void foo(void *);
int test() {
const int a = 8;
char tmp[AA + sizeof(struct t) + a];
foo(tmp);
...
}
Paul helped bisect that the following llvm commit is
responsible:
552c6c2328 ("PR44406: Follow behavior of array bound constant
folding in more recent versions of GCC.")
Basically, before the above commit, clang frontend did constant
folding for array size "AA + sizeof(struct t) + a" to be 68,
so used alloca for stack allocation. After the above commit,
clang frontend didn't do constant folding for array size
any more, which results in a VLA and llvm.stacksave/llvm.stackrestore
is generated.
BPF architecture API does not support stack pointer (sp) register.
The LLVM internally used R11 to indicate sp register but it should
not be in the final code. Otherwise, kernel verifier will reject it.
The early patch ([2]) tried to fix the issue in clang frontend.
But the upstream discussion considered frontend fix is really a
hack and the backend should properly undo llvm.stacksave/llvm.stackrestore.
This patch implemented a bpf IR phase to remove these intrinsics
unconditionally. If eventually the alloca can be resolved with
constant size, r11 will not be generated. If alloca cannot be
resolved with constant size, SelectionDag will complain, the same
as without this patch.
[1] https://lore.kernel.org/bpf/20210809151202.GB1012999@Mem/
[2] https://reviews.llvm.org/D107882
Differential Revision: https://reviews.llvm.org/D111897
The MIPS ABI requires the thread pointer be accessed via rdhwr $3, $r29.
This is currently represented by (CopyToReg $3, (RDHWR $29)) followed by
a (CopyFromReg $3). However, there is no glue between these, meaning
scheduling can break those apart. In particular, PR51691 is a report
where PseudoSELECT_I was moved to between the CopyToReg and CopyFromReg,
and since its expansion uses branches, it split the def and use of the
physical register between two basic blocks, resulting in the def being
eliminated and the use having no def. It also seems possible that a
similar situation could arise splitting up the CopyToReg from the RDHWR,
causing the RDHWR to use a destination register other than $3, violating
the ABI requirement.
Thus, add glue between all three nodes to ensure they aren't split up
during instruction selection. No regression test is added since any test
would be implictly relying on specific scheduling behaviour, so whilst
it might be testing that glue is preventing reordering today, changes to
scheduling behaviour could result in the test no longer being able to
catch a regression here, as the reordering might no longer happen for
other unrelated reasons.
Fixes PR51691.
Reviewed By: atanasyan, dim
Differential Revision: https://reviews.llvm.org/D111967
Try to widen element type to get a new mask value for a better permutation
sequence, so that we can use NEON shuffle instructions, such as zip1/2,
UZP1/2, TRN1/2, REV, INS, etc.
For example:
shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 6, i32 7, i32 2, i32 3>
is equivalent to:
shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 3, i32 1>
Finally, we can get:
mov v0.d[0], v1.d[1]
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D111619
TargetPassConfig::addPass takes a "bool verifyAfter" argument which lets
you skip machine verification after a particular pass. Unfortunately
this is used in generic code in TargetPassConfig itself to skip
verification after a generic pass, only because some previous target-
specific pass damaged the MIR on that specific target. This is bad
because problems in one target cause lack of verification for all
targets.
This patch replaces that mechanism with a new MachineFunction property
called "FailsVerification" which can be set by (usually target-specific)
passes that are known to introduce problems. Later passes can reset it
again if they are known to clean up the previous problems.
Differential Revision: https://reviews.llvm.org/D111397
Add patterns for i8/i16 local atomic load/store.
Added tests for new patterns.
Copied atomic_[store/load]_local.ll to GlobalISel directory.
Differential Revision: https://reviews.llvm.org/D111869
This patch is to order the AVX instructions ahead of AVX512 instructions
in the matching table so that the AVX instructions can be matched first.
Thanks Craig and Shengchen for the idea.
Differential Revision: https://reviews.llvm.org/D111538
The change adds divergence predicates for fused logical operations.
The problem with selecting a scalar fused op such as S_NOR_B32 is
that it does not have a VALU counterpart and will be split in
moveToVALU. At the same time it prevents selection of a better
opcode on the VALU side (such as V_OR3_B32) which does not have a
counterpart on SALU side.
XNOR opcodes are left as is and selected as scalar to get advantage
of the SIInstrInfo::lowerScalarXnor() code which can commute
operations to keep one of two opcodes on SALU if possible. See
xnor.ll test for this.
Differential Revision: https://reviews.llvm.org/D111907
Create new virtual register for the definition of new AND instruction and
replace old register by the new one to keep SSA form.
Differential Revision: https://reviews.llvm.org/D109963
Both ports are required for BitTest ops. Update the uops counts + port usage based off the most recent llvm-exegesis captures and what Intel AoM / Agner reports as well.
A few more tuples are being queried after D111546. Might be good to model them,
They all require a lot of manual assembly surgery.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/9bnKrefcG - for intels `Block RThroughput: =40.0`; for ryzens, `Block RThroughput: =16.0`
So could pick cost of `40`
For store we have:
https://godbolt.org/z/5s3s14dEY - for intels `Block RThroughput: =40.0`; for ryzens, `Block RThroughput: =16.0`
So we could pick cost of `40`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111945
A few more tuples are being queried after D111546. Might be good to model them,
They all require a lot of manual assembly surgery.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/MTaKboejM - for intels `Block RThroughput: =32.0`; for ryzens, `Block RThroughput: <=16.0`
So could pick cost of `32`
For store we have:
https://godbolt.org/z/v7xPj3Wd4 - for intels `Block RThroughput: =32.0`; for ryzens, `Block RThroughput: <=32.0`
So we could pick cost of `32`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111944
A few more tuples are being queried after D111546. Might be good to model them,
They all require a lot of manual assembly surgery.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/11rcvdreP - for intels `Block RThroughput: <=68.0`; for ryzens, `Block RThroughput: <=48.0`
So could pick cost of `68`
For store we have:
https://godbolt.org/z/6aM11fWcP - for intels `Block RThroughput: <=64.0`; for ryzens, `Block RThroughput: <=32.0`
So we could pick cost of `64`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111943
A few more tuples are being queried after D111546. Might be good to model them,
They all require a lot of manual assembly surgery.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/s5b6E6jsP - for intels `Block RThroughput: <=32.0`; for ryzens, `Block RThroughput: <=24.0`
So could pick cost of `32`
For store we have:
https://godbolt.org/z/efh99d93b - for intels `Block RThroughput: <=48.0`; for ryzens, `Block RThroughput: <=32.0`
So we could pick cost of `48`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111942
A few more tuples are being queried after D111546. Might be good to model them,
They all require a lot of manual assembly surgery.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/YTeT9M7fW - for intels `Block RThroughput: <=212.0`; for ryzens, `Block RThroughput: <=64.0`
So could pick cost of `212`
For store we have:
https://godbolt.org/z/vc954KEGP - for intels `Block RThroughput: <=90.0`; for ryzens, `Block RThroughput: <=24.0`
So we could pick cost of `90`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111940
SVE has predicated literal forms of some instructions for specific
literals, which currently are generated correctly when using ACLE
but not when those instructions are generated directly.
This adds the patterns to generate those instructions when
generating from standard LLVM IR instructions.
Differential Revision: https://reviews.llvm.org/D99074
How many place you need to modify when implementing a new extension for RISC-V?
At least 7 places as I know:
- Add new SubtargetFeature at RISCV.td
- -march parser in RISCV.cpp
- RISCVTargetInfo::initFeatureMap@RISCV.cpp for handling feature vector.
- RISCVTargetInfo::getTargetDefines@RISCV.cpp for pre-define marco.
- Arch string parser for ELF attribute in RISCVAsmParser.cpp
- ELF attribute emittion in RISCVAsmParser.cpp, and make sure it's in
canonical order...
- ELF attribute emittion in RISCVTargetStreamer.cpp, and again, must in
canonical order...
And now, this patch provide an unified infrastructure for handling (almost)
everything of RISC-V arch string.
After this patch, you only need to update 2 places for implement an extension
for RISC-V:
- Add new SubtargetFeature at RISCV.td, hmmm, it's hard to avoid.
- Add new entry to RISCVSupportedExtension@RISCVISAInfo.cpp or
SupportedExperimentalExtensions@RISCVISAInfo.cpp .
Most codes are come from existing -march parser, but with few new feature/bug
fixes:
- Accept version for -march, e.g. -march=rv32i2p0.
- Reject version info with `p` but without minor version number like `rv32i2p`.
Differential Revision: https://reviews.llvm.org/D105168
This patch detects the absolute value pattern on the RHS of a
subtract. If we find it we swap the CMOV true/false values and
replace the subtract with an ADD.
There may be a more generic way to do this, but I'm not sure.
Targets that don't have legal or custom ISD::ABS use a generic
expand in DAG combiner already when it sees (neg (abs(x))). I
haven't checked what happens if the neg is a more general subtract.
Fixes PR50991 for X86.
Reviewed By: RKSimon, spatel
Differential Revision: https://reviews.llvm.org/D111858
These cases use the same codegen as AVX2 (pshuflw/pshufd) for the sub-128bit vector deinterleaving, and unpcklqdq for v2i64.
It's going to take a while to add full interleaved cost coverage, but since these are the same for SSE2 -> AVX2 it should be an easy win.
Fixes PR47437
Differential Revision: https://reviews.llvm.org/D111938
And another attempt to start untangling this ball of threads around gather.
There's `TTI::prefersVectorizedAddressing()`hoop, which confusingly defaults to `true`,
which tells LV to try to vectorize the addresses that lead to loads,
but X86 generally can not deal with vectors of addresses,
the only instructions that support that are GATHER/SCATTER,
but even those aren't available until AVX2, and aren't really usable until AVX512.
This specializes the hook for X86, to return true only if we have AVX512 or AVX2 w/ fast gather.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111546
By default clang emits complete contructors as alias of base constructors if they are the same.
The backend is supposed to emit symbols for the alias, otherwise it causes undefined symbols.
@yaxunl observed that this issue is related to the llvm options `-amdgpu-early-inline-all=true`
and `-amdgpu-function-calls=false`. This issue is resolved by only inlining global values
with internal linkage. The `getCalleeFunction()` in AMDGPUResourceUsageAnalysis also had
to be extended to support aliases to functions. inline-calls.ll was corrected appropriately.
Reviewed By: yaxunl, #amdgpu
Differential Revision: https://reviews.llvm.org/D109707
- When a redundant MBB is being erased from MDT, check whether its
single successor is dominiated by it. If yes, update that successor's
idom before erasing MBB; otherwise, it implies MBB is a leaf node and
could be erased directly.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D111831
The isU24() and isI24() calls numBits to make its decision. This change
replaces them with the internal numBits call so that we can use its
result for the > 32 bit width cases.
Differential Revision: https://reviews.llvm.org/D111864
This changes fixes a case in which the highest set bit of the original
result is at bit 31 and sign-extending the mul24 for it would make the
result negative.
Differential Revision: https://reviews.llvm.org/D111823
Similar to D111236, this improves the lowering of vector fptosi.sat and
fptoui.sat, using legal converts and further saturating from there with
min/max. f64 are excluded for the moment due to producing worse code in
places compared to the unrolling.
Differential Revision: https://reviews.llvm.org/D111787
Improve the lowering of scalar fptosi.sat and fptoui.sat for saturating
widths smaller than legal types by using the fact that the legal type
will saturate under aarch64, and saturating the result further using
min/max.
Differential Revision: https://reviews.llvm.org/D111236
The MOVCC peephole eliminates a MOVCC by making one of its inputs a
conditional instruction, but when doing this it should be using both
inputs of the MOVCC to decide on the register class to use as
otherwise we can get an error when using -verify-machineinstrs.
Differential Revision: https://reviews.llvm.org/D111714
Use SH1ADD/SH2ADD/SH3ADD along with LUI+ADDI to compose int32*3,
int32*5 and int32*9.
Reviewed By: craig.topper, luismarques
Differential Revision: https://reviews.llvm.org/D111484
While i've modelled most of the relevant tuples for AVX2,
that only covered fully-interleaved groups.
By definition, interleaving load of stride N means:
load N*VF elements, and shuffle them into N VF-sized vectors,
with 0'th vector containing elements `[0, VF)*stride + 0`,
and 1'th vector containing elements `[0, VF)*stride + 1`.
Example: https://godbolt.org/z/df561Me5E (i64 stride 4 vf 2 => cost 6)
Now, not fully interleaved load, is when not all of these vectors is demanded.
So at worst, we could just pretend that everything is demanded,
and discard the non-demanded vectors. What this means is that the cost
for not-fully-interleaved group should be not greater than the cost
for the same fully-interleaved group, but perhaps somewhat less.
Examples:
https://godbolt.org/z/a78dK5Geq (i64 stride 4 (indices 012u) vf 2 => cost 4)
https://godbolt.org/z/G91ceo8dM (i64 stride 4 (indices 01uu) vf 2 => cost 2)
https://godbolt.org/z/5joYob9rx (i64 stride 4 (indices 0uuu) vf 2 => cost 1)
As we have established over the course of last ~70 patches, (wow)
`BaseT::getInterleavedMemoryOpCos()` is absolutely bogus,
it is usually almost an order of magnitude overestimation,
so i would claim that we should at least use the hardcoded costs
of fully interleaved load groups.
We could go further and adjust them e.g. by the number of demanded indices,
but then i'm somewhat fearful of underestimating the cost.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111174
CMOVGE reads SF and OF. CMOVNS only reads SF. This matches with
other recent changes to use a single flag where possible. It also
matches gcc codegen.
I believe this technically changes whether the conditioanl move happens
on INT_MIN, but for INT_MIN both registers are the same so it doesn't
matter.
Differential Revision: https://reviews.llvm.org/D111826
This a NFC refactor patch to merge the AVX2 interleaved cost handling back into the getInterleavedMemoryOpCost base method - while getInterleavedMemoryOpCostAVX512 uses instruction and patterns very specific to AVX512+, much of the costs analysis for AVX2 can be reused for all SSE targets.
This is the first step towards improving SSE and AVX1 costs that will reuse the relevant AVX2 costs by splitting some of the tables - for instance AVX1 has very similar costs for most vXi64/vXf64 interleave patterns and many sub-128bit vector costs are the same all the way down to SSE2 (or at least SSSE3).
Differential Revision: https://reviews.llvm.org/D111822
I have some upcoming refactoring for SSE/AVX1 interleaving cost support, and the diff is a lot nicer if the (unaltered) AVX512 implementation isn't stuck between getInterleavedMemoryOpCost and getInterleavedMemoryOpCostAVX2
I've removed the Zbs W instructions that are not part of the frozen spec.
References to B as an extension name have been removed. Tests are updated or split accordingly.
Reviewed By: luismarques
Differential Revision: https://reviews.llvm.org/D110669
These registers are used as operands for instructions that expect an
integer register, so they should be added to Int32Regs or Int64Regs
register classes. Otherwise the machine verifier emits an error for
the following LIT tests when LLVM_ENABLE_MACHINE_VERIFIER=1
environment variable is set:
*** Bad machine code: Illegal physical register for instruction ***
- function: kernel_func
- basic block: %bb.0 entry (0x55c8903d5438)
- instruction: %3:int64regs = LEA_ADDRi64 $vrframelocal, 0
- operand 1: $vrframelocal
$vrframelocal is not a Int64Regs register.
CodeGen/NVPTX/call-with-alloca-buffer.ll
CodeGen/NVPTX/disable-opt.ll
CodeGen/NVPTX/lower-alloca.ll
CodeGen/NVPTX/lower-args.ll
CodeGen/NVPTX/param-align.ll
CodeGen/NVPTX/reg-types.ll
DebugInfo/NVPTX/dbg-declare-alloca.ll
DebugInfo/NVPTX/dbg-value-const-byref.ll
Differential Revision: https://reviews.llvm.org/D110164
The patch attempts to optimize a sequence of SIMD loads from the same
base pointer:
%0 = gep float*, float* base, i32 4
%1 = bitcast float* %0 to <4 x float>*
%2 = load <4 x float>, <4 x float>* %1
...
%n1 = gep float*, float* base, i32 N
%n2 = bitcast float* %n1 to <4 x float>*
%n3 = load <4 x float>, <4 x float>* %n2
For AArch64 the compiler generates a sequence of LDR Qt, [Xn, #16].
However, 32-bit NEON VLD1/VST1 lack the [Wn, #imm] addressing mode, so
the address is computed before every ld/st instruction:
add r2, r0, #32
add r0, r0, #16
vld1.32 {d18, d19}, [r2]
vld1.32 {d22, d23}, [r0]
This can be improved by computing address for the first load, and then
using a post-indexed form of VLD1/VST1 to load the rest:
add r0, r0, #16
vld1.32 {d18, d19}, [r0]!
vld1.32 {d22, d23}, [r0]
In order to do that, the patch adds more patterns to DAGCombine:
- (load (add ptr inc1)) and (add ptr inc2) are now folded if inc1
and inc2 are constants.
- (or ptr inc) is now recognized as a pointer increment if ptr is
sufficiently aligned.
In addition to that, we now search for all possible base updates and
then pick the best one.
Differential Revision: https://reviews.llvm.org/D108988
Without SSE41 sext/zext instructions the extensions will be split, meaning that the MUL->PMADDWD fold will split the sext_i32(x) into zext_i32(sext_i16(x))
This patch fixes the bug that consisted of treating variable / immediate
length mem operations (such as memcpy, memset, ...) differently. The variable
length case needs to have the length minus 1 passed due to the use of EXRL
target instructions. However, the DAGCombiner can convert a register length
argument into a constant one, and whenever that happened one byte too little
would end up being performed.
This is also a refactorization by reducing the number of opcodes and variants
involved. For any opcode (variable or constant length), only the length minus
one is passed on to the ISD node. The rest of the logic is now instead
handled during isel pseudo expansion.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D111729
Opitimize immediate materialisation in the following way if profitable:
1. Use BCLRI for upper 32 bits if the lower 32 bits are negative int32.
2. Use BSETI for upper 32 bits if the lower 32 bits are positive int32.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D111508
The 24-bit mul intrinsics yields the low-order 32 bits. We should only
do the transformation if the operands are known to be not wider than 24
bits and the result is known to be not wider than 32 bits.
Differential Revision: https://reviews.llvm.org/D111523
`X86TTIImpl::getGSScalarCost()` has (at least) two issues:
* it naively computes the cost of sequence of `insertelement`/`extractelement`.
If we are operating not on the XMM (but YMM/ZMM),
this widely overestimates the cost of subvector insertions/extractions.
* Gather/scatter takes a vector of pointers, and scalarization results in us performing
scalar memory operation for each of these pointers, but we never account for the cost
of extracting these pointers out of the vector of pointers.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111222
NFC. This check does not verify any functional property since size 8
was added. Remove it for simplicity.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D111737
Change-Id: Ifd7cbd324a137f939d8dc04acb8fbd54c9527a42
This PR implements the save of the XPLINK callee-saved registers
on z/OS.
Reviewed By: uweigand, Kai
Differential Revision: https://reviews.llvm.org/D111653
This has a couple of benefits:
1. It can sometimes fix clusters that got broken apart when the register
allocator inserted a copy.
2. Post-RA scheduling does not have to worry about increasing register
pressure, which in some cases gives it more freedom to reorder
instructions.
Testing on a collection of 10,000 graphics shaders compiled for gfx1010
showed:
- The average length of each run of one or more load instructions
increased by about 1%.
- The number of runs of two or more load instructions increased by
about 4%.
Differential Revision: https://reviews.llvm.org/D111646
This patch adds patterns to match the following with INC/DEC:
- @llvm.aarch64.sve.cnt[b|h|w|d] intrinsics + ADD/SUB
- vscale + ADD/SUB
For some implementations of SVE, INC/DEC VL is not as cheap as ADD/SUB and
so this behaviour is guarded by the "use-scalar-inc-vl" feature flag, which for SVE
is off by default. There are no known issues with SVE2, so this feature is
enabled by default when targeting SVE2.
Reviewed By: david-arm
Differential Revision: https://reviews.llvm.org/D111441
This makes Wasm EH work with dynamic linking. So far we were only able
to handle destructors, which do not use any tags or LSDA info.
1. This uses `TargetExternalSymbol` for `GCC_except_tableN` symbols,
which points to the address of per-function LSDA info. It is more
convenient to use than `MCSymbol` because it can take additional
target flags.
2. When lowering `wasm_lsda` intrinsic, if PIC is enabled, make the
symbol relative to `__memory_base` and generate the `add` node. If
PIC is disabled, continue to use the absolute address.
3. Make tag symbols (`__cpp_exception` and `__c_longjmp`) undefined in
the backend, because it is hard to make it work with dynamic
linking's loading order. Instead, we make all tag symbols undefined
in the LLVM backend and import it from JS.
4. Add support for undefined tags to the linker.
Companion patches:
- https://github.com/WebAssembly/binaryen/pull/4223
- https://github.com/emscripten-core/emscripten/pull/15266
Reviewed By: sbc100
Differential Revision: https://reviews.llvm.org/D111388
Do the following optimization for immediate materialisation:
1. For values in range 0xffffffff 7fffffff ~ 0xffffffff 00000000, first
generate the lower 32-bit with Val|0x80000000 (which is expected be an
int32), then emit (BCLRI r, 31).
2. For values in range 0x80000000 ~ 0xffffffff, first generate the lower
32-bit with Val&~0x80000000 (which is expected to be an int32), then
emit (BSETI r, 31).
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D111532
GCC 9.1 removed Intel MPX support. Linux kernel removed MPX in 2019.
glibc 2.35 will remove MPX.
Our support is limited: we support assembling of bndmov but not bnd.
Just remove it.
Reviewed By: pengfei, skan
Differential Revision: https://reviews.llvm.org/D111517
On AIX, the system assembler does not support the extended mnemonics
dcbtt and dcbtstt. This patch stops them from being emitted on
AIX and emits the base mnemonics instead, dcbt X, X, 16 and
dcbtstt X, X, 16 respectively.
Differential revision: https://reviews.llvm.org/D111258
The new pass walks kernel's pointer arguments, then loads from them.
If a loaded value is a pointer and loaded pointer is unmodified in
the kernel before the load, then promote loaded pointer to global.
Then recursively continue.
Differential Revision: https://reviews.llvm.org/D111464
As noted in https://github.com/halide/Halide/pull/6302,
we hilariously fail to match PAVG if we even as much
as look at it the wrong way.
In this particular case, the problem stems from the fact that
`PAVG` root (def) is a `trunc`, and leafs (uses) are `zext`'s,
and InstCombine really loves to get rid of both of these,
for example replace them with a bit mask. So we may not have
said `zext`.
Instead of checking for that + type match,
i think we should rely on the actual active type,
as per the knownbits.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111571
This has a couple of benefits:
1. It can sometimes fix clusters that got broken apart when the register
allocator inserted a copy.
2. Post-RA scheduling does not have to worry about increasing register
pressure, which in some cases gives it more freedom to reorder
instructions.
Testing on a collection of 10,000 graphics shaders compiled for gfx1010
showed:
- The average length of each run of one or more load instructions
increased by about 1%.
- The number of runs of two or more load instructions increased by
about 4%.
Rename vfredsum and vfwredsum to vfredusum and vfwredusum. Add aliases for vfredsum and vfwredsum.
Reviewed By: luismarques, HsiangKai, khchen, frasercrmck, kito-cheng, craig.topper
Differential Revision: https://reviews.llvm.org/D105690
Per discussion in https://reviews.llvm.org/D111199,
the existing btf_tag attribute will be renamed to
btf_decl_tag. This patch updated BTF backend to
use btf_decl_tag attribute name and also
renamed BTF_KIND_TAG to BTF_KIND_DECL_TAG.
Differential Revision: https://reviews.llvm.org/D111592
All those frame indices which are dead after sgpr spill should be removed from
the function frame. Othewise, there is a side effect such as re-mapping of free
frame index ids by the later pass(es) like "stack slot coloring" which in turn
could mess-up with the book keeping of "frame index to VGPR lane".
Reviewed By: cdevadas
Differential Revision: https://reviews.llvm.org/D111150
This shouldn't be called for non-neon targets at the moment in either
case, but it is good to be expliit about the CombineBaseUpdate being a
NEON function, not expecting to be run under MVE.
These instructions should allow src0 to be a literal with the same
value as the mandatory other literal. Enable it by introducing an
operand that defers adding its value to the MI when decoding till
the mandatory literal is parsed.
Reviewed By: dp, foad
Differential Revision: https://reviews.llvm.org/D111067
Change-Id: I22b0ae0d35bad17b6f976808e48bffe9a6af70b7
armv9-a, armv9.1-a and armv9.2-a can be targeted using the -march option
both in ARM and AArch64.
- Armv9-A maps to Armv8.5-A.
- Armv9.1-A maps to Armv8.6-A.
- Armv9.2-A maps to Armv8.7-A.
- The SVE2 extension is enabled by default on these architectures.
- The cryptographic extensions are disabled by default on these
architectures.
The Armv9-A architecture is described in the Arm® Architecture Reference
Manual Supplement Armv9, for Armv9-A architecture profile
(https://developer.arm.com/documentation/ddi0608/latest).
Reviewed By: SjoerdMeijer
Differential Revision: https://reviews.llvm.org/D109517
The select(pshufb,pshufb) -> or(pshufb,pshufb) fold uses getConstVector to create the refreshed pshufb masks, which treats all negative indices as undef.
PR52122 shows that if we were selecting an element that the PSHUFB has set to zero we must set it back to 0x80 when we recreate the PSHUFB mask and not just leave it as SM_SentinelZero
The lowering for EXTRACT_SUBVECTOR should not be called during type
legalization, only as part of lowering, hence return SDValue() when
called on illegal types.
This also adds missing tests for extracting fixed types from illegal
scalable types.
Differential Revision: https://reviews.llvm.org/D111412
AAPCS requires i1 argument to be zero-extended to 8-bits by the
caller. Emit a new AArch64ISD::ASSERT_ZEXT_BOOL hint (or AssertZExt
for GlobalISel) to enable some optimization opportunities. In
particular, when the argument is forwarded to the callee, we can avoid
zero-extension and use it as-is.
Differential Revision: https://reviews.llvm.org/D107160
This patch adds further support for vectorisation of loops that involve
selecting an integer value based on a previous comparison. Consider the
following C++ loop:
int r = a;
for (int i = 0; i < n; i++) {
if (src[i] > 3) {
r = b;
}
src[i] += 2;
}
We should be able to vectorise this loop because all we are doing is
selecting between two states - 'a' and 'b' - both of which are loop
invariant. This just involves building a vector of values that contain
either 'a' or 'b', where the final reduced value will be 'b' if any lane
contains 'b'.
The IR generated by clang typically looks like this:
%phi = phi i32 [ %a, %entry ], [ %phi.update, %for.body ]
...
%pred = icmp ugt i32 %val, i32 3
%phi.update = select i1 %pred, i32 %b, i32 %phi
We already detect min/max patterns, which also involve a select + cmp.
However, with the min/max patterns we are selecting loaded values (and
hence loop variant) in the loop. In addition we only support certain
cmp predicates. This patch adds a new pattern matching function
(isSelectCmpPattern) and new RecurKind enums - SelectICmp & SelectFCmp.
We only support selecting values that are integer and loop invariant,
however we can support any kind of compare - integer or float.
Tests have been added here:
Transforms/LoopVectorize/AArch64/sve-select-cmp.ll
Transforms/LoopVectorize/select-cmp-predicated.ll
Transforms/LoopVectorize/select-cmp.ll
Differential Revision: https://reviews.llvm.org/D108136
We would like to start pushing -mcpu=generic towards enabling the set of
features that improves performance for some CPUs, without hurting any
others. A blend of the performance options hopefully beneficial to all
CPUs. The largest part of that is enabling in-order scheduling using the
Cortex-A55 schedule model. This is similar to the Arm backend change
from eecb353d0e which made -mcpu=generic perform in-order scheduling
using the cortex-a8 schedule model.
The idea is that in-order cpu's require the most help in instruction
scheduling, whereas out-of-order cpus can for the most part out-of-order
schedule around different codegen. Our benchmarking suggests that
hypothesis holds. When running on an in-order core this improved
performance by 3.8% geomean on a set of DSP workloads, 2% geomean on
some other embedded benchmark and between 1% and 1.8% on a set of
singlecore and multicore workloads, all running on a Cortex-A55 cluster.
On an out-of-order cpu the results are a lot more noisy but show flat
performance or an improvement. On the set of DSP and embedded
benchmarks, run on a Cortex-A78 there was a very noisy 1% speed
improvement. Using the most detailed results I could find, SPEC2006 runs
on a Neoverse N1 show a small increase in instruction count (+0.127%),
but a decrease in cycle counts (-0.155%, on average). The instruction
count is very low noise, the cycle count is more noisy with a 0.15%
decrease not being significant. SPEC2k17 shows a small decrease (-0.2%)
in instruction count leading to a -0.296% decrease in cycle count. These
results are within noise margins but tend to show a small improvement in
general.
When specifying an Apple target, clang will set "-target-cpu apple-a7"
on the command line, so should not be affected by this change when
running from clang. This also doesn't enable more runtime unrolling like
-mcpu=cortex-a55 does, only changing the schedule used.
A lot of existing tests have updated. This is a summary of the important
differences:
- Most changes are the same instructions in a different order.
- Sometimes this leads to very minor inefficiencies, such as requiring
an extra mov to move variables into r0/v0 for the return value of a test
function.
- misched-fusion.ll was no longer fusing the pairs of instructions it
should, as per D110561. I've changed the schedule used in the test
for now.
- neon-mla-mls.ll now uses "mul; sub" as opposed to "neg; mla" due to
the different latencies. This seems fine to me.
- Some SVE tests do not always remove movprfx where they did before due
to different register allocation giving different destructive forms.
- The tests argument-blocks-array-of-struct.ll and arm64-windows-calls.ll
produce two LDR where they previously produced an LDP due to
store-pair-suppress kicking in.
- arm64-ldp.ll and arm64-neon-copy.ll are missing pre/postinc on LPD.
- Some tests such as arm64-neon-mul-div.ll and
ragreedy-local-interval-cost.ll have more, less or just different
spilling.
- In aarch64_generated_funcs.ll.generated.expected one part of the
function is no longer outlined. Interestingly if I switch this to use
any other scheduled even less is outlined.
Some of these are expected to happen, such as differences in outlining
or register spilling. There will be places where these result in worse
codegen, places where they are better, with the SPEC instruction counts
suggesting it is not a decrease overall, on average.
Differential Revision: https://reviews.llvm.org/D110830
This moves the registry higher in the LLVM library dependency stack.
Every client of the target registry needs to link against MC anyway to
actually use the target, so we might as well move this out of Support.
This allows us to ensure that Support doesn't have includes from MC/*.
Differential Revision: https://reviews.llvm.org/D111454
PR45875 notes an instance where exception handling crashes on aarch64-fuchsia
where SCS is enabled by default. The underlying issue seems to be that within libunwind,
various _Unwind_* functions, the x18 register is not updated if a function is marked
with nounwind. This removes the check for nounwind and emits the CFI instruction that updates x18.
Differential Revision: https://reviews.llvm.org/D79822
For PS shaders we can use the input SPI_PS_INPUT_ENA and SPI_PS_INPUT_ADDR
registers
Calculate the number of VGPR registers used as input VGPRs based on these
registers rather than the arguments passed in (this conservatively always
allocates the maximum).
Differential Revision: https://reviews.llvm.org/D101633
Change-Id: Idf7c060cbbd5f7e3300102c55ecee3c07f209de6
At this point it looks like a B extension will never exist. Instead
Zba, Zbb, Zbc, and Zbs are individual extensions being ratified
together as a package. Unknown at this time when or if the other
Zb* extensions will be ratified.
This patch removes references to the B extension. I've updated and
split tests accordingly.
This has been split from D110669 to make review a little easier.
Differential Revision: https://reviews.llvm.org/D111338
Updating the MachineDominatorTree is easy since SILowerControlFlow only
splits and removes basic blocks. This should save a bit of compile time
because previously we would recompute the dominator tree from scratch
after this pass.
Another reason for doing this is that SILowerControlFlow preserves
LiveIntervals which transitively requires MachineDominatorTree. I think
that means that SILowerControlFlow is obliged to preserve
MachineDominatorTree too as explained here:
https://lists.llvm.org/pipermail/llvm-dev/2020-November/146923.html
although it does not seem to have caused any problems in practice yet.
Differential Revision: https://reviews.llvm.org/D111313
X86InstrInfo::convertToThreeAddress would convert this:
%1:gr32 = ADD32rr killed %0:gr32(tied-def 0), %0:gr32, implicit-def dead $eflags
to this:
undef %2.sub_32bit:gr64 = COPY killed %0:gr32
undef %3.sub_32bit:gr64_nosp = COPY %0:gr32
%1:gr32 = LEA64_32r killed %2:gr64, 1, killed %3:gr64_nosp, 0, $noreg
Note that in the ADD32rr, %0 was used twice and the first use had a kill
flag, which is what MachineInstr::addRegisterKilled does.
In the converted code, each use of %0 is copied to a new reg, and the
first COPY inherits the kill flag from the ADD32rr. This causes
machine verification to fail (if you force it to run after
TwoAddressInstructionPass) because the second COPY uses %0 after it is
killed. Note that machine verification is currently disabled after
TwoAddressInstructionPass but this is a step towards being able to
enable it.
Fix this by not inserting more than one COPY from the same source
register.
Differential Revision: https://reviews.llvm.org/D110829
Based on the reasoning of D53903, register operands of DBG_VALUE are
invariably treated as RegState::Debug operands. This change enforces
this invariant as part of MachineInstr::addOperand so that all passes
emit this flag consistently.
RegState::Debug is inconsistently set on DBG_VALUE registers throughout
LLVM. This runs the risk of a filtering iterator like
MachineRegisterInfo::reg_nodbg_iterator to process these operands
erroneously when not parsed from MIR sources.
This issue was observed in the development of the llvm-mos fork which
adds a backend that relies on physical register operands much more than
existing targets. Physical RegUnit 0 has the same numeric encoding as
$noreg (indicating an undef for DBG_VALUE). Allowing debug operands into
the machine scheduler correlates $noreg with RegUnit 0 (i.e. a collision
of register numbers with different zero semantics). Eventually, this
causes an assert where DBG_VALUE instructions are prohibited from
participating in live register ranges.
Reviewed By: MatzeB, StephenTozer
Differential Revision: https://reviews.llvm.org/D110105
Currently when creating tail predicated loops, we need to validate that
all the live-outs of a loop will be equivalent with and without tail
predication, and if they are not we cannot legally create a
tail-predicated loop, leaving expensive vctp and vpst instructions in
the loop. These notably can include register-allocation instructions
like stack loads and stores, and copys lowered from COPYs to MVE_VORRs.
Instead of trying to prove this is valid late in the pipeline, this
patch introduces a MQPRCopy pseudo instruction that COPY is lowered to.
This can then either be converted to a MVE_VORR where possible, or to a
couple of VMOVD instructions if not. This way they do not behave
differently within and outside of tail-predications regions, and we can
know by construction that they are always valid. The idea is that we can
do the same with stack load and stores, converting them to VLDR/VSTR or
VLDM/VSTM where required to prove tail predication is always valid.
This does unfortunately mean inserting multiple VMOVD instructions,
instead of a single MVE_VORR, but my experiments show it to be an
improvement in general.
Differential Revision: https://reviews.llvm.org/D111048
Changes in architecture revision 00eac1:
* Renamed to PSEL.
* Copies whole source register.
* Element type suffix removed from destination.
* Element index no longer optional and '#' prefix has been removed.
The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2021-09
Depends on D111212.
Reviewed By: kmclaughlin
Differential Revision: https://reviews.llvm.org/D111213
Changes in architecture revision 00eac1:
* Tile slice index offset no longer prefixed with '#'.
* The syntax for 128-bit (.Q) ZA tile slice accesses must now include
an explicit zero index.
The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2021-09
Reviewed By: david-arm
Differential Revision: https://reviews.llvm.org/D111212
To better reflect the meaning of the now-disambiguated {GlobalValue,
GlobalAlias}::getBaseObject after breaking off GlobalIFunc::getResolverFunction
(D109792), the function is renamed to getAliaseeObject.
Without popcnt we had a special case for using the parity flag from a single test i8 test instruction if only bits 7:0 could be non-zero. That special case is still useful when we have popcnt.
To reach this special case, we enable custom lowering of parity for i16/i32/i64 even when popcnt is enabled. The check for POPCNT being enabled is now after the special case in LowerPARITY.
Fixes PR52093
Differential Revision: https://reviews.llvm.org/D111249
Lowering of byval parameters with sizes that are not represented by a single
store require multiple stores to properly address the correct size of the
parameter.
Sizes that cannot be done with a single store are 3 bytes, 5 bytes, 6 bytes,
7 bytes. It is not correct to simply perform an 8 byte store and for these
elements because then the store would be larger than the element and alias
analysis would assume that this is undefined behaivour and return NoAlias
for them.
This patch adds the correct stores so that the size of the store is not larger
than the size of the element.
Reviewed By: nemanjai, #powerpc
Differential Revision: https://reviews.llvm.org/D108795
D100244 missed a check on the ResNo of the extract's operand 0 when finding a
pair of extracts to combine into a VMOVRRD (extract(x, n); extract(x, n+1) ->
VMOVRRD(extract x, n/2)). As a result, it can incorrectly pair an extract(x, n)
with another extract(x:3, n+1) for example. This patch fixes the bug by adding
the proper check on ResNo.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D111188
As described on D111049, we're trying to remove the <string> dependency from error handling and replace uses of report_fatal_error(const std::string&) with the Twine() variant which can be forward declared.
The X86 backend only needs to know whether structure return is via an
sret pointer. This removes the categorization enumeration and
adjusts, templatizes and renames the related functions.
Differential Revision: https://reviews.llvm.org/D109966
As suggested on D111024, we should treat getCmpSelInstrCost calls without a specific predicate as matching the worst case predicate cost.
These regressions will be addressed with a mixture of D111024 and fixing other specific getCmpSelInstrCost calls to have realistic predicates.
Seem to cause test failures in compiler-rt.
Revert "[SystemZ] Implement memcmp of variable length with CLC."
This reverts commit 7a4e9a0c73.
Revert "[SystemZ] Implement memcpy of variable length with MVC."
This reverts commit c6c13c58ee.
The currently implementation of funcrefs is broken since it is putting
the funcref itself on the stack before the call_indirect. Instead what
should be on the stack is the constant 0, which is the index at which
we store the funcref in __funcref_call_table.
Reviewed By: tlively
Differential Revision: https://reviews.llvm.org/D111152
This is a non-functional change to remove the duplicate
WasmAddressSpace enum and refactor reftype predicates by moving them
to the Utilities source file.
Reviewed By: tlively
Differential Revision: https://reviews.llvm.org/D111144
Without this change _term instructions can be removed during
critical edge splitting.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D111126
This removes `WasmTagType`. `WasmTagType` contained an attribute and a
signature index:
```
struct WasmTagType {
uint8_t Attribute;
uint32_t SigIndex;
};
```
Currently the attribute field is not used and reserved for future use,
and always 0. And that this class contains `SigIndex` as its property is
a little weird in the place, because the tag type's signature index is
not an inherent property of a tag but rather a reference to another
section that changes after linking. This makes tag handling in the
linker also weird that tag-related methods are taking both `WasmTagType`
and `WasmSignature` even though `WasmTagType` contains a signature
index. This is because the signature index changes in linking so it
doesn't have any info at this point. This instead moves `SigIndex` to
`struct WasmTag` itself, as we did for `struct WasmFunction` in D111104.
In this CL, in lib/MC and lib/Object, this now treats tag types in the
same way as function types. Also in YAML, this removes `struct Tag`,
because now it only contains the tag index. Also tags set `SigIndex` in
`WasmImport` union, as functions do.
I think this makes things simpler and makes tag handling more in line
with function handling. These two shares similar properties in that both
of them have signatures, but they are kind of nominal so having the same
signature doesn't mean they are the same element.
Also a drive-by fix: the reserved 'attirubute' part's encoding changed
from uleb32 to uint8 a while ago. This was fixed in lib/MC and
lib/Object but not in YAML. This doesn't change object files because the
field's value is always 0 and its encoding is the same for the both
encoding.
This is effectively NFC; I didn't mark it as such just because it
changed YAML test results.
Reviewed By: sbc100, tlively
Differential Revision: https://reviews.llvm.org/D111086
As described on D111049, we're trying to remove the <string> dependency from error handling and replace uses of report_fatal_error(const std::string&) with the Twine() variant which can be forward declared.
We can use the raw_string_ostream::str() method to perform the implicit flush() and return a reference to the std::string container that we can then wrap inside Twine().
Following the same pattern of memset/memcpy, this patch implements a variable
length memcmp with a CLC loop followed by an EXRL instruction.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D107380
Instead of making a memcpy libcall, emit an MVC loop and an EXRL instruction
the same way as is already done for memset 0.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D106874
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/1jfGddcre - for intels `Block RThroughput: =36.0`; for ryzens, `Block RThroughput: =12.0`
So could pick cost of `36`
For store we have:
https://godbolt.org/z/ao9srMT8r - for intels `Block RThroughput: =30.0`; for ryzens, `Block RThroughput: =12.0`
So we could pick cost of `30`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111094
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/rc8jYxW6M - for intels `Block RThroughput: =18.0`; for ryzens, `Block RThroughput: =6.0`
So could pick cost of `18`.
For store we have:
https://godbolt.org/z/9PhPEr65G - for intels `Block RThroughput: =15.0`; for ryzens, `Block RThroughput: =6.0`
So we could pick cost of `15`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111093
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/onese7rec - for intels `Block RThroughput: =6.0`; for ryzens, `Block RThroughput: =3.0`
So could pick cost of `6`.
For store we have:
https://godbolt.org/z/bMd7dddnT - for intels `Block RThroughput: =8.0`; for ryzens, `Block RThroughput: <=6.0`
So we could pick cost of `8`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111092
This one required quite a bit of an assembly surgery, but i think it's in the right ballpark..
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/na97Kb96o - for intels `Block RThroughput: <=64.0`; for ryzens, `Block RThroughput: <=32.0`
So could pick cost of `64`.
For store we have:
https://godbolt.org/z/GG1WeoKar - for intels `Block RThroughput: =66.0`; for ryzens, `Block RThroughput: <=27.5`
So we could pick cost of `66`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111091
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/jK85GWKaK - for intels `Block RThroughput: =31.0`; for ryzens, `Block RThroughput: <=17.0`
So could pick cost of `31`.
For store we have:
https://godbolt.org/z/hPWWhEEf9 - for intels `Block RThroughput: =33.0`; for ryzens, `Block RThroughput: <=13.8`
So we could pick cost of `33`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111089
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/szEj1ceee - for intels `Block RThroughput: =15.0`; for ryzens, `Block RThroughput: <=8.8`
So could pick cost of `15`.
For store we have:
https://godbolt.org/z/81bq4fTo1 - for intels `Block RThroughput: =12.0`; for ryzens, `Block RThroughput: <=10.0`
So we could pick cost of `12`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111087
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/aec96Thee - for intels `Block RThroughput: =6.0`; for ryzens, `Block RThroughput: <=3.3`
So could pick cost of `6`.
For store we have:
https://godbolt.org/z/aec96Thee - for intels `Block RThroughput: =9.0`; for ryzens, `Block RThroughput: <=3.0`
So we could pick cost of `9`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111083
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/3M3hbq7n8 - for intels `Block RThroughput: =20.0`; for ryzens, `Block RThroughput: =8.0`
So could pick cost of `20`.
For store we have:
https://godbolt.org/z/zvnPYWTx7 - for intels `Block RThroughput: =20.0`; for ryzens, `Block RThroughput: =8.0`
So we could pick cost of `20`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111076
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/MTKdzjvnr - for intels `Block RThroughput: =8.0`; for ryzens, `Block RThroughput: <=4.0`
So could pick cost of `8`.
For store we have:
https://godbolt.org/z/cMYEvqoah - for intels `Block RThroughput: =8.0`; for ryzens, `Block RThroughput: <=4.0`
So we could pick cost of `8`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111075
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/z197317d1 - for intels `Block RThroughput: =6.0`; for ryzens, `Block RThroughput: =2.0`
So could pick cost of `6`.
For store we have:
https://godbolt.org/z/8dzszjf9q - for intels `Block RThroughput: =6.0`; for ryzens, `Block RThroughput: <=4.0`
So we could pick cost of `6`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111073
This one required quite a bit of assembly surgery, but the trend continues, so i think this is right.
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/EKWdj8cKT - for intels `Block RThroughput: <=32.0`; for ryzens, `Block RThroughput: <=24.0`
So could pick cost of `32`.
For store we have:
https://godbolt.org/z/zj4bb9P75 - for intels `Block RThroughput: =32.0`; for ryzens, `Block RThroughput: <=16.0`
So we could pick cost of `32`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111064
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/a6rxMG6ec - for intels `Block RThroughput: =16.0`; for ryzens, `Block RThroughput: <=12.0`
So could pick cost of `16`.
For store we have:
https://godbolt.org/z/ced1bdqc9 - for intels `Block RThroughput: =16.0`; for ryzens, `Block RThroughput: <=8.0`
So we could pick cost of `16`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111063
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/avq1oz98W - for intels `Block RThroughput: =8.0`; for ryzens, `Block RThroughput: =4.0`
So could pick cost of `8`.
For store we have:
https://godbolt.org/z/89PGMc1qs - for intels `Block RThroughput: =6.0`; for ryzens, `Block RThroughput: <=6.0`
So we could pick cost of `6`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111061
Finally, we are getting to the heavy-hitter stuff!
The only sched models that for cpu's that support avx2
but not avx512 are: haswell, broadwell, skylake, zen1-3
For load we have:
https://godbolt.org/z/7crGWoar6 - for intels `Block RThroughput: =4.0`; for ryzens, `Block RThroughput: <=2.0`
So could pick cost of `4`.
For store we have:
https://godbolt.org/z/T8aq3MszM - for intels `Block RThroughput: =5.0`; for ryzens, `Block RThroughput: <=2.0`
So we could pick cost of `5`.
I'm directly using the shuffling asm the llc produced,
without any manual fixups that may be needed
to ensure sequential execution.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111060
The current way to detect hostcalls by looking for "ockl_hostcall_internal()" function in the module seems to be not reliable enough. The LTO may rename the "ockl_hostcall_internal()" function when an application is compiled with "-fgpu-rdc", and MetadataStreamer pass to fail to detect hostcalls, therefore it does not set the "hidden_hostcall_buffer" kernel argument.
This change adds a new module flag: hostcall that can be used to detect whether GPU functions use host calls for printf.
Differential revision: https://reviews.llvm.org/D110337
Scalarize before narrowing because the narrowing implementation does not
work on vectors. This matches what we do for regular G_MUL.
Differential Revision: https://reviews.llvm.org/D111129
Simon Pilgrim