WebAssembly doesn't require full RPO; topological sorting is sufficient and
can preserve more of the MachineBlockPlacement ordering. Unfortunately, this
still depends a lot on heuristics, because while we use the
MachineBlockPlacement ordering as a guide, we can't use it in places where
it isn't topologically ordered. This area will require further attention.
llvm-svn: 260978
This avoids some complications updating LiveIntervals to be aware of the new
register lifetimes, because we can just compute new intervals from scratch
rather than describe how the old ones have been changed.
llvm-svn: 260971
Once a pointer is turned into a reference it cannot be nullptr, clang
rightfully warns about this assert being a tautology. Put the assert
before the reference is created.
llvm-svn: 260949
Summary:
In order to pass the tests, this required marking R_MIPS_16 relocations
as needing to point to the symbol and not the section.
Reviewers: vkalintiris, dsanders
Subscribers: dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D17200
llvm-svn: 260896
Summary:
This section is used for debug information and has no need to be
in memory at runtime. With this patch, LLVM now emits the same flags as
the GNU assembler. This patch also fixes an error when compiling
the Linux kernel, The error is that there are relocations within the
.pdr section in a VDSO.
Reviewers: vkalintiris, dsanders
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D17199
llvm-svn: 260879
If KMOVB not supported (require AVX512DQ) only KMOVW can be used so store size should be 2 bytes.
Differential Revision: http://reviews.llvm.org/D17138
llvm-svn: 260878
r180893 added an indirect include of llvm/Config/Targets.def to
llvm/Support/CodeGen.h, which in turn is included by things like
llvm/IR/Module.h. After a full build of LLVM and Clang, ninja had to
rebuild 1274 files after reconfiguring.
This commit strips CodeGen.h back down to just a pile of enums and moves
the expensive includes over to CodeGenCWrappers.h (which is only
included in two places). This gets ninja down to 88 files if you
reconfigure with, e.g., -DLLVM_TARGETS_TO_BUILD=X86.
llvm-svn: 260835
This patch attempts to represent a shuffle as a repeating shuffle (recognisable by is128BitLaneRepeatedShuffleMask) with the source input(s) in their original lanes, followed by a single permutation of the 128-bit lanes to their final destinations.
On AVX2 we can additionally attempt to match using 64-bit sub-lane permutation. AVX2 can also now match a similar 'broadcasted' repeating shuffle.
This patch has several benefits:
* Avoids prematurely matching with lowerVectorShuffleByMerging128BitLanes which can require both inputs to have their input lanes permuted before shuffling.
* Can replace PERMPS/PERMD instructions - although these are useful for cross-lane unary shuffling, they require their shuffle mask to be pre-loaded (and increase register pressure).
* Matching the repeating shuffle makes use of a lot of existing shuffle lowering.
There is an outstanding minor AVX1 regression (combine_unneeded_subvector1 in vector-shuffle-combining.ll) of a previously 128-bit shuffle + subvector splat being converted to a subvector splat + (2 instruction) 256-bit shuffle, I intend to fix this in a followup patch for review.
Differential Revision: http://reviews.llvm.org/D16537
llvm-svn: 260834
As shown in:
https://llvm.org/bugs/show_bug.cgi?id=23203
...we currently die because lowering believes that mfence is allowed without SSE2 on x86-64,
but the instruction def doesn't know that.
I don't know if allowing mfence without SSE is right, but if not, at least now it's consistently wrong. :)
Differential Revision: http://reviews.llvm.org/D17219
llvm-svn: 260828
Tests for the new scalarize all private access options will be
included with a future commit.
The only functional change is to make the split/scalarize behavior
for private access of > 4 element vectors to be consistent
with the flat/global handling. This makes the spilling worse
in the two changed tests.
llvm-svn: 260804
Replace spills to memory with spills to registers, if possible. This
applies mostly to predicate registers (both scalar and vector), since
they are very limited in number. A spill of a predicate register may
happen even if there is a general-purpose register available. In cases
like this the stack spill/reload may be eliminated completely.
This optimization will consider all stack objects, regardless of where
they came from and try to match the live range of the stack slot with
a dead range of a register from an appropriate register class.
llvm-svn: 260758
Rewrite the code to handle all pseudo-instructions in a single pass.
This temporarily reverts spill slot optimization that used general-
purpose registers to hold values of spilled predicate registers.
llvm-svn: 260696
Historically, AMD internal sp3 assembler has flat_store* addr, data
format. To match existing code and to enable reuse, change LLVM
definitions to match. Also update MC and CodeGen tests.
Differential Revision: http://reviews.llvm.org/D16927
Patch by: Nikolay Haustov
llvm-svn: 260694
Summary:
It is possible that the loop condition can be a boolean constant (infinite loop,
for example). So we sould handle constant condition in annotating a loop. This
patch adds this functionality to support annotating constant condition.
Reviewers: tstellarAMD, arsenm
Subscribers: llvm-commits, arsenm
Differential Revision: http://reviews.llvm.org/D15093
llvm-svn: 260692
We can generate the actual instructions from the intrinsics without the
need for pseudo-instructions. Also, since the intrinsics have a side-
effect in a form of a store, attempt to optimize away loads from the
store location.
llvm-svn: 260690
Summary:
Before this change, callee-save registers would be rounded up to even
pairs of GPRs and FPRs. This change eliminates these extra padding
load/stores, though it does keep the stack allocation the same size
unless both the GPR and FPR sets have an odd size, in which case one
full pair stack slot (16 bytes) is saved.
This optimization cannot currently be done for MachO targets since they
rely on a fast-path .debug_frame equivalent that can only encode
callee-save registers as pairs.
Reviewers: t.p.northover, rengolin, mcrosier, jmolloy
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D17000
llvm-svn: 260689
The DataLayout can calculate alignment of vectors based on the alignment
of the element type and the number of elements. In fact, it is the product
of these two values. The problem is that for vectors of N x i1, this will
return the alignment of N bytes, since the alignment of i1 is 8 bits. The
vector types of vNi1 should be aligned to N bits instead. Provide explicit
alignment for HVX vectors to avoid such complications.
llvm-svn: 260678
This was hardcoded to the static private size, but this
would be missing the offset and additional size for someday
when we have dynamic sizing.
Also stops always initializing flat_scratch even when unused.
In the future we should stop emitting this unless flat instructions
are used to access private memory. For example this will initialize
it almost always on VI because flat is used for global access.
llvm-svn: 260658
Introduce a subtarget feature for this, and leave the default with
the current behavior which assumes up to 16-byte loads/stores can
be used. The field also seems to have the ability to be set to 2 bytes,
but I'm not sure what that would be used for.
llvm-svn: 260651
This is just a trivial implementation:
- Support only arguments passed in registers.
- Support only "plain" arguments, i.e., no sext/zext attribute.
At this point, it is possible to play with the IRTranslator on AArch64:
llc -mtriple arm64-<vendor>-<os> -print-machineinstrs <input.ll> -o - -global-isel
For now, we only support the translation of program with adds and returns.
Follow-up patches are on their way to add a test case (the MIRParser is
not ready as it is).
llvm-svn: 260600
Summary:
It's possible to have resource descriptors and samplers stored in
VGPRs, either by a VMEM instruction or in the case of samplers,
floating-point calculations. When this happens, we need to use
v_readfirstlane to copy these values back to sgprs.
Reviewers: mareko, arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D17102
llvm-svn: 260599
Summary:
When we split SMRD instructions into two MUBUFs we were adding the users
of the newly created MUBUFs to the VALU worklist. However, the only
users these instructions had was the REG_SEQUENCE that was inserted
by splitSMRD when the original SMRD instruction was split.
We need to make sure to add the users of the original SMRD to the VALU
worklist before it is split.
I have a test case, but it requires one other bug fix, so it will be
added in a later commt.
Reviewers: mareko, arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D17101
llvm-svn: 260588
Summary:
Added support for "VOP3Only" attribute in VOP3bInst encoding.
Set VOP3Only=1 for V_DIV_SCALE_F64/32 insns.
Added support for multi-dest instructions in AMDGPUAs::cvt*().
Added lit test for "V_DIV_SCALE_F64|F32 vreg,vcc|sreg,vreg,vreg,vreg".
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm, SamWot, nhaustov, vpykhtin
Differential Revision: http://reviews.llvm.org/D16995
Patch By: Artem Tamazov
llvm-svn: 260560
This patch allows the mixing of scaled and unscaled load/stores to form
load/store pairs.
This is a reapplication of r259812, which had an incorrect assert. The
test_stur_str_no_assert() test is a reduced version of the issue hit in
the AArch64 self-host.
PR24465
llvm-svn: 260523
If the two operands to an instruction were both
subregisters of the same super register, it would incorrectly
think this counted as the same constant bus use.
This fixes the verifier error in fmin_legacy.ll which
was missing -verify-machineinstrs.
llvm-svn: 260495
Separate methods to convert parsed instructions to MCInst:
- VOP3 only instructions (always create modifiers as operands in MCInst)
- VOP2 instrunctions with modifiers (create modifiers as operands
in MCInst when e64 encoding is forced or modifiers are parsed)
- VOP2 instructions without modifiers (do not create modifiers
as operands in MCInst)
- Add VOP3Only flag. Pass HasMods flag to VOP3Common.
- Simplify code that deals with modifiers (-1 is now same as
0). This is no longer needed.
- Add few tests (more will be added separately).
Update error message now correct.
Patch By: Nikolay Haustov
Differential Revision: http://reviews.llvm.org/D16778
llvm-svn: 260483
Summary:
Refactor common value, scope, and label tracking logic out of DwarfDebug
into a common base class called DebugHandlerBase.
Update an old LLVM IR test case to avoid an assertion in LexicalScopes.
Reviewers: dblaikie, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D16931
llvm-svn: 260432
Summary:
This fixes a crash where subsequent spills would be unable to scavenge
a register. In particular, it fixes a crash in piglit's
spec@glsl-1.50@execution@geometry@max-input-components (the test still
has a shader that fails to compile because of too many SGPR spills, but
at least it doesn't crash any more).
This is a candidate for the release branch.
Reviewers: arsenm, tstellarAMD
Subscribers: qcolombet, arsenm
Differential Revision: http://reviews.llvm.org/D16558
llvm-svn: 260427
Instead of passing varargs directly on the user stack, allocate a buffer in
the caller's stack frame and pass a pointer to it. This simplifies the C
ABI (e.g. non-C callers of C functions do not need to use C's user stack if
they have their own mechanism) and allows further optimizations in the future
(e.g. fewer functions may need to use the stack).
Differential Revision: http://reviews.llvm.org/D17048
llvm-svn: 260421
The encodings for floating point conditions A(lways) and N(ever) were
incorrectly specified for the assembly parser, per Sparc manual v8 page
121. This change corrects that mistake.
Also, strangely, all of the branch instructions already had MC test
cases, except for the broken ones. Added the tests.
Patch by Chris Dewhurst
Differential Revision: http://reviews.llvm.org/D17074
llvm-svn: 260390
Now the parser supports `%got(sym)` expressions only but `%got(sym + const)`
variant is also valid and accepted by GAS.
Differential Revision: http://reviews.llvm.org/D16885
llvm-svn: 260305
I reinvented this functionality in http://reviews.llvm.org/D16828 because it was
hidden away as a static function. The changes in x86 are not based on a complete
audit. I suspect there are other possible uses there, and there are almost certainly
more potential users in other targets.
llvm-svn: 260295
Summary:
Fix case where a pre-inc/dec load/store would not be formed if the
add/sub that forms the inc/dec part of the operation was the first
instruction in the block being examined.
Reviewers: mcrosier, jmolloy, t.p.northover, junbuml
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D16785
llvm-svn: 260275
The logic to pair instructions and merge narrow instructions has become cloogy
and error prone. This patch beings to unravel these two similar, but distinct
optimizations.
llvm-svn: 260242
As mentioned in http://reviews.llvm.org/D16828 , the related masked load transform
will need this logic, so I'm moving it out to make that patch smaller.
llvm-svn: 260240
On AVX2 target we are poorly legalizing SIGN_EXTEND ops for which the input's legalized type doesn't have the same number of elements as the destination, resulting in an ANY_EXTEND followed by a SIGN_EXTEND_INREG.
This patch uses the existing SIGN_EXTEND -> SIGN_EXTEND_VECTOR_INREG combine to extend the input to the size of the result and using SIGN_EXTEND_VECTOR_INREG instead.
Differential Revision: http://reviews.llvm.org/D16994
llvm-svn: 260210
As discussed on PR26491, this patch adds support for lowering v4f32 shuffles to the MOVLHPS/MOVHLPS instructions. It also adds support for memory folding with their MOVLPS/MOVHPS load equivalents.
This first patch only really helps SSE1 targets as SSE2+ targets will widen the shuffle mask and use v2f64 equivalents (although they still combine to MOVLHPS/MOVHLPS for v2f64 splats). This will have to be addressed in a future patch, most likely when we add support for binary target shuffle combines.
Differential Revision: http://reviews.llvm.org/D16956
llvm-svn: 260168
Another opportunity to reduce masked stores: in D16691, we decided not to attempt the 'one mask element is set'
transform in InstCombine, but this should be a win for any AVX machine.
Code comments note that this transform could be extended for other targets / cases.
Differential Revision: http://reviews.llvm.org/D16828
llvm-svn: 260145
Summary:
We will hit this once we have enabled uniform branches. The
smrd-vccz-bug.ll test will be added with the uniform branch commit.
Reviewers: mareko, arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D16725
llvm-svn: 260137
This matches GCC and MSVC's behaviour, and saves on code size.
We were already not extending i1 return values on x86_64 after r127766. This
takes that patch further by applying it to x86 target as well, and also for i8
and i16.
The ABI docs have been unclear about the required behaviour here. The new i386
psABI [1] clearly states (Table 2.4, page 14) that i1, i8, and i16 return
vales do not need to be extended beyond 8 bits. The x86_64 ABI doc is being
updated to say the same [2].
Differential Revision: http://reviews.llvm.org/D16907
[1]. https://01.org/sites/default/files/file_attach/intel386-psabi-1.0.pdf
[2]. https://groups.google.com/d/msg/x86-64-abi/E8O33onbnGQ/_RFWw_ixDQAJ
llvm-svn: 260133
The combineX86ShufflesRecursively only supports unary shuffles, but was missing the opportunity to combine binary shuffles with a zero / undef second input.
This patch resolves target shuffle inputs, converting the shuffle mask elements to SM_SentinelUndef/SM_SentinelZero where possible. It then resolves the updated mask to check if we have created a faux unary shuffle.
Additionally, we now attempt to recursively call combineX86ShufflesRecursively for all input operands (we used to just recurse for unary integer shuffles and unary unpacks) - it safely returns early if its not a target shuffle.
Differential Revision: http://reviews.llvm.org/D16683
llvm-svn: 260063
Pulled out the code used by PSHUFB/VPERMV/VPERMV3 shuffle mask decoding into common helper functions.
The helper functions handle masks coming from BROADCAST/BUILD_VECTOR and ConstantPool nodes respectively.
llvm-svn: 260032
First step towards being able to decode AVX512 PMOVZX instructions without a massive bloat in the shuffle decode switch statement.
This should also make it easier to decode X86ISD::VZEXT target shuffles in the future.
llvm-svn: 259995
The current situation isn't great, because the amount of padding
requires is determined by the inverse order of the first encountered
use. We should eventually somehow sort these to minimize wasted space.
Another problem is the alignment of kernel arguments isn't
respected. The group_segment_alignment is always emitted as
the default 16, and typed arguments with higher alignments
or an explicitly set alignment are also ignored.
llvm-svn: 259912
Using the load immediate only when the immediate (whether signed or unsigned)
can fit in a 16-bit signed field. Namely, from -32768 to 32767 for signed and
0 to 65535 for unsigned. This patch also ensures that we sign-extend under the
right conditions.
llvm-svn: 259840
This only impacts the creation of pre-/post-index instructions. The bound was
set high enough such that it did not change code generation for SPEC200X.
llvm-svn: 259828
Choose between MOVD/MOVSS and MOVQ/MOVSD depending on the target vector type.
This has a lot fewer test changes than trying to add this to X86InstrInfo::setExecutionDomain.....
llvm-svn: 259816
This patch allows the mixing of scaled and unscaled load/stores to form
load/store pairs.
PR24465
http://reviews.llvm.org/D12116
Many thanks to Ahmed and Michael for fixes and code review.
This is a reapplication of r246769 and r259790. The tramp3d failure was caused
by an incorrect refactoring in the patch. Specifically, we weren't always
properly clearing the SExtIdx flag.
llvm-svn: 259812
During instruction selection, the AArch64 backend can recognise the
following pattern and generate an [U|S]MADDL instruction, i.e. a
multiply of two 32-bit operands with a 64-bit result:
(mul (sext i32), (sext i32))
However, when one of the operands is constant, the sign extension
gets folded into the constant in SelectionDAG::getNode(). This means
that the instruction selection sees this:
(mul (sext i32), i64)
...which doesn't match the pattern. Sign-extension and 64-bit
multiply instructions are generated, which are slower than one 32-bit
multiply.
Add a pattern to match this and generate the correct instruction, for
both signed and unsigned multiplies.
Patch by Chris Diamand!
llvm-svn: 259800
This patch adds support for consecutive (load/undef elements) 32-bit loads, followed by trailing undef/zero elements to be combined to a single MOVD load.
Differential Revision: http://reviews.llvm.org/D16729
llvm-svn: 259796
This patch allows the mixing of scaled and unscaled load/stores to form
load/store pairs.
PR24465
http://reviews.llvm.org/D12116
Many thanks to Ahmed and Michael for fixes and code review.
This is a reapplication of r246769, which was reverted in r246782 due to a
test-suite failure. I'm unable to reproduce the issue at this time.
llvm-svn: 259790
Use hash table (key is a memory operand) to store found LEA instructions to reduce compile time.
Differential Revision: http://reviews.llvm.org/D16404
llvm-svn: 259770
Add support for TLS access for Windows on ARM. This generates a similar access
to MSVC for ARM.
The changes to the tablegen data is needed to support loading an external symbol
global that is not for a call. The adjustments to the DAG to DAG transforms are
needed to preserve the 32-bit move.
llvm-svn: 259676
The GNU toolchain emits __aeabi_divmod for soft-divide on ARM cores
which happens to be a lot faster than __divsi3/__modsi3 when the core
has hardware divide instructions. Do the same here.
Fixes PR26450.
llvm-svn: 259657
MIPS ABI states that .sbss and .sdata sections must have SHF_MIPS_GPREL
flag. See Figure 4–7 on page 69 in the following document:
ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf.
Differential Revision: http://reviews.llvm.org/D15740
llvm-svn: 259641
Follow up to D16217 and D16729
This change uncovered an odd pattern where VZEXT_LOAD v4i64 was being lowered to a load of the lower v2i64 (so the 2nd i64 destination element wasn't being zeroed), I can't find any use/reason for this and have removed the pattern and replaced it so only the 1st i64 element is loaded and the upper bits all zeroed. This matches the description for X86ISD::VZEXT_LOAD
Differential Revision: http://reviews.llvm.org/D16768
llvm-svn: 259635
The purpose of PPCVSXFMAMutate is to elide copies by changing FMA forms
on PPC.
%vreg6<def> = COPY %vreg96
%vreg6<def,tied1> = XSMADDASP %vreg6<tied0>, %vreg5<kill>, %vreg7
;v6 = v6 + v5 * v7
is replaced by
%vreg5<def,tied1> = XSMADDMSP %vreg5<tied0>, %vreg7, %vreg96
;v5 = v5 * v7 + v96
This was broken in the case where the target register was also used as a
multiplicand. Fix this case by checking for it and replacing both uses
with the copied register.
%vreg6<def> = COPY %vreg96
%vreg6<def,tied1> = XSMADDASP %vreg6<tied0>, %vreg5<kill>, %vreg6
;v6 = v6 + v5 * v6
is replaced by
%vreg5<def,tied1> = XSMADDMSP %vreg5<tied0>, %vreg96, %vreg96
;v5 = v5 * v96 + v96
llvm-svn: 259617
If we can't assume the pointer value isn't within the bounds
of the object, it seems risky to try to replace the pointer
calculations.
llvm-svn: 259573
When promoting allocas to LDS, we know we are indexing
into a specific area just created, and the calculation
will also never overflow.
Also emit some of the muls as nsw nuw, because instcombine
infers this already from the range metadata. I think
putting this on the other adds and muls might be OK too,
but I'm not 100% sure.
llvm-svn: 259545
Summary:
Enables eip-based addressing, e.g.,
lea constant(%eip), %rax
lea constant(%eip), %eax
in MC, (used for the x32 ABI). EIP-base addressing is also valid in x86_64,
it is left enabled for that architecture as well.
Patch by João Porto
Differential Revision: http://reviews.llvm.org/D16581
llvm-svn: 259528
Re-commit of r258951 after fixing layering violation.
The BPF and WebAssembly backends had identical code for emitting errors
for unsupported features, and AMDGPU had very similar code. This merges
them all into one DiagnosticInfo subclass, that can be used by any
backend.
There should be minimal functional changes here, but some AMDGPU tests
have been updated for the new format of errors (it used a slightly
different format to BPF and WebAssembly). The AMDGPU error messages will
now benefit from having precise source locations when debug info is
available.
llvm-svn: 259498
description and changed the regression test accordingly.
The default configuration of a Cortex-R7 is to implement the
VFPv3-D16 architecture and the feature line as it was is too
restrictive.
llvm-svn: 259480
Fix a crash in `getMemOpBaseRegImmOfs` that happens if the base of
`MemOp` is a frame index memory operand. The fix is to have
`getMemOpBaseRegImmOfs` bail out in such cases. We can possibly be more
clever here, if needed.
llvm-svn: 259456
Officially, we don't acknowledge non-default configurations of MXCSR,
as getting there would require usage of the FENV_ACCESS pragma (at
least insofar as rounding mode is concerned).
We don't support the pragma, so we can assume that the default
rounding mode - round to nearest, ties to even - is always used.
However, it's inconsistent with the rest of the instruction set,
where MXCSR is always effective (unless otherwise specified).
Also, it's an unnecessary obstacle to the few brave souls that use
fenv.h with LLVM.
Avoid the hard-coded rounding mode for fp_to_f16; use MXCSR instead.
llvm-svn: 259448
These sets do linear searching in small mode; It is not a good idea to
use huge numbers as the small value here, save people from themselves by
adding a static_assert.
Differential Revision: http://reviews.llvm.org/D16706
llvm-svn: 259419
Summary:
This is an extension to the existing implementation of r242436 which
restricts to only select inputs. This version fixes missed opportunities
in pr26084 by attempting to lower conditional compare sequences of
and/or trees with setcc leafs. This will additionaly handle the case
when a tree with select input is not a conjunction-disjunction tree
but some of the sub trees are conjunction-disjunction trees.
Reviewers: jmolloy, t.p.northover, mcrosier, MatzeB
Subscribers: mcrosier, llvm-commits, junbuml, haicheng, mssimpso, gberry
Differential Revision: http://reviews.llvm.org/D16291
llvm-svn: 259387
Summary:
Factor out common code for callee-save register pair calculation. This
is intended to simplify follow-on changes that reduce the number of
registers saved/restored.
Depends on D16732
Reviewers: mcrosier, jmolloy, t.p.northover
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D16734
llvm-svn: 259384
We've found another bug in the code generation logic conditions for a
certain class of always-false conditions, those of the form
if ((a & 1) < 0)
These only reach the back end when compiling without optimization.
The bug was introduced by the choice of using TEST UNDER MASK
to implement a check for
if ((a & MASK) < VAL)
as
if ((a & MASK) == 0)
where VAL is less than the the lowest bit of MASK. This is correct
in all cases except for VAL == 0, in which case the original
condition is always false, but the replacement isn't.
Fixed by excluding that particular case.
llvm-svn: 259381
Summary:
Simplify callee-save register save/restore code generation by
remembering the size of the callee-save area when it is computed so we
don't have to scan the prologue/epilogue instructions again later to
reconstruct it.
This is intended to simplify follow-on changes that reduce the number of
registers saved/restored.
Reviewers: mcrosier, jmolloy, t.p.northover
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D16732
llvm-svn: 259365
Summary:
The bugs were:
* teq and similar take 4-bit unsigned immediates on microMIPS.
* teqi and similar have side-effects like teq do.
* shll_s.w and shra_r.w take 5-bit unsigned immediates.
* The various DSP ext* instructions take a 5-bit immediate.
* repl.qh takes an 8-bit unsigned immediate.
* repl.ph takes a 10-bit unsigned immediate.
* rddsp/wrdsp take a 10-bit unsigned immediate.
* teqi and similar take signed 16-bit immediates (10-bit for microMIPS).
* Out-of-range immediate macros for or/xor take a simm32/simm64 depending
on architecture. I'll fix the simm64 case properly when I reach simm32.
lui is a bit more lenient than GAS and accepts signed immediates in addition
to unsigned. This is because MipsMCExpr can produce signed values when
constant folding and it currently lacks a way of knowing it should fold to
an unsigned value.
Reviewers: vkalintiris
Subscribers: dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D15446
llvm-svn: 259360
Minor patch to trace back through target shuffles to the source of the inserted element in a (V)INSERTPS shuffle.
Differential Revision: http://reviews.llvm.org/D16652
llvm-svn: 259343
Previously the code assumed all uses of FI on loads and stores were as
addresses. This checks whether the use is the address or a value and
handles the latter case as it does for non-memory instructions.
llvm-svn: 259306
The previous code was incorrect (can't getReg a frameindex). We could instead optimize it to reduce tree height, but I'm not sure that's worthwhile yet because we then try to eliminate the frameindex.
This patch also fixes frame index elimination for operations which may load or store: it used to assume the base was operand 2 and immediate offset operand 1. That's not true for stores, where they're 4 and 3.
llvm-svn: 259305
The AMDGPUPromoteAlloca pass was emitting the read.local.size
calls, which with HSA was incorrectly selected to reading from
the offset mesa uses off of the kernarg pointer.
Error on intrinsics which aren't supported by HSA, and start
emitting the correct IR to read the workgroup size
out of the dispatch pointer.
Also initialize the pass so it can be tested with opt, and
start moving towards not depending on the subtarget as an
argument.
Start emitting errors for the intrinsics not handled with HSA.
llvm-svn: 259297
Only the dispatch.ptr intrinsic is supposed to be used now to get
the workgroup size, and the read.local.size intrinsics do not
work correctly.
llvm-svn: 259296
Refine the test for whether an instruction is in an expression tree so that
it detects when one tree ends and another begins, so we can place a block
at that point, rather than continuing to find the first instruction not in
a tree at all.
llvm-svn: 259294
The basic optimisation was to convert (mul $LHS, $complex_constant) into
roughly "(shl (mul $LHS, $simple_constant), $simple_amt)" when it was expected
to be cheaper. The original logic checks that the mul only has one use (since
we're mangling $complex_constant), but when used in even more complex
addressing modes there may be an outer addition that can pick up the wrong
value too.
I *think* the ARM addressing-mode problem is actually unreachable at the
moment, but that depends on complex assessments of the profitability of
pre-increment addressing modes so I've put a real check in there instead of an
assertion.
llvm-svn: 259228
Add support for frame pointer use in prolog/epilog.
Supports dynamic allocas but not yet over-aligned locals.
Target-independend CG generates SP updates, but we still need to write
back the SP value to memory when necessary.
llvm-svn: 259220
The trap instruction is emitted as a data-in-text rather
than an instruction. This patch uses the .inst directive
for emitting trap.
Differential Revision: http://reviews.llvm.org/D16684
llvm-svn: 259182
check that the sign extended constant fits into 16-bits if we want a
zero extended value, otherwise go ahead and put it together piecemeal.
Fixes PR26356.
llvm-svn: 259177
Since we only have pair - not single - nontemporal store instructions,
we have to extract the high part into a separate register to be able
to use them.
When the initial nontemporal codegen support was added, I wrote the
extract using the nonsensical UBFX [0,32[.
Use the correct LSR form instead.
llvm-svn: 259134
Summary:
Also delete all the stub functions that are identical to the
implementations in TargetInstrInfo.cpp.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D16609
llvm-svn: 259054
Enable truncate 128/256bit packed byte/word with AVX512BW but without AVX512VL, use 512bit instructions.
Differential Revision: http://reviews.llvm.org/D16531
llvm-svn: 259044
Author: milena.vujosevic.janicic
Reviewers: dsanders
FastIsel is not supported for microMIPS, thus it needs to be disabled.
Test micromips-zero-mat-uses.ll is deleted since the tested sequence of instructions is not generated for microMIPS without FastISel.
Differential Revision: http://reviews.llvm.org/D15892
llvm-svn: 259039
Re-commit of r258951 after fixing layering violation.
The related LLVM patch adds a backend diagnostic type for reporting
unsupported features, this adds a printer for them to clang.
In the case where debug location information is not available, I've
changed the printer to report the location as the first line of the
function, rather than the closing brace, as the latter does not give the
user any information. This also affects optimisation remarks.
Differential Revision: http://reviews.llvm.org/D16590
llvm-svn: 259035
move ptestm{q|d} intrinsics from patterns form (in td file) to the intrinsics table
Differential Revision: http://reviews.llvm.org/D16633
llvm-svn: 259029
This patch revamps the RegStackifier pass with a new tree traversal mechanism,
enabling three major new features:
- Stackification of values with multiple uses, using the result value of set_local
- More aggressive stackification of instructions with side effects
- Reordering operands in commutative instructions to enable more stackification.
llvm-svn: 259009
This patch is part of the work to make PPCLoopDataPrefetch
target-independent
(http://thread.gmane.org/gmane.comp.compilers.llvm.devel/92758).
As it was discussed in the above thread, getPrefetchDistance is
currently using instruction count which may change in the future.
llvm-svn: 258995
Summary:
Just does the simple allocation of a stack object and passes
a pointer to the callee.
Differential Revision: http://reviews.llvm.org/D16610
llvm-svn: 258989
Various bits we want to use the new ABI actually compile with "-arch armv7k
-miphoneos-version-min=9.0". Not ideal, but also not ridiculous given how
slices work.
llvm-svn: 258975
The BPF and WebAssembly backends had identical code for emitting errors
for unsupported features, and AMDGPU had very similar code. This merges
them all into one DiagnosticInfo subclass, that can be used by any
backend.
There should be minimal functional changes here, but some AMDGPU tests
have been updated for the new format of errors (it used a slightly
different format to BPF and WebAssembly). The AMDGPU error messages will
now benefit from having precise source locations when debug info is
available.
The implementation of DiagnosticInfoUnsupported::print must be in
lib/Codegen rather than in the existing file in lib/IR/ to avoid
introducing a dependency from IR to CodeGen.
Differential Revision: http://reviews.llvm.org/D16590
llvm-svn: 258951
Summary:
We didn't have entries in the commuting table for the 32-bit
instructions. I don't think we hit this problem now, but we
will once uniform branching is enabled. Tests will come in
a later commit.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D16600
llvm-svn: 258936
Summary:
This is a candidate for stable, along with all patches that add the "stoney"
processor.
Reviewers: tstellarAMD
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16485
llvm-svn: 258922
When no device name is specified, default to kaveri
for HSA since SI is not supported and it woud fail.
Default to "tahiti" instead of "SI" since these are
effectively the same, and tahiti is an actual device.
Move default device handling to the TargetMachine
rather than the AMDGPUSubtarget. The module ISA version
is computed from the device name provided with the target
machine, so the attributes printed by the AsmPrinter were
inconsistent with those computed in the subtarget.
Also remove DevName field from subtarget since it's redundant
with getCPU() in the superclass.
llvm-svn: 258901
This brings the compile time of Function.cpp from ~40s down to ~4s for
me locally. It also shaves off about 400KB of object file size in a
release+asserts build.
I also realized that the AMDGPU backend does not have any GCC builtin
names to match, so the extra lookup was a no-op. I removed it to silence
a zero-length string table array warning. There should be no functional
change here.
This change really ends the story of PR11951.
llvm-svn: 258897
The AMDGPU backend was the last user of the old StringMatcher
recognition code. Move it over to the new lookupLLVMIntrinsicName
funciton, which is now improved to handle all of the interesting edge
cases exposed by AMDGPU intrinsic names.
llvm-svn: 258875
Summary:
This patch is provided in preparation for removing autoconf on 1/26. The proposal to remove autoconf on 1/26 was discussed on the llvm-dev thread here: http://lists.llvm.org/pipermail/llvm-dev/2016-January/093875.html
"I felt a great disturbance in the [build system], as if millions of [makefiles] suddenly cried out in terror and were suddenly silenced. I fear something [amazing] has happened."
- Obi Wan Kenobi
Reviewers: chandlerc, grosbach, bob.wilson, tstellarAMD, echristo, whitequark
Subscribers: chfast, simoncook, emaste, jholewinski, tberghammer, jfb, danalbert, srhines, arsenm, dschuff, jyknight, dsanders, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D16471
llvm-svn: 258861
r258781 optimized memcpy/memmove/memcpy so the intrinsic call can return its first argument, but missed the frame index case. Teach it to ignore that case so C code doesn't assert out in these cases.
llvm-svn: 258851
Currently, AnalyzeBranch() fails non-equality comparison between floating points
on X86 (see https://llvm.org/bugs/show_bug.cgi?id=23875). This is because this
function can modify the branch by reversing the conditional jump and removing
unconditional jump if there is a proper fall-through. However, in the case of
non-equality comparison between floating points, this can turn the branch
"unanalyzable". Consider the following case:
jne.BB1
jp.BB1
jmp.BB2
.BB1:
...
.BB2:
...
AnalyzeBranch() will reverse "jp .BB1" to "jnp .BB2" and then "jmp .BB2" will be
removed:
jne.BB1
jnp.BB2
.BB1:
...
.BB2:
...
However, AnalyzeBranch() cannot analyze this branch anymore as there are two
conditional jumps with different targets. This may disable some optimizations
like block-placement: in this case the fall-through behavior is enforced even if
the fall-through block is very cold, which is suboptimal.
Actually this optimization is also done in block-placement pass, which means we
can remove this optimization from AnalyzeBranch(). However, currently
X86::COND_NE_OR_P and X86::COND_NP_OR_E are not reversible: there is no defined
negation conditions for them.
In order to reverse them, this patch defines two new CondCode X86::COND_E_AND_NP
and X86::COND_P_AND_NE. It also defines how to synthesize instructions for them.
Here only the second conditional jump is reversed. This is valid as we only need
them to do this "unconditional jump removal" optimization.
Differential Revision: http://reviews.llvm.org/D11393
llvm-svn: 258847
This patch adds support for trailing zero elements to VZEXT_LOAD loads (and checks that no zero elts occur within the consecutive load).
It also generalizes the 64-bit VZEXT_LOAD load matching to work for loads other than 2x32-bit loads.
After this patch it will also be easier to add support for other basic load patterns like 32-bit VZEXT_LOAD loads, PMOVZX and subvector load insertion.
Differential Revision: http://reviews.llvm.org/D16217
llvm-svn: 258798
Their opcodes are used as part of the VEX prefix in 64-bit mode. Clearly the disassembler implicitly decoded them as AVX instructions in 64-bit mode, but I think the AsmParser would have encoded them.
llvm-svn: 258793
Make comments and indentation more consistent.
Rearrange a few things to be in a more consistent order,
such as organizing subtarget features from those describing
an actual device property, and those used as options.
llvm-svn: 258789
I did my best to try to update all the uses in tests that
just happened to use the old ones to the newer intrinsics.
I'm not sure I got all of the immediate operand conversions
correct, since the value seems to have been ignored by the
old pattern but I don't think it really matters.
llvm-svn: 258787
Some of the special intrinsics now that now correspond to a instruction
also have special setting of some registers, e.g. llvm.SI.sendmsg sets
m0 as well as use s_sendmsg. Using these explicit register intrinsics
may be a better option.
Reading the exec mask and others may be useful for debugging. For this
I'm not sure this is entirely correct because we would want this to
be convergent, although it's possible this is already treated
sufficently conservatively.
llvm-svn: 258785
These calls return their first argument, but because LLVM uses an intrinsic
with a void return type, they can't use the returned attribute. Generalize
the store results pass to optimize these calls too.
llvm-svn: 258781
Step one towards using a simple binary search to lookup intrinsic IDs
instead of our crazy table generated switch+memcmp+startswith code that
makes Function.cpp take about a minute to compile. See PR24785 and
PR11951 for why we should do this.
The X86 backend contains tables that need to be sorted on intrinsic ID,
so reorder those.
llvm-svn: 258757
There's a special case in EmitLoweredSelect() that produces an improved
lowering for cmov(cmov) patterns. However this special lowering is
currently broken if the inner cmov has multiple users so this patch
stops using it in this case.
If you wonder why this wasn't fixed by continuing to use the special
lowering and inserting a 2nd PHI for the inner cmov: I believe this
would incur additional copies/register pressure so the special lowering
does not improve upon the normal one anymore in this case.
This fixes http://llvm.org/PR26256 (= rdar://24329747)
llvm-svn: 258729
Instructions can be DCE'd after the RegStackify pass. If the instruction which
would be the pop for what would be a push is removed, don't use a push.
llvm-svn: 258694
This patch was originally committed as r257885, but was reverted due to windows
failures. The cause of these failures has been fixed under r258677, hence
re-committing the original patch.
llvm-svn: 258683
This patch was originally committed as r257884, but was reverted due to windows
failures. The cause of these failures has been fixed under r258677, hence
re-committing the original patch.
llvm-svn: 258682
This patch was originally committed as r257883, but was reverted due to windows
failures. The cause of these failures has been fixed under r258677, hence
re-committing the original patch.
llvm-svn: 258681
VPMADD52LUQ - Packed Multiply of Unsigned 52-bit Integers and Add the Low 52-bit Products to Qword Accumulators
VPMADD52HUQ - Packed Multiply of Unsigned 52-bit Unsigned Integers and Add High 52-bit Products to 64-bit Accumulators
Differential Revision: http://reviews.llvm.org/D16407
llvm-svn: 258680
This was originally committed as r255762, but reverted as it broke windows
bots. Re-commitiing the exact same patch, as the underlying cause was fixed by
r258677.
ARMv8.2-A adds 16-bit floating point versions of all existing VFP
floating-point instructions. This is an optional extension, so all of
these instructions require the FeatureFullFP16 subtarget feature.
The assembly for these instructions uses S registers (AArch32 does not
have H registers), but the instructions have ".f16" type specifiers
rather than ".f32" or ".f64". The top 16 bits of each source register
are ignored, and the top 16 bits of the destination register are set to
zero.
These instructions are mostly the same as the 32- and 64-bit versions,
but they use coprocessor 9 rather than 10 and 11.
Two new instructions, VMOVX and VINS, have been added to allow packing
and extracting two 16-bit floats stored in the top and bottom halves of
an S register.
New fixup kinds have been added for the PC-relative load and store
instructions, but no ELF relocations have been added as they have a
range of 512 bytes.
Differential Revision: http://reviews.llvm.org/D15038
llvm-svn: 258678
This is a fix for https://llvm.org/bugs/show_bug.cgi?id=22796.
The previous implementation of ClassInfo::operator< allowed cycles of classes
such that x < y < z < x, meaning that a list of them cannot be correctly
sorted, and the sort order could differ with different standard libraries.
The original implementation sorted classes by ValueName if they were otherwise
equal. This isn't strictly necessary, but some backends seem to accidentally
rely on it. If I reverse this comparison I get 8 test failures spread across
the AArch64, Mips and X86 backends, so I have left it in until those backends
can be fixed.
There was one case in the X86 backend where the observable behaviour of the
assembler is changed by this patch. This was because some of the memory asm
operands were not marked as children of X86MemAsmOperand.
Differential Revision: http://reviews.llvm.org/D16141
llvm-svn: 258677
Use AVX1 FP instructions (vmaskmovps/pd) in place of the AVX2 int instructions (vpmaskmovd/q).
Differential Revision: http://reviews.llvm.org/D16528
llvm-svn: 258675
Changes in X86.td:
I set features of Intel processors in incremental form: IVB = SNB + X HSW = IVB + X ..
I added Skylake client processor and defined it's features
FeatureADX was missing on KNL
Added some new features to appropriate processors SMAP, IFMA, PREFETCHWT1, VMFUNC and others
Differential Revision: http://reviews.llvm.org/D16357
llvm-svn: 258659
Generalised mask generation / subvector extraction to use the input/output types directly instead of an if/else through all the currently accepted types.
llvm-svn: 258645
Summary:
Previously, we would just output "foo = bar" in the assembly, and then
ptxas would choke. Now we die before emitting any invalid code.
Reviewers: echristo
Subscribers: jholewinski, llvm-commits, jhen, tra
Differential Revision: http://reviews.llvm.org/D16490
llvm-svn: 258638
If the INSERTPS zeroes out all the referenced elements from either of the 2 input vectors (and the input is not already UNDEF), then set that input to UNDEF to reduce dependencies.
llvm-svn: 258622
Some of the conditions necessary to produce ccmp sequences were only
checked in recursive calls to emitConjunctionDisjunctionTree() after
some of the earlier expressions were already built. Move all checks over
to isConjunctionDisjunctionTree() so they are all checked before we
start emitting instructions.
Also rename some variable to better reflect their usage.
llvm-svn: 258605
Previously it failed to add NumArgRegs to the offset and so clobbered an
already-used register. Now just start the numbering after the arg regs
and don't duplicate the add. Test coverage for this coming shortly with
the implementation of byval.
llvm-svn: 258597
The intrinsic target prefix should match the target name
as it appears in the triple.
This is not yet complete, but gets most of the important ones.
llvm.AMDGPU.* intrinsics used by mesa and libclc are still handled
for compatability for now.
llvm-svn: 258557
The promote alloca pass didn't handle these intrinsics and crashed.
These intrinsics should accept any address space, but for now just
erase them to avoid breaking.
llvm-svn: 258537
The current behavior is incorrect, as the two CCs returned by
changeFPCCToAArch64CC, intended to be OR'ed, are instead used
in an AND ccmp chain.
Consider:
define i32 @t(float %a, float %b, float %c, float %d, i32 %e, i32 %f) {
%cc1 = fcmp one float %a, %b
%cc2 = fcmp olt float %c, %d
%and = and i1 %cc1, %cc2
%r = select i1 %and, i32 %e, i32 %f
ret i32 %r
}
Assuming (%a < %b) and (%c < %d); we used to do:
fcmp s0, s1 # nzcv <- 1000
orr w8, wzr, #0x1 # w8 <- 1
csel w9, w8, wzr, mi # w9 <- 1
csel w8, w8, w9, gt # w8 <- 1
fcmp s2, s3 # nzcv <- 1000
cset w9, mi # w9 <- 1
tst w8, w9 # (w8 & w9) == 1, so: nzcv <- 0000
csel w0, w0, w1, ne # w0 <- w0
We now do:
fcmp s2, s3 # nzcv <- 1000
fccmp s0, s1, #0, mi # mi, so: nzcv <- 1000
fccmp s0, s1, #8, le # !le, so: nzcv <- 1000
csel w0, w0, w1, pl # !pl, so: w0 <- w1
In other words, we transformed:
(c < d) && ((a < b) || (a > b))
into:
(c < d) && (a u>= b) && (a u<= b)
whereas, per De Morgan's, we wanted:
(c < d) && !((a u>= b) && (a u<= b))
Note that this problem doesn't occur in the test-suite.
changeFPCCToAArch64CC produces disjunct CCs; here, one -> mi/gt.
We can't represent that in the fccmp chain; it can't express
arbitrary OR sequences, as one comment explains:
In general we can create code for arbitrary "... (and (and A B) C)"
sequences. We can also implement some "or" expressions, because
"(or A B)" is equivalent to "not (and (not A) (not B))" and we can
implement some negation operations. [...] However there is no way
to negate the result of a partial sequence.
Instead, introduce changeFPCCToANDAArch64CC, which produces the
conjunct cond codes:
- (a one b)
== ((a olt b) || (a ogt b))
== ((a ord b) && (a une b))
- (a ueq b)
== ((a uno b) || (a oeq b))
== ((a ule b) && (a uge b))
Note that, at first, one might think that, when PushNegate is true,
we should use the disjunct CCs, in effect doing:
(a || b)
= !(!a && !(b))
= !(!a && !(b1 || b2)) <- changeFPCCToAArch64CC(b, b1, b2)
= !(!a && !b1 && !b2)
However, we can take advantage of the fact that the CC is already
negated, which lets us avoid special-casing PushNegate and doing
the simpler to reason about:
(a || b)
= !(!a && (!b))
= !(!a && (b1 && b2)) <- changeFPCCToANDAArch64CC(!b, b1, b2)
= !(!a && b1 && b2)
This makes both emitConditionalCompare cases behave identically,
and produces correct ccmp sequences for the 2-CC fcmps.
llvm-svn: 258533
We verify that the op tree is eligible for CCMP emission in
isConjunctionDisjunctionTree, but it's also possible that
emitConjunctionDisjunctionTree fails later.
The initial check is useful, as it avoids building nodes
that will get discarded.
Still, make sure that inconsistencies don't happen with
an assert.
llvm-svn: 258532
Summary:
SETCC with f16 vectors has OperationAction set to Expand but still gets
lowered to FCM* intrinsics based on its result type. This patch skips
lowering of VSETCC if the operand is an f16 vector.
v4 and v8 tests included.
Reviewers: ab, jmolloy
Subscribers: srhines, llvm-commits
Differential Revision: http://reviews.llvm.org/D15361
llvm-svn: 258471
Better handling of the annoying pshuflw/pshufhw ops which only shuffle lower/upper halves of a vector.
Added vXi16 unary shuffle support for cases where i16 elements (from the same half of the source) are being splatted to the whole of one of the halves. This avoids the general lowering case which must shuffle the 32-bit elements first - meaning that we used to end up with unnecessary duplicate pshuflw/pshufhw shuffles.
Note this has the side effect of a lot of SSSE3 test cases no longer needing to use PSHUFB, as it falls below the 3 op combine threshold for when PSHUFB is typically worth it. I've raised PR26183 to discuss if the threshold should be changed and whether we need to make it more specific to the target CPU.
Differential Revision: http://reviews.llvm.org/D14901
llvm-svn: 258440
Summary:
And use it in PPCLoopDataPrefetch.cpp.
@hfinkel, please let me know if your preference would be to preserve the
ppc-loop-prefetch-cache-line option in order to be able to override the
value of TTI::getCacheLineSize for PPC.
Reviewers: hfinkel
Subscribers: hulx2000, mcrosier, mssimpso, hfinkel, llvm-commits
Differential Revision: http://reviews.llvm.org/D16306
llvm-svn: 258419
Summary:
This is now the same as the behaviour of the GNU assembler. This was done
as it is required in order to build the Linux kernel with the integrated
assembler enabled.
Reviewers: dsanders, vkalintiris
Subscribers: dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D13594
llvm-svn: 258400
There's an overloading of the "movsd" and "cmpsd" instructions, e.g. movsd can be either "Move Data from String to String" or "Move or Merge Scalar Double-Precision Floating-Point Value".
The former should produce warnings when parsing a memory operand that is not ESI/EDI, but the latter should not.
Fixed the code to produce warnings only after making sure we're dealing with the first case.
Expanded the tests of the produced warnings + fixed RUN line of the test so that it would check both stdout and stderr
Differential Revision: http://reviews.llvm.org/D16359
llvm-svn: 258393
Summary:
Currently the SI scheduler can be selected via command line option,
but it turned out it would be better if it was selectable via a Target Attribute.
This patch adds "si-scheduler" attribute to the backend.
Reviewers: tstellarAMD, echristo
Subscribers: echristo, arsenm
Differential Revision: http://reviews.llvm.org/D16192
llvm-svn: 258386
Summary:
While working on uniform branching, I've hit a few cases where we emit
i1 SETCC operations.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D16233
llvm-svn: 258352
There was support for writing the AArch64 big endian data fixup entries in
the .eh_frame section in BE. This is changed to write all such fixup
entries in BE with no restriction on the section. This is similar to
the existing support for fixup entries for ARM.
A test is added to check the length field in the .debug_line section as
this is an example of where such a fixup occurs.
Differential Revision: http://reviews.llvm.org/D16064
llvm-svn: 258320
There was a bug in my rL258132 because there's an overloading of the "movsd" and "cmpsd" instructions, e.g. movsd can be either "Move Data from String to String" (the case I wanted to handle) or "Move or Merge Scalar Double-Precision Floating-Point Value" (the case that causes the asserts).
Added code for escaping the unfamiliar scenarios and falling back to old behviour.
Also changed the asserts to llvm_unreachable.
llvm-svn: 258312
The AArch64 .inst directive was implemented using EmitIntValue, which resulted
in both $x and $d (code and data) mapping symbols being emitted at the same
address. This fixes it to only emit the $x mapping symbol.
EmitIntValue also emits the value in big-endian order when targeting big-endian
systems, but instructions are always emitted in little-endian order for
AArch64.
Differential Revision: http://reviews.llvm.org/D16349
llvm-svn: 258308
calling convention.
The implementation of the related callbacks in the x86 backend for such
functions are not ready to deal with a prologue block that is not the entry
block of the function.
This fixes PR26107, but the longer term solution would be to fix those callbacks.
llvm-svn: 258221
As vector shuffles can only reference two inputs many (V)INSERTPS patterns end up being split over two targets shuffles.
This patch adds combines to attempt to combine (V)INSERTPS nodes with input/output nodes that are just zeroing out these additional vector elements.
Differential Revision: http://reviews.llvm.org/D16072
llvm-svn: 258205
Summary:
GEPOperator: provide getResultElementType alongside getSourceElementType.
This is made possible by adding a result element type field to GetElementPtrConstantExpr, which GetElementPtrInst already has.
GEP: replace get(Pointer)ElementType uses with get{Source,Result}ElementType.
Reviewers: mjacob, dblaikie
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D16275
llvm-svn: 258145
Teach the register stackifier to rematerialize constants that have multiple
uses instead of leaving them in registers. In the WebAssembly encoding, it's
the same code size to materialize most constants as it is to read a value
from a register.
llvm-svn: 258142
According to x86 spec "xlat m8" is a legal instruction and it is equivalent to "xlatb".
Differential Revision: http://reviews.llvm.org/D15150
llvm-svn: 258135
The following are legal according to X86 spec:
ins mem, DX
outs DX, mem
lods mem
stos mem
scas mem
cmps mem, mem
movs mem, mem
Differential Revision: http://reviews.llvm.org/D14827
llvm-svn: 258132
This breaks the tests that were meant for testing
64-bit inline immediates, so move those to shl where
they won't be broken up.
This should be repeated for the other related bit ops.
llvm-svn: 258095
AVX2 can only broadcast from the zero'th element of a vector, but if the broadcastable element is the zero'th element of a 128-bit subvector its advantageous to extract the subvector, broadcast from that and avoid the loading of shuffle mask data that would be needed for VPERMPS/VPERMD. The only exception being when the source type is 4f64 or 4i64 which can directly use the immediate shuffle VPERMPD/VPERMQ directly.
Differential Revision: http://reviews.llvm.org/D16050
llvm-svn: 258081
When the shift immediate is zero, PKHTB is an alias for PKHBT, but the order of
the input operands needs to be swapped.
Differential Revision: http://reviews.llvm.org/D16288
llvm-svn: 258044
The feature flag is for VPERMB,VPERMI2B,VPERMT2B and VPMULTISHIFTQB instructions.
More about the instruction can be found in:
hattps://software.intel.com/sites/default/files/managed/07/b7/319433-023.pdf
Differential Revision: http://reviews.llvm.org/D16190
llvm-svn: 258012
Added support for the extraction of the upper 128-bit subvectors for lower/upper half undef shuffles if it would reduce the number of extractions/insertions or avoid loads of AVX2 permps/permd shuffle masks.
Minor follow up to D15477.
llvm-svn: 258000
%RBP can't be handled explicitly. We generate the following code:
pushq %rbp
movq %rsp, %rbp
...
movq %rbx, (%rbp) ## 8-byte Spill
where %rbp will be overwritten by the spilled value.
The fix is to let PEI handle %RBP.
PR26136
llvm-svn: 257997
FIXME: Add more targets to use emutls into clang/test/Driver/emulated-tls.cpp.
FIXME: Add cygwin tests into llvm/test/CodeGen/X86. Working in progress.
llvm-svn: 257984
WebAssembly's stack will never be executable by default, so it isn't
necessary to declare .note.GNU-stack sections to request a non-executable
stack.
Differential Revision: http://reviews.llvm.org/D15969
llvm-svn: 257962
When we have a single basic block, the explicit copy-back instructions should
be inserted right before the terminator. Before this fix, they were wrongly
placed at the beginning of the basic block.
PR26136
llvm-svn: 257930
When we have a single basic block, the explicit copy-back instructions should
be inserted right before the terminator. Before this fix, they were wrongly
placed at the beginning of the basic block.
I will commit fixes to other platforms as well.
PR26136
llvm-svn: 257929
When we have a single basic block, the explicit copy-back instructions should
be inserted right before the terminator. Before this fix, they were wrongly
placed at the beginning of the basic block.
I will commit fixes to other platforms as well.
PR26136
llvm-svn: 257925
Some compilers don't do exhaustive switch checking. For those compilers,
add an initialization to prevent un-initialized variable warnings from
firing. For compilers with exhaustive switch checking, we still get a
guarantee that the switch is exhaustive, and hence the initializations
are redundant, and a non-functional change.
llvm-svn: 257923
# The first commit's message is:
Revert "[ARM] Add DSP build attribute and extension targeting"
This reverts commit b11cc50c0b4a7c8cdb628abc50b7dc226ff583dc.
# This is the 2nd commit message:
Revert "[ARM] Add new system registers to ARMv8-M Baseline/Mainline"
This reverts commit 837d08454e3e5beb8581951ac26b22fa07df3cd5.
llvm-svn: 257916
This method has no callers.
Also remove X86ELFRelocationInfo.cpp and X86MachORelocationInfo.cpp
which only existed to provide an implementation of that method.
Ok'd by Rafael and Jim.
llvm-svn: 257859
Summary:
This pass may modify the Cmp operands. However, the flag reg may be used by both the branch and CSEL.
Modifying CMP will have side effect on CSEL.
Reviewers: t.p.northover
Subscribers: llvm-commits, aemerson, rengolin
Differential Revision: http://reviews.llvm.org/D16147
llvm-svn: 257844
A recent change introduced this assertion failure in some corner cases.
Repro:
mkdir /s/wasm/torture-out ; time /s/wasm/waterfall/src/compile_torture_tests.py --c /s/llvm/out/bin/clang --cxx /s/llvm/out/bin/clang++ --testsuite /s/gcc/gcc/testsuite --fails /s/llvm/llvm/lib/Target/WebAssembly/known_gcc_test_failures.txt --out /s/wasm/torture-out
Or look on the wasm integration bot:
https://build.chromium.org/p/client.wasm.llvm/console
llvm-svn: 257733
We rely on HasOpaqueSPAdjustment not changing after we've calculated
things based on it. Things like whether or not we can use 'rep;movs' to
copy bytes around, that sort of thing. If it changes, invariants in the
backend will quietly break. This situation arose when we had a call to
memcpy *and* a COPY of the FLAGS register where we would attempt to
reference local variables using %esi, a register that was clobbered by
the 'rep;movs'.
This fixes PR26124.
llvm-svn: 257730
Summary:
v2: Make ReturnsVoid private, so that I can another 8 lines of code and
look more productive.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16034
llvm-svn: 257622
Summary:
Return values can be stored in SGPRs (i32) and VGPRs (f32).
This will be used by functions which expect some bytecode or other binary to
be appended at the end. It allows defining in which registers the return
values will be stored.
v2: don't do this for compute shaders
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16033
llvm-svn: 257621
Summary:
It is off by default, but can be used
with --misched=si
Patch by: Axel Davy
Reviewers: arsenm, tstellarAMD, nhaehnle
Subscribers: nhaehnle, solenskiner, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D11885
llvm-svn: 257609
The global entry point prologue currently assumes that the TOC
associated with a function is less than 2GB away from the function
entry point. This is always true when using the medium or small
code model, but may not be the case when using the large code model.
This patch adds a new variant of the ELFv2 global entry point prologue
that lifts the 2GB restriction when building with -mcmodel=large.
This works by emitting a quadword containing the distance from the
function entry point to its associated TOC immediately before the
entry point, and then using a prologue like:
ld r2,-8(r12)
add r2,r2,r12
Since creation of the entry point prologue is now split across two
separate routines (PPCLinuxAsmPrinter::EmitFunctionEntryLabel emits
the data word, PPCLinuxAsmPrinter::EmitFunctionBodyStart the prolog
code), I've switched to using named labels instead of just temporaries
to indicate the locations of the global and local entry points and the
new TOC offset data word.
These names are provided by new routines in PPCFunctionInfo modeled
after the existing PPCFunctionInfo::getPICOffsetSymbol.
Note that a corresponding change was committed to GCC here:
https://gcc.gnu.org/ml/gcc-patches/2015-12/msg00355.html
Reviewers: hfinkel
Differential Revision: http://reviews.llvm.org/D15500
llvm-svn: 257597
Summary:
With the ability to concatenate shader binaries, the limit of 15 no longer
applies.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16031
llvm-svn: 257592
Summary:
This allows Mesa to pass initial SPI_PS_INPUT_ADDR to LLVM.
The register assigns VGPR locations to PS inputs, while the ENA register
determines whether or not they are loaded.
Mesa needs to set some inputs as not-movable, so that a pixel shader prolog
binary appended at the beginning can assume where some inputs are.
v2: Make PSInputAddr private, because there is never enough silly getters
and setters for people to read.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16030
llvm-svn: 257591
Summary: ret.ll will contain a test for this
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm
Differential Revision: http://reviews.llvm.org/D16029
llvm-svn: 257590
Make x86 OptimizeLEAs pass remove LEA instruction if there is another LEA
(in the same basic block) which calculates address differing only be a
displacement. Works only for -Oz.
Differential Revision: http://reviews.llvm.org/D13295
llvm-svn: 257589
AnalyzeBranch on X86 (and, previously, SPARC, which implementation was
copied from X86) tries to modify the branches based on block
layout (e.g. checking isLayoutSuccessor), when AllowModify is true.
The rest of the architectures leave that up to the caller, which can
call InsertBranch, RemoveBranch, and ReverseBranchCondition as
appropriate. That appears to be the preferred way to do it nowadays.
This commit makes SPARC like the rest: replaces AnalyzeBranch with an
implementation cribbed from AArch64, and adds a ReverseBranchCondition
implementation.
Additionally, a test-case has been added (also cribbed from AArch64)
demonstrating that redundant branch sequences no longer get emitted.
E.g., it used to emit code like this:
bne .LBB1_2
nop
ba .LBB1_1
nop
.LBB1_2:
And now emits:
cmp %i0, 42
be .LBB1_1
nop
llvm-svn: 257572
Summary:
BFC instructions are available in ARMv6T2 and above.
Reviewers: t.p.northover
Subscribers: aemerson
Differential Revision: http://reviews.llvm.org/D16076
llvm-svn: 257546
VMOVs are not strictly speaking cheap, but they are as expensive as a vector
copy (VORR), so we should prefer rematerialization over splitting when it
applies.
rdar://problem/23754176
llvm-svn: 257545
Only non-weighted predicates were handled in PPCInstrInfo::insertSelect. Handle
the weighted predicates as well.
This latent bug was triggered by r255398, because it added use of the
branch-weighted predicates.
While here, switch over an enum instead of an int to get the compiler to enforce
totality in the future.
llvm-svn: 257518
A request has been made to the official registry, but an official value is
not yet available. This patch uses a temporary value in order to support
development. When an official value is recieved, the value of EM_WEBASSEMBLY
will be updated.
llvm-svn: 257517
Refactor .param, .result, .local, and .endfunc, as directives, using the
proper MCTargetStreamer mechanism, rather than fake instructions.
llvm-svn: 257511
This patch changes the way labels are referenced. Instead of referencing the
basic-block label name (eg. .LBB0_0), instructions now just have an immediate
which indicates the depth in the control-flow stack to find a label to jump to.
This makes them much closer to what we expect to have in the binary encoding,
and avoids the problem of basic-block label names not being explicit in the
binary encoding.
Also, it terminates blocks and loops with end_block and end_loop instructions,
rather than basic-block label names, for similar reasons.
This will also fix problems where two constructs appear to have the same label,
because we no longer explicitly use labels, so consumers that need labels will
presumably create their own labels, and presumably they won't reuse labels
when they do.
This patch does make the code a little more awkward to read; as a partial
mitigation, this patch also introduces comments showing where the labels are,
and comments on each branch showing where it's branching to.
llvm-svn: 257505
This is a very limited implementation of DFG-based copy propagation.
It only handles actual COPY instructions (does not handle other equivalents
such as add-immediate with a 0 operand).
The major limitation is that it does not update the DFG: that will be the
change required to make it more robust (hopefully coming up soon).
llvm-svn: 257490
Summary: The result register is the second operand as per the other mt* instructions.
Reviewers: vkalintiris
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D15993
llvm-svn: 257478
Target independent, SSA-based data flow framework for representing
data flow between physical registers.
This commit implements the creation of the actual data flow graph.
llvm-svn: 257477
Summary:
This fixes three bugs, in all of which state is not or incorrecly reset between
objects (i.e. when reusing the same pass manager to create multiple object
files):
1) AttributeSection needs to be reset to nullptr, because otherwise the backend
will try to emit into the old object file's attribute section causing a
segmentation fault.
2) MappingSymbolCounter needs to be reset, otherwise the second object file
will start where the first one left off.
3) The MCStreamer base class resets the Streamer's e_flags settings. Since
EF_ARM_EABI_VER5 is set on streamer creation, we need to set it again
after the MCStreamer was rest.
Also rename Reset (uppser case) to EHReset to avoid confusion with
reset (lower case).
Reviewers: rengolin
Differential Revision: http://reviews.llvm.org/D15950
llvm-svn: 257473
(64 to 128-bit) matches against the pattern fragment 'vzmovl_v2i64'
(a zero-extended 64-bit load).
However, a change in r248784 teaches the instruction combiner that only
the lower 64 bits of the input to a 128-bit vcvtph2ps are used. This means
the instruction combiner will ordinarily optimize away the upper 64-bit
insertelement instruction in the zero-extension and so we no longer select
the memory-register form. To fix this a new pattern has been added.
Differential Revision: http://reviews.llvm.org/D16067
llvm-svn: 257470
Currently WebAssembly has two kinds of relocations; data addresses and
function addresses. This adds ELF relocations for them, as well as an
MC symbol kind to indicate which type of relocation is needed.
llvm-svn: 257416
Always expect tglobaladdr and texternalsym to be wrapped in
WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so
that it can special-case global addresses, as they can be folded in more
cases.
Unfortunately this doesn't enable any new optimizations yet due to
SelectionDAG limitations. I'll be submitting changes to the SelectionDAG
infrastructure, along with tests, in a separate patch.
llvm-svn: 257394
The hardware instruction's output on 0 is -1 rather than 32.
Eliminate a test and select to -1. This removes an extra instruction
from the compatability function with HSAIL's firstbit instruction.
llvm-svn: 257352
Summary:
It actually takes an offset into the current PC-region.
This fixes the 'expr' command in lldb.
Reviewers: vkalintiris, jaydeep, bhushan
Subscribers: dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D16054
llvm-svn: 257339
In the OptimizeLEA pass keep instructions' positions in the basic block saved and use them for calculation of the distance between two instructions instead of std::distance. This reduces complexity of the pass from O(n^3) to O(n^2) and thus the compile time.
Differential Revision: http://reviews.llvm.org/D15692
llvm-svn: 257328
AVX1 v8i32/v4i64 shuffles are bitcasted to v8f32/v4f64, this patch peeks through any bitcast to check for a load node to allow broadcasts to occur.
This is a re-commit of r257055 after r257264 fixed 32-bit broadcast loads of i64 scalars.
llvm-svn: 257266
Summary:
r255334 matches bit-reverse pattern in InstCombine and generates calls to Instrinsic::bitreverse.
RBIT instruction is only available for ARMv6t2 and above. This patch has the intrinsic expanded during legalization for ARMv4 and ARMv5.
Patch by Z. Zheng <zhaoshiz@codeaurora.org>
Reviewers: apazos, jmolloy, weimingz
Subscribers: aemerson, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D15932
llvm-svn: 257188
Summary: In ARMConstantIslandPass, which runs after Shrink Wrap pass, long jumps will be fixed up as BL (tBfar) which depends on spilling LR in epilogue. However, shrink-wrap may remove the LR, which causes issues when the function returns.
Reviewers: qcolombet, rengolin
Subscribers: aemerson, rengolin
Differential Revision: http://reviews.llvm.org/D15984
llvm-svn: 257187
This patch corresponds to review:
http://reviews.llvm.org/D15930
Moves to and from CR fields depend on shifts/masks that depend on the
target/source CR field. Thus, post-ra anti-dep breaking must not later
change that CR register assignment.
llvm-svn: 257168
This is a fix for bug http://llvm.org/bugs/show_bug.cgi?id=25839.
For a PIC TLS variable access in a function, prologue (mflr followed by std and
stdu) gets scheduled after a tls_get_addr call. tls_get_addr messed up LR but
no one saves/restores it.
Also added a test for save/restore clobbered registers during calling __tls_get_addr.
Patch by Tim Shen
llvm-svn: 257137
.zero is confusing when used with two arguments. Documentation:
This directive emits SIZE 0-valued bytes. SIZE must be an absolute
expression. This directive is actually an alias for the '.skip'
directive so in can take an optional second argument of the value to
store in the bytes instead of zero. Using '.zero' in this way would be
confusing however.
Ref: https://sourceware.org/bugzilla/show_bug.cgi?id=18353
Hexagon and Sparc do the same, and it's all the same to WebAssembly so
let's pick the less confusing of the two.
llvm-svn: 257111
Previously we only supported putting the FI into memory operand offset
fields if there was nothing there already. Now combine them.
Differential Revision: http://reviews.llvm.org/D15941
llvm-svn: 257084
The MC assembler doesn't like using the empty string as a private label
prefix because then it treats all labels as private. This commit reverts
back to the default prefix, which is .L, which is common in ELF targets
and consistent with the LLVM name mangler.
llvm-svn: 257083
Summary:
Multi-dword constant loads generated unnecessary moves from SGPRs into VGPRs,
increasing the code size and VGPR pressure. These moves are now folded away.
Note that this lack of operand folding was not a problem for VMEM loads,
because COPY nodes from VReg_Nnn to VGPR32 are eliminated by the register
coalescer.
Some tests are updated, note that the fsub.ll test explicitly checks that
the move is elided.
With the IR generated by current Mesa, the changes are obviously relatively
minor:
7063 shaders in 3531 tests
Totals:
SGPRS: 351872 -> 352560 (0.20 %)
VGPRS: 199984 -> 200732 (0.37 %)
Code Size: 9876968 -> 9881112 (0.04 %) bytes
LDS: 91 -> 91 (0.00 %) blocks
Scratch: 1779712 -> 1767424 (-0.69 %) bytes per wave
Wait states: 295164 -> 295337 (0.06 %)
Totals from affected shaders:
SGPRS: 65784 -> 66472 (1.05 %)
VGPRS: 38064 -> 38812 (1.97 %)
Code Size: 1993828 -> 1997972 (0.21 %) bytes
LDS: 42 -> 42 (0.00 %) blocks
Scratch: 795648 -> 783360 (-1.54 %) bytes per wave
Wait states: 54026 -> 54199 (0.32 %)
Reviewers: tstellarAMD, arsenm, mareko
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15875
llvm-svn: 257074
Summary:
Somehow, I first interpreted the docs as saying space for xnack_mask is only
reserved when XNACK is enabled via SH_MEM_CONFIG. I felt uneasy about this and
went back to actually test what is happening, and it turns out that xnack_mask
is always reserved at least on Tonga and Carrizo, in the sense that flat_scr
is always fixed below the SGPRs that are used to implement xnack_mask, whether
or not they are actually used.
I confirmed this by writing a shader using inline assembly to tease out the
aliasing between flat_scratch and regular SGPRs. For example, on Tonga, where
we fix the number of SGPRs to 80, s[74:75] aliases flat_scratch (so
xnack_mask is s[76:77] and vcc is s[78:79]).
This patch changes both the calculation of the total number of SGPRs and the
various register reservations to account for this.
It ought to be possible to use the gap left by xnack_mask when the feature
isn't used, but this patch doesn't try to do that. (Note that the same applies
to vcc.)
Note that previously, even before my earlier change in r256794, the SGPRs that
alias to xnack_mask could end up being used as well when flat_scr was unused
and the total number of SGPRs happened to fall on the right alignment
(e.g. highest regular SGPR being used s29 and VCC used would lead to number
of SGPRs being 32, where s28 and s29 alias with xnack_mask). So if there
were some conflict due to such aliasing, we should have noticed that already.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15898
llvm-svn: 257073
AVX1 v8i32/v4i64 shuffles are bitcasted to v8f32/v4f64, this patch peeks through bitcasts to check for a load node to allow broadcasts to occur.
Follow up to D15310
llvm-svn: 257055
Follow up to D15378, added INSERTPS to the list of decodable target shuffles and enabled XFormVExtractWithShuffleIntoLoad to handle target shuffles with SentinelZero and tested this with INSERTPS.
llvm-svn: 257046
Darwin TLS accesses most closely resemble ELF's general-dynamic situation,
since they have to be able to handle all possible situations. The descriptors
and so on are obviously slightly different though.
llvm-svn: 257039
Serialize will perform a hardware serialization operation, and is
acting as a memory barrier. Therefore it must have the hasSideEffects
flag set so it will be treated as a global memory object.
Reviewed by Ulrich Weigand
llvm-svn: 257036
getTargetShuffleMask may return shuffle masks with SM_SentinelZero (-2) values (currently just for PSHUFB but VPERM2X128 as well with this patch). Although some calling functions can make use of this (mainly for shuffle combining), others can not and their inclusion makes shuffle mask comparisons more difficult.
This patch adds a flag to getTargetShuffleMask to indicate if the calling function can't handle SM_SentinelZero; getTargetShuffleMask will then return false if it occurs to make handling much easier.
I've tidied up some uses of getTargetShuffleMask to better indicate what is going on - more could be done but at present I don't have test cases to demonstrate it.
Some upcoming patches will make use of this to both support more uses where SM_SentinelZero is not permitted (e.g. combineShuffleToAddSub), and also will allow us to add INSERTPS support to getTargetShuffleMask as part of better zero handling discussed in D14261.
Differential Revision: http://reviews.llvm.org/D15378
llvm-svn: 256992
Summary:
This is admittedly something that you could only run into by manually
playing around with shader assembly because the SITypeWriter pass is
skipped for compute.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15902
llvm-svn: 256980
The first instruction in a block is what the rend() iterator points to, so
if it moves, we need to re-evaluate rend() so that we continue to iterate
through the rest of the instructions.
llvm-svn: 256953
As discussed on D15378, move the mask.empty() tests to after the switch statement and consider any shuffle decode where the extracted target shuffle mask is empty as a failure.
llvm-svn: 256921
In the discussion on http://reviews.llvm.org/D15730, Andy pointed out we had a utility function for merging MMO lists. Since it turned we actually had two copies and there's another review in progress (http://reviews.llvm.org/D15230) which needs the same, extract it into a utility function and clean up the interfaces to make it easier to use with a MachineInstBuilder.
I introduced a pair here to track size and allocation together. I think we should probably move in the direction of the MachineOperandsRef helper class, but I'm leaving that for further work. I want to get the poison state introduced before I make major changes to the interface.
Differential Revision: http://reviews.llvm.org/D15757
llvm-svn: 256909
Due to the SGPR init bug, every program claims to use the same number
of SGPRs anyway, so there's no point in trying to shift those registers
down from their initial spot of reservation.
Add a test that uses VGPR spilling and blocks most SGPRs from being used for
the scratch resource register. Previously, this would run into an assertion.
Differential Revision: http://reviews.llvm.org/D15724
llvm-svn: 256870
We queried hasFP before we hit ExpandISelPseudos. ExpandISelPseudos
manipulated state that hasFP relied on, potentially changing the result
after it has been queried elsewhere.
While I am not aware of any particular bug due to this state of affairs,
it seems best to avoid it entirely by changing the state during DAG
construction.
llvm-svn: 256849
PBLEND/BLENDPD/BLENDPS are no different to the other target shuffles and this will make future improvements to the target shuffle combines more straightforward.
llvm-svn: 256819
The red zone consists of 128 bytes beyond the stack pointer so that the
allocation of objects in leaf functions doesn't require decrementing
rsp. In r255656, we introduced an optimization that would cheaply
materialize certain constants via push/pop. Push decrements the stack
pointer and stores it's result at what is now the top of the stack.
However, this means that using push/pop would encroach on the red zone.
PR26023 gives an example where this corrupts an object in the red zone.
llvm-svn: 256808
Summary:
We had to sets of identical FLAT patterns one inside the
HasFlatAddressSpace predicate and one inside the useFlatForGloabl
predicate. This patch merges these sets into a single pattern
under the isCIVI predicate.
The reason we can remove the predicates is that when MUBUF instructions
are legal, the instruction selector will prefer selecting those over
FLAT instructions because MUBUF patterns have a higher complexity score.
So, in this case having patterns for FLAT instructions will have no effect.
This change also simplifies the process for forcing global address space
loads to use FLAT instructions, since we no only have to disable the
MUBUF patterns instead of having to disable the MUBUF patterns and
enable the FLAT patterns.
Reviewers: arsenm, cfang
Subscribers: llvm-commits
llvm-svn: 256807
Unfortunately this fix had the effect of exposing the
-verify-machineinstrs FIXME of X86InstrInfo.cpp in two testcases for
which I disabled it for now.
Two testcases also have additional pushq/popq where the corrected code
cannot prove that %rax is dead any longer. Looking at the examples, this
could potentially be fixed by improving computeRegisterLiveness() to check
the live-in lists of the successors blocks when reaching the end of a
block.
This fixes http://llvm.org/PR25951.
llvm-svn: 256799
Summary:
Enabling this feature will account for the two SGPRs used by the hardware
to store the XNACK_MASK physically.
The hardware only requires this reservation when the XNACK feature is
explicitly enabled. At some point, HSA will probably want to do that, but
it does increase SGPR register pressure, so leave it disabled by default
for now (but do add a small test).
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15869
llvm-svn: 256794
Summary: This was accidently moved to CIInstructions.td in r256282
Reviewers: cfang, arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15763
llvm-svn: 256775
Summary:
The comment explains it: emitError does not necessarily exit the compilation
process, and then using NoRegister leads to assertions later on.
This generates incorrect code, of course, but the user should know to not use
the result when an error has been emitted.
It would be nice to have a test-case for this inside the LLVM repository,
but llc exits on error. shader-db tests trigger the underlying issue at least
on Tonga.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15826
llvm-svn: 256757
We need a frame pointer if there is a push/pop sequence after the
prologue in order to unwind the stack. Scanning the instructions to
figure out if this happened made hasFP not constant-time which is a
violation of expectations. Let's compute this up-front and reuse that
computation when we need it.
llvm-svn: 256730
Summary:
There are a number of files in the tree which have been accidentally checked in with DOS line endings. Convert these to native line endings.
There are also a few files which have DOS line endings on purpose, and I have set the svn:eol-style property to 'CRLF' on those.
Reviewers: joerg, aaron.ballman
Subscribers: aaron.ballman, sanjoy, dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D15848
llvm-svn: 256707
LLVM's targets need to know if stack pointer adjustments occur after the
prologue. This is needed to correctly determine if the red-zone is
appropriate to use or if a frame pointer is required.
Normally, LLVM can figure this out very precisely by reasoning about the
contents of the MachineFunction. There is an interesting corner case:
inline assembly.
The vast majority of inline assembly which will perform a push or pop is
done so to pair up with pushf or popf as appropriate. Unfortunately,
this inline assembly doesn't mark the stack pointer as clobbered
because, well, it isn't. The stack pointer is decremented and then
immediately incremented. Because of this, LLVM was changed in r256456
to conservatively assume that inline assembly contain a sequence of
stack operations. This is unfortunate because the vast majority of
inline assembly will not end up manipulating the stack pointer in any
way at all.
Instead, let's provide a more principled solution: an intrinsic.
FWIW, other compilers (MSVC and GCC among them) also provide this
functionality as an intrinsic.
llvm-svn: 256685
Not folding these cases tends to avoid partial register updates:
sqrtss (%eax), %xmm0
Has a partial update of %xmm0, while
movss (%eax), %xmm0
sqrtss %xmm0, %xmm0
Has a clobber of the high lanes immediately before the partial update,
avoiding a potential stall.
Given this, we only want to fold when optimizing for size.
This is consistent with the patterns we already have for some of
the fp/int converts, and in X86InstrInfo::foldMemoryOperandImpl()
Differential Revision: http://reviews.llvm.org/D15741
llvm-svn: 256671
This restores the previous behavior of not including the mnemonic in the classes table for every target that starts instruction lines with the mnemonic. Not only did the table size increase by 1 entry, but the class enum increased in size which caused every class in the array to increase in size. It also grew the size of the function that parsers tokens into classes by a substantial amount.
This adds a new HasMnemonicFirst flag to all AsmParsers. It's set to 1 by default and Hexagon target overrides it to 0.
For the X86 target alone this recovers 324KB of size on the llvm-mc executable.
I believe the current state is still a bad design choice for the Hexagon target as it causes most of the parsing to do a linear search through the entire match table to comparing operands against every instruction until it finds one that works. At least for the other targets we do a binary search based on mnemonic over which to do the linear scan.
llvm-svn: 256669
Summary:
* avoid generating POP {LR} in Thumb1 epilogues
* combine MOV LR, Rx + BX LR -> BX Rx in a peephole optimization pass
* combine POP {LR} + B + BX LR -> POP {PC} on v5T+
Test cases by Ana Pazos
Differential Revision: http://reviews.llvm.org/D15707
llvm-svn: 256523
The cost is calculated for all X86 targets. When gather/scatter instruction
is not supported we calculate the cost of scalar sequence.
Differential revision: http://reviews.llvm.org/D15677
llvm-svn: 256519
This adds support for the MCU psABI in a way different from r251223 and r251224,
basically reverting most of these two patches. The problem with the approach
taken in r251223/4 is that it only handled libcalls that originated from the backend.
However, the mid-end also inserts quite a few libcalls and assumes these use the
platform's default calling convention.
The previous patch tried to insert inregs when necessary both in the FE and,
somewhat hackily, in the CG. Instead, we now define a new default calling convention
for the MCU, which doesn't use inreg marking at all, similarly to what x86-64 does.
Differential Revision: http://reviews.llvm.org/D15054
llvm-svn: 256494
lower broadcast<type>x<vector> to shuffles.
there are two cases:
1.src is 128 bits and dest is 512 bits: in this case we will lower it to shuffle with imm = 0.
2.src is 256 bit and dest is 512 bits: in this case we will lower it to shuffle with imm = 01000100b (0x44) that way we will broadcast the 256bit source: ymm[0,1,2,3] => zmm[0,1,2,3,0,1,2,3] then it will mask it with the passthru value (in case it's mask op).
Differential Revision: http://reviews.llvm.org/D15790
llvm-svn: 256490
I believe this also fixes a case where a 64-bit memory form that is documented as being unsupported in 32-bit mode was able to be selected there.
llvm-svn: 256483
Fix TRUNCATE lowering vector to vector i1, use LSB and not MSB.
Implement VPMOVB/W/D/Q2M intrinsic.
Differential Revision: http://reviews.llvm.org/D15675
llvm-svn: 256470
A frame pointer must be used if stack pointer is modified after the
prologue. LLVM will emit pushf/popf if we need to save/restore the
FLAGS register, requiring us to have a frame pointer for the function.
There is a small twist: this sequence might exist in user code via
inline-assembly. For now, conservatively assume that such functions
require a frame pointer. For real world justification, please see
clang's implementation of __readeflags.
This fixes PR25945.
llvm-svn: 256456
Move RegStackify after coalescing and teach it to use LiveIntervals instead
of depending on SSA form. This avoids a problem where a register in a COPY
instruction is stackified and then subsequently coalesced with a register
that is not stackified.
This also puts it after the scheduler, which allows us to simplify the
EXPR_STACK constraint, as we no longer have instructions being reordered
after stackification and before coloring.
llvm-svn: 256402
The patterns that set a mask register to 0/1
KXOR %kn, %kn, %kn / KXNOR %kn, %kn, %kn
are replaced with
KXOR %k0, %k0, %kn / KXNOR %k0, %k0, %kn - AVX-512 targets optimization.
KNL does not recognize dependency-breaking idioms for mask registers,
so kxnor %k1, %k1, %k2 has a RAW dependence on %k1.
Using %k0 as the undef input register is a performance heuristic based
on the assumption that %k0 is used less frequently than the other mask
registers, since it is not usable as a write mask.
Differential Revision: http://reviews.llvm.org/D15739
llvm-svn: 256365
First step towards making better use of AVX's implicit zeroing of the upper half of a 256-bit vector by instructions that only act on the lower 128-bit vector - discussed on D14151.
As well as the fact that 128-bit shuffle instructions are generally more capable, this can be performant for older CPUs with 128-bit ALUs (e.g. Jaguar, Sandy Bridge) that must treat 256-bit vectors as multiple micro-ops.
Moved the similar subvector extraction shuffle combines from PerformShuffleCombine256 to lowerVectorShuffle as well.
Note: I've avoided combining shuffles that reference elements from the upper halves of the input vectors - this may be reviewed in future work as well (AVX1 would probably always gain, but AVX2 does have some cross-lane shuffle instructions).
Differential Revision: http://reviews.llvm.org/D15477
llvm-svn: 256332
Summary:
For some reason doing executing an MUBUF instruction with the addr64
bit set and a zero base pointer in the resource descriptor causes
the memory operation to be dropped when the shader is executed using
the HSA runtime.
This kind of MUBUF instruction is commonly used when the pointer is
stored in VGPRs. The base pointer field in the resource descriptor
is set to zero and and the pointer is stored in the vaddr field.
This patch resolves the issue by only using flat instructions for
global memory operations when targeting HSA. This is an overly
conservative fix as all other configurations of MUBUF instructions
appear to work.
NOTE: re-commit by fixing a failure in Codegen/AMDGPU/llvm.dbg.value.ll
Reviewers: tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15543
llvm-svn: 256282
Summary:
For some reason doing executing an MUBUF instruction with the addr64
bit set and a zero base pointer in the resource descriptor causes
the memory operation to be dropped when the shader is executed using
the HSA runtime.
This kind of MUBUF instruction is commonly used when the pointer is
stored in VGPRs. The base pointer field in the resource descriptor
is set to zero and and the pointer is stored in the vaddr field.
This patch resolves the issue by only using flat instructions for
global memory operations when targeting HSA. This is an overly
conservative fix as all other configurations of MUBUF instructions
appear to work.
Reviewers: tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15543
llvm-svn: 256273
This patch removes all weight-related interfaces from BPI and replace
them by probability versions. With this patch, we won't use edge weight
anymore in either IR or MC passes. Edge probabilitiy is a better
representation in terms of CFG update and validation.
Differential revision: http://reviews.llvm.org/D15519
llvm-svn: 256263
This is a recommit of r256004 which was reverted in r256160. The issue was the
incorrect promotion for half and byte loads transformed into mov instructions.
This fix will replace half and byte type loads only with bit field extracts.
Original commit message:
This change promotes load instructions which directly read from stored by
replacing them with mov instructions. If the store is wider than the load,
the load will be replaced with a bitfield extract.
For example :
STRWui %W1, %X0, 1
%W0 = LDRHHui %X0, 3
becomes
STRWui %W1, %X0, 1
%W0 = UBFMWri %W1, 16, 31
llvm-svn: 256249
Today, we always take into account the possibility that object files
produced by MC may be consumed by an incremental linker. This results
in us initialing fields which vary with time (TimeDateStamp) which harms
hermetic builds (e.g. verifying a self-host went well) and produces
sub-optimal code because we cannot assume anything about the relative
position of functions within a section (call sites can get redirected
through incremental linker thunks).
Let's provide an MCTargetOption which controls this behavior so that we
can disable this functionality if we know a-priori that the build will
not rely on /incremental.
llvm-svn: 256203
This patch transforms truncation between vectors of integers into
X86ISD::PACKUS/PACKSS operations during DAG combine. We don't do it in
lowering phase because after type legalization, the original truncation
will be turned into a BUILD_VECTOR with each element that is extracted
from a vector and then truncated, and from them it is difficult to do
this optimization. This greatly improves the performance of truncations
on some specific types.
Cost table is updated accordingly.
Differential revision: http://reviews.llvm.org/D14588
llvm-svn: 256194
Summary:
These register has different encodings on CI and VI, so we add pseudo
FLAT_SCRACTH registers to be used before MC, and subtarget specific
registers to be used by the MC layer.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15661
llvm-svn: 256178
This patch adds to the target description two additional patterns for matching
extract-extend operations to SMOV. The patterns catch the v16i8-to-i64 and
v8i16-to-i64 cases. The existing patterns miss these cases because the
extracted elements must first be legalized to i32, resulting in any_extend
nodes.
This was originally implemented as a DAG combine (r255895), but was reverted
due to failing out-of-tree tests.
llvm-svn: 256176
This allows the AsmMatcherEmitter to properly tokenize the AsmStrings for
load and store instructions. This is a step towards asm parsing.
llvm-svn: 256166
Disable post-ra scheduler for perturbed tests to appease the bots and to
preserve the history of the tests.
http://reviews.llvm.org/D15652
llvm-svn: 256158
It resolves clang selfhosting with std::once() for Cygwin.
FIXME: It may be EmulatedTLS-generic also for X86-Android.
FIXME: Pass EmulatedTLS to LLVM CodeGen from Clang with -femulated-tls.
llvm-svn: 256134
This allows "icmp ugt %a, 4294967295" and "icmp uge %a, 4294967296" to be optimized into right shifts by 32 which can fold the immediate into the shift instruction. These patterns show up with some regularity in real code.
Unfortunately, since getImmCost can't see the icmp predicate we can't be tell if we're only catching these specific cases.
llvm-svn: 256126
Summary:
This adds the core AVR TableGen file, along with the register descriptions.
Lines in AVR.td which require other TableGen files which haven't been committed
yet are commented out.
This is a fairly trivial patch, and should only require a quick review.
I kept the line width smaller than 80 columns, but there are a few exceptions
because I'm not sure how to split a string over several lines.
Reviewers: stoklund
Subscribers: dylanmckay, agnat
Differential Revision: http://reviews.llvm.org/D14684
llvm-svn: 256120
Summary:
r250697 fixed the mapping for ARM mode. We have to do the same for Thumb2 otherwise the same llvm.arm.ssat() will generate different saturating amount for ARM and Thumb.
r250697: http://reviews.llvm.org/rL250697
Reviewers: rmaprath
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D15653
llvm-svn: 256115
Summary:
The analysis of shader inputs was completely wrong. We were passing the
wrong index to AttributeSet::hasAttribute() and the logic for which
inputs where in SGPRs was wrong too.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15608
llvm-svn: 256082
noduplicate prevents unrolling of small loops that happen to have
barriers in them. If a loop has a barrier in it, it is OK to duplicate
it for the unroll.
llvm-svn: 256075
Summary:
When copying aggregate registers within the same register class, there may
be an overlap between source and destination that forces us to do the copy
backwards.
Do the simplest possible thing that guarantees the correct order of moves
when there are overlaps, and does whatever when there is no overlap. (The
last part forces some trivial adjustments to test cases.)
Together with r255906, this fixes a VM fault in Unreal Elemental Demo.
While at it, change the generation of kill and def flags to something that
looks more reasonable. This method is used very late during compilation, so
it probably doesn't matter in practice, and to be honest, I don't know if
this change is actually correct because the semantics in connection with
aggregate registers vs. sub-registers are not clear to me.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93264
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15622
llvm-svn: 256072
This change promotes load instructions which directly read from stores by
replacing them with mov instructions. If the store is wider than the load,
the load will be replaced with a bitfield extract.
For example :
STRWui %W1, %X0, 1
%W0 = LDRHHui %X0, 3
becomes
STRWui %W1, %X0, 1
%W0 = UBFMWri %W1, 16, 31
llvm-svn: 256004
Use the 3-byte (4 with REX prefix) push-pop sequence for materializing
small constants. This is smaller than using a mov (5, 6 or 7 bytes
depending on size and REX prefix), but it's likely to be slower, so
only used for 'minsize'.
This is a follow-up to r255656.
Differential Revision: http://reviews.llvm.org/D15549
llvm-svn: 255936
Summary:
The method insertNOPs expected the number of wait states to be passed as
parameter, while eliminateFrameIndex passed the immediate argument for the
S_NOP, leading to an off-by-one error. Rename the method to make the
meaning of its parameter clearer. The number of 4 / 5 wait states (which
is what the method has always _tried_ to do according to the comment) is
correct according to the hardware docs.
I stumbled upon this while trying to track down the cause of
https://bugs.freedesktop.org/show_bug.cgi?id=93264. While clearly needed,
this patch unfortunately does not fix that bug...
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15542
llvm-svn: 255906
These days relocations are created and stored in a deterministic way.
The order they are created is also suitable for the .o file, so we don't
need an explicit sort.
The last remaining exception is MIPS.
llvm-svn: 255902
This patch enables PostRAScheduler specifically for AArch64 generic build,
which is beneficial from the performance perspective.
Speedups up to 2 to 7% for some benchmarks on A57 and A53 are observed.
Also benchmarks from LLVM test-suite did not regress.
Differential Revision: http://reviews.llvm.org/D15557
llvm-svn: 255896
This patch adds a DAG combine for (any_extend (extract_vector_elt v, i)) ->
(extract_vector_elt v, i). The combine enables us to better match some SMOV
patterns.
Differential Revision: http://reviews.llvm.org/D15515
llvm-svn: 255895
Add option to enable/disable LEA optimization pass. By default the pass is disabled.
Differential Revision: http://reviews.llvm.org/D15573
llvm-svn: 255881
This creates the initial infrastructure for writing ELF output files. It
doesn't yet have any implementation for encoding instructions.
Differential Revision: http://reviews.llvm.org/D15555
llvm-svn: 255869
Summary:
Implement eliminateCallFramePsuedo to handle ADJCALLSTACKUP/DOWN
pseudo-instructions. Add a test calling a vararg function which causes non-0
adjustments. This revealed an issue with RegisterCoalescer wherein it
eliminates a COPY from SP32 to a vreg but failes to update the live ranges
of EXPR_STACK, causing a machineinstr verifier failure (so this test
is commented out).
Also add a dynamic alloca test, which causes a callseq_end dag node with
a 0 (instead of undef) second argument to be generated. We currently fail to
select that, so adjust the ADJCALLSTACKUP tablegen code to handle it.
Differential Revision: http://reviews.llvm.org/D15587
llvm-svn: 255844
This matches the other MIB methods, none of which modify the builder.
Without this, we can't chain copyImplicitOps.
Also reformat the few users, in PPCEarlyReturn.
llvm-svn: 255828
The access function has a short entry and a short exit, the initialization
block is only run the first time. To improve the performance, we want to
have a short frame at the entry and exit.
We explicitly handle most of the CSRs via copies. Only the CSRs that are not
handled via copies will be in CSR_SaveList.
Frame lowering and prologue/epilogue insertion will generate a short frame
in the entry and exit according to CSR_SaveList. The majority of the CSRs will
be handled by register allcoator. Register allocator will try to spill and
reload them in the initialization block.
We add CSRsViaCopy, it will be explicitly handled during lowering.
1> we first set FunctionLoweringInfo->SplitCSR if conditions are met (the target
supports it for the given machine function and the function has only return
exits). We also call TLI->initializeSplitCSR to perform initialization.
2> we call TLI->insertCopiesSplitCSR to insert copies from CSRsViaCopy to
virtual registers at beginning of the entry block and copies from virtual
registers to CSRsViaCopy at beginning of the exit blocks.
3> we also need to make sure the explicit copies will not be eliminated.
The target independent portion was committed as r255353.
rdar://problem/23557469
Differential Revision: http://reviews.llvm.org/D15341
llvm-svn: 255821
The SystemZ linkers provide an optimization to transform a general-
or local-dynamic TLS sequence into an initial-exec sequence if possible.
Do do that, the compiler generates a function call to __tls_get_offset,
which is a brasl instruction annotated with *two* relocations:
- a R_390_PLT32DBL to install __tls_get_offset as branch target
- a R_390_TLS_GDCALL / R_390_TLS_LDCALL to inform the linker that
the TLS optimization should be performed if possible
If the optimization is performed, the brasl is replaced by an ld load
instruction.
However, *both* relocs are processed independently by the linker.
Therefore it is crucial that the R_390_PLT32DBL is processed *first*
(installing the branch target for the brasl) and the R_390_TLS_GDCALL
is processed *second* (replacing the whole brasl with an ld).
If the relocs are swapped, the linker will first replace the brasl
with an ld, and *then* install the __tls_get_offset branch target
offset. Since ld has a different layout than brasl, this may even
result in a completely different (or invalid) instruction; in any
case, the resulting code is corrupted.
Unfortunately, the way the MC common code sorts relocations causes
these two to *always* end up the wrong way around, resulting in
wrong code generation by the linker and crashes.
This patch overrides the sortRelocs routine to detect this particular
pair of relocs and enforce the required order.
llvm-svn: 255787
When comparing a zero-extended value against a constant small enough to
be in range of the inner type, it doesn't matter whether a signed or
unsigned compare operation (for the outer type) is being used. This is
why the code in adjustSubwordCmp had this assertion:
assert(C.ICmpType == SystemZICMP::Any &&
"Signedness shouldn't matter here.");
assuming the the caller had already detected that fact. However, it
turns out that there cases, in particular with always-true or always-
false conditions that have not been eliminated when compiling at -O0,
where this is not true.
Instead of failing an assertion if C.ICmpType is not SystemZICMP::Any
here, we can simply *set* it safely to SystemZICMP::Any, however.
llvm-svn: 255786
This removes an unpleasant hack involving a global variable for special
lowering of certain memcpy calls. These are now lowered as intended in
EmitTargetCodeForMemcpy in the same way that other targets do it.
llvm-svn: 255785
ARMv8.2-A adds 16-bit floating point versions of all existing SIMD
floating-point instructions. This is an optional extension, so all of
these instructions require the FeatureFullFP16 subtarget feature.
Note that VFP without SIMD is not a valid combination for any version of
ARMv8-A, but I have ensured that these instructions all depend on both
FeatureNEON and FeatureFullFP16 for consistency.
Differential Revision: http://reviews.llvm.org/D15039
llvm-svn: 255764
ARMv8.2-A adds 16-bit floating point versions of all existing VFP
floating-point instructions. This is an optional extension, so all of
these instructions require the FeatureFullFP16 subtarget feature.
The assembly for these instructions uses S registers (AArch32 does not
have H registers), but the instructions have ".f16" type specifiers
rather than ".f32" or ".f64". The top 16 bits of each source register
are ignored, and the top 16 bits of the destination register are set to
zero.
These instructions are mostly the same as the 32- and 64-bit versions,
but they use coprocessor 9 rather than 10 and 11.
Two new instructions, VMOVX and VINS, have been added to allow packing
and extracting two 16-bit floats stored in the top and bottom halves of
an S register.
New fixup kinds have been added for the PC-relative load and store
instructions, but no ELF relocations have been added as they have a
range of 512 bytes.
Differential Revision: http://reviews.llvm.org/D15038
llvm-svn: 255762
This folds (ashr (shl a, [56,48,32,24,16]), SarConst)
into (shl, (sext (a), [56,48,32,24,16] - SarConst))
or into (lshr, (sext (a), SarConst - [56,48,32,24,16]))
depending on sign of (SarConst - [56,48,32,24,16])
sexts in X86 are MOVs.
The MOVs have the same code size as above SHIFTs (only SHIFT by 1 has lower code size).
However the MOVs have 2 advantages to SHIFTs on x86:
1. MOVs can write to a register that differs from source.
2. MOVs accept memory operands.
This fixes PR24373.
Patch by: evgeny.v.stupachenko@intel.com
Differential Revision: http://reviews.llvm.org/D13161
llvm-svn: 255761
It adjusts from RSP-after-prologue to RBP, which is what SEH filters
need to do before they can use llvm.localrecover.
Fixes SEH filter captures, which were broken in r250088.
Issue reported by Alex Crichton.
llvm-svn: 255707
This patch improves on the suggested codegen from PR24475:
https://llvm.org/bugs/show_bug.cgi?id=24475
but only for the fmaxf() case to start, so we can sort out any bugs before
extending to fmin, f64, and vectors.
The fmax / maxnum definitions provide us flexibility for signed zeros, so the
only thing we have to worry about in this replacement sequence is NaN handling.
Note 1: It may be better to implement this as lowerFMAXNUM(), but that exposes
a problem: SelectionDAGBuilder::visitSelect() transforms compare/select
instructions into FMAXNUM nodes if we declare FMAXNUM legal or custom. Perhaps
that should be checking for NaN inputs or global unsafe-math before transforming?
As it stands, that bypasses a big set of optimizations that the x86 backend
already has in PerformSELECTCombine().
Note 2: The v2f32 test reveals another bug; the vector is extended to v4f32, so
we have completely unnecessary operations happening on undef elements of the
vector.
Differential Revision: http://reviews.llvm.org/D15294
llvm-svn: 255700
Summary: I'm not sure how things worked before without this.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15492
llvm-svn: 255692
Add instruction patterns for matching load and store instructions with constant
offsets in addresses. The code is fairly redundant due to the need to replicate
everything between imm, tglobaldadr, and texternalsym, but this appears to be
common tablegen practice. The main alternative appears to be to introduce
matching functions with C++ code, but sticking with purely generated matchers
seems better for now.
Also note that this doesn't yet support offsets from getelementptr, which will
be the most common case; that will depend on a change in target-independent code
in order to set the NoUnsignedWrap flag, which I'll submit separately. Until
then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add.
Also implement isLegalAddressingMode with an approximation of this.
Differential Revision: http://reviews.llvm.org/D15538
llvm-svn: 255681
Summary:
We were previously selecting all constant loads to SMRD instructions and legalizing
the SMRDs with non-uniform addresses during the SIFixSGPRCopesPass.
This new solution is more simple and also generates much better code, because
the instruction selector is able to take advantage of all the MUBUF addressing
modes that are legalization pass wasn't able to.
We also no longer need to generate v_add_* instructions when we
have a uniform pointer and a non-uniform offset, as this is now folded into the
MUBUF instruction during instruction selection.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15425
llvm-svn: 255672
A large number of loop utility functions take a `Pass *` and reach
into it to find out which analyses to preserve. There are a number of
problems with this:
- The APIs have access to pretty well any Pass state they want, so
it's hard to tell what they may or may not do.
- Other APIs have copied these and pass around a `Pass *` even though
they don't even use it. Some of these just hand a nullptr to the API
since the callers don't even have a pass available.
- Passes in the new pass manager don't work like the current ones, so
the APIs can't be used as is there.
Instead, we should explicitly thread the analysis results that we
actually care about through these APIs. This is both simpler and more
reusable.
llvm-svn: 255669
On SparcV8, doubles get passed in two 32-bit integer registers. The call
code was already handling endianness correctly, but the incoming
argument code was not -- it got the two halves in opposite order.
Also remove some dead code in LowerFormalArguments_32 to handle
less-than-32bit values, which can't actually happen.
Finally, add some test cases for the 32-bit calling convention, cribbed
from the 64abi.ll test, and run for both big and little-endian.
llvm-svn: 255668
We only want to emit CFI adjustments when actually using DWARF.
This fixes PR25828.
Differential Revision: http://reviews.llvm.org/D15522
llvm-svn: 255664
The radeonsi fp64 support can hit these now that some redundant bitcasts
are folded.
Patch by: Michel Dänzer
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 255657
"movl $-1, %eax" is 5 bytes, "xorl %eax, %eax; decl %eax" is 3 bytes.
This commit makes LLVM use the latter when optimizing for size.
Differential Revision: http://reviews.llvm.org/D14971
llvm-svn: 255656
Summary:
These are meant to be used instead of the llvm.SI.tid intrinsic which will
be deprecated at some point.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15475
llvm-svn: 255652
Summary:
These are meant to be used instead of the llvm.SI.fs.interp intrinsic which
will be deprecated at some point.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15474
llvm-svn: 255651
the feature flag is essential for RDPKRU and WRPKRU instruction
more about the instruction can be found in the SDM rev 56, vol 2 from http://www.intel.com/sdm
Differential Revision: http://reviews.llvm.org/D15491
llvm-svn: 255644
This patch corresponds to review:
http://reviews.llvm.org/D15117
In preparation for supporting IEEE Quad precision floating point,
this patch simply defines a feature to specify the target supports this.
For now, nothing is done with the target feature, we just don't want
warnings from the Clang FE when a user specifies -mfloat128.
Calling convention and other related work will add to this patch in
the near future.
llvm-svn: 255642
Full type legalizer that works with all vectors length - from 2 to 16, (i32, i64, float, double).
This intrinsic, for example
void @llvm.masked.scatter.v2f32(<2 x float>%data , <2 x float*>%ptrs , i32 align , <2 x i1>%mask )
requires type widening for data and type promotion for mask.
Differential Revision: http://reviews.llvm.org/D13633
llvm-svn: 255629
Prior to this patch, we would wrongly stick to the variant with imm8 encoding
even when the relocation could not fit that size.
rdar://problem/23785506
llvm-svn: 255583
Prior to this patch, we would wrongly stick to the variant with imm8 encoding
even when the relocation could not fit that size.
rdar://problem/23785506
llvm-svn: 255570
Add return type information to call and call_indirect instructions. This
allows them to be disambiguated without knowledge of the callee.
Differential Revision: http://reviews.llvm.org/D15484
llvm-svn: 255565
Implement a new BLOCK scope placement algorithm which better handles
early-return blocks and early exists from nested scopes.
Differential Revision: http://reviews.llvm.org/D15368
llvm-svn: 255564
Part 1 was submitted in http://reviews.llvm.org/D15134.
Changes in this part:
* X86RegisterInfo.td, X86RecognizableInstr.cpp: Add FR128 register class.
* X86CallingConv.td: Pass f128 values in XMM registers or on stack.
* X86InstrCompiler.td, X86InstrInfo.td, X86InstrSSE.td:
Add instruction selection patterns for f128.
* X86ISelLowering.cpp:
When target has MMX registers, configure MVT::f128 in FR128RegClass,
with TypeSoftenFloat action, and custom actions for some opcodes.
Add missed cases of MVT::f128 in places that handle f32, f64, or vector types.
Add TODO comment to support f128 type in inline assembly code.
* SelectionDAGBuilder.cpp:
Fix infinite loop when f128 type can have
VT == TLI.getTypeToTransformTo(Ctx, VT).
* Add unit tests for x86-64 fp128 type.
Differential Revision: http://reviews.llvm.org/D11438
llvm-svn: 255558
The WebAssemblyStoreResults pass runs before LiveVariables, so it doesn't
expect to have to keep dead flags up to date; check this with an assert.
llvm-svn: 255551
This will make the depedence graph more accurate if an alias analysis
is provided. If nullptr is specified in its place, the behavior will
remain as it is currently.
llvm-svn: 255540
This is the second in a set of patches for soft float support for ppc32,
it enables soft float operations.
Patch by Strahinja Petrovic.
Differential Revision: http://reviews.llvm.org/D13700
llvm-svn: 255516
The .even directive aligns content to an evan-numbered address.
In at&t syntax .even
In Microsoft syntax even (without the dot).
Differential Revision: http://reviews.llvm.org/D15413
llvm-svn: 255462
This patch adds some missing calls to MBB::normalizeSuccProbs() in several
locations where it should be called. Those places are found by checking if the
sum of successors' probabilities is approximate one in MachineBlockPlacement
pass with some instrumented code (not in this patch).
Differential revision: http://reviews.llvm.org/D15259
llvm-svn: 255455
EABI attributes should only be emitted on EABI targets. This prevents the
emission of the optimization goals EABI attribute on Windows ARM.
llvm-svn: 255448
While we have successfully implemented a funclet-oriented EH scheme on
top of LLVM IR, our scheme has some notable deficiencies:
- catchendpad and cleanupendpad are necessary in the current design
but they are difficult to explain to others, even to seasoned LLVM
experts.
- catchendpad and cleanupendpad are optimization barriers. They cannot
be split and force all potentially throwing call-sites to be invokes.
This has a noticable effect on the quality of our code generation.
- catchpad, while similar in some aspects to invoke, is fairly awkward.
It is unsplittable, starts a funclet, and has control flow to other
funclets.
- The nesting relationship between funclets is currently a property of
control flow edges. Because of this, we are forced to carefully
analyze the flow graph to see if there might potentially exist illegal
nesting among funclets. While we have logic to clone funclets when
they are illegally nested, it would be nicer if we had a
representation which forbade them upfront.
Let's clean this up a bit by doing the following:
- Instead, make catchpad more like cleanuppad and landingpad: no control
flow, just a bunch of simple operands; catchpad would be splittable.
- Introduce catchswitch, a control flow instruction designed to model
the constraints of funclet oriented EH.
- Make funclet scoping explicit by having funclet instructions consume
the token produced by the funclet which contains them.
- Remove catchendpad and cleanupendpad. Their presence can be inferred
implicitly using coloring information.
N.B. The state numbering code for the CLR has been updated but the
veracity of it's output cannot be spoken for. An expert should take a
look to make sure the results are reasonable.
Reviewers: rnk, JosephTremoulet, andrew.w.kaylor
Differential Revision: http://reviews.llvm.org/D15139
llvm-svn: 255422
We don't need to pass OutStreamer as a parameter to LowerSTACKMAP and
LowerPATCHPOINT. It is a member variable of PPCAsmPrinter, and thus, is already
available. NFC.
llvm-svn: 255418
Summary: This patch adds support of conversion (mul x, 2^N + 1) => (add (shl x, N), x) and (mul x, 2^N - 1) => (sub (shl x, N), x) if the multiplication can not be converted to LEA + SHL or LEA + LEA. LLVM has already supported this on ARM, and it should also be useful on X86. Note the patch currently only applies to cases where the constant operand is positive, and I am planing to add another patch to support negative cases after this.
Reviewers: craig.topper, RKSimon
Subscribers: aemerson, llvm-commits
Differential Revision: http://reviews.llvm.org/D14603
llvm-svn: 255415
This branch adds hints for highly biased branches on the PPC architecture. Even
in absence of profiling information, LLVM will mark code reaching unreachable
terminators and other exceptional control flow constructs as highly unlikely to
be reached.
Patch by Tom Jablin!
llvm-svn: 255398
Many tests are now passing due to eliminateFrameIndex implementation and
the list needs to be re-triaged because it unblocks other failures, and
some previous failures are different. However I'm about to churn it more
by implementing more lowering, so will wait on that.
llvm-svn: 255396
Summary:
Use the SP32 physical register as the base for FrameIndex
lowering. Update it and the __stack_pointer global var in the prolog and
epilog. Extend the mapping of virtual registers to wasm locals to
include the physical registers.
Rather than modify the target-independent PrologEpilogInserter (which
asserts that there are no virtual registers left) include a
slightly-modified copy for Wasm that does not have this assertion and
only clears the virtual registers if scavenging was needed (which of
course it isn't for wasm).
Differential Revision: http://reviews.llvm.org/D15344
llvm-svn: 255392
Summary: This patch adds support of conversion (mul x, 2^N + 1) => (add (shl x, N), x) and (mul x, 2^N - 1) => (sub (shl x, N), x) if the multiplication can not be converted to LEA + SHL or LEA + LEA. LLVM has already supported this on ARM, and it should also be useful on X86. Note the patch currently only applies to cases where the constant operand is positive, and I am planing to add another patch to support negative cases after this.
Reviewers: craig.topper, RKSimon
Subscribers: aemerson, llvm-commits
Differential Revision: http://reviews.llvm.org/D14603
llvm-svn: 255391
After much discussion, ending here:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151123/315620.html
it has been decided that, instead of having the vectorizer directly generate
special absdiff and horizontal-add intrinsics, we'll recognize the relevant
reduction patterns during CodeGen. Accordingly, these intrinsics are not needed
(the operations they represent can be pattern matched, as is already done in
some backends). Thus, we're backing these out in favor of the current
development work.
r248483 - Codegen: Fix llvm.*absdiff semantic.
r242546 - [ARM] Use [SU]ABSDIFF nodes instead of intrinsics for VABD/VABA
r242545 - [AArch64] Use [SU]ABSDIFF nodes instead of intrinsics for ABD/ABA
r242409 - [Codegen] Add intrinsics 'absdiff' and corresponding SDNodes for absolute difference operation
llvm-svn: 255387
computeRegisterLiveness() was broken in that it reported dead for a
register even if a subregister was alive. I assume this was because the
results of analayzePhysRegs() are hard to understand with respect to
subregisters.
This commit: Changes the results of analyzePhysRegs (=struct
PhysRegInfo) to be clearly understandable, also renames the fields to
avoid silent breakage of third-party code (and improve the grammar).
Fix all (two) users of computeRegisterLiveness() in llvm: By reenabling
it and removing workarounds for the bug.
This fixes http://llvm.org/PR24535 and http://llvm.org/PR25033
Differential Revision: http://reviews.llvm.org/D15320
llvm-svn: 255362
These are redundant pairs of nodes defined for
INSERT_VECTOR_ELEMENT/EXTRACT_VECTOR_ELEMENT.
insertelement/extractelement are slightly closer to the corresponding
C++ node name, and has stricter type checking so prefer it.
Update targets to only use these nodes where it is trivial to do so.
AArch64, ARM, and Mips all have various type errors on simple replacement,
so they will need work to fix.
Example from AArch64:
def : Pat<(sext_inreg (vector_extract (v16i8 V128:$Rn), VectorIndexB:$idx), i8),
(i32 (SMOVvi8to32 V128:$Rn, VectorIndexB:$idx))>;
Which is trying to do sext_inreg i8, i8.
llvm-svn: 255359
Summary:
ADJCALLSTACK{DOWN,UP} (aka CALLSEQ_{START,END}) MIs are supposed to use
and def the stack pointer. Since they do not, all the nodes are being
eliminated by DeadMachineInstructionElim, so they aren't in the IR when
PrologEpilogInserter/eliminateCallFramePseudo needs them.
This change fixes that, but since RegStackify will not stackify across
them (and it runs early, before PEI), change LowerCall to only emit them
when the call frame size is > 0. That makes the current code work the
same way and makes code handled by D15344 also work the same way. We can
expand the condition beyond NumBytes > 0 in the future if needed.
Reviewers: sunfish, jfb
Subscribers: jfb, dschuff, llvm-commits
Differential Revision: http://reviews.llvm.org/D15459
llvm-svn: 255356
Access to aligned globals gives us a chance to peephole optimize nonzero
offsets. If a struct is 4 byte aligned, then accesses to bytes 0-3 won't
overflow the available displacement. For example:
addis 3, 2, b4v@toc@ha
addi 4, 3, b4v@toc@l
lbz 5, b4v@toc@l(3) ; This is the result of the current peephole
lbz 6, 1(4) ; optimizer
lbz 7, 2(4)
lbz 8, 3(4)
If b4v is 4-byte aligned, we can skip using register 4 because we know
that b4v@toc@l+{1,2,3} won't overflow 32K, and instead generate:
addis 3, 2, b4v@toc@ha
lbz 4, b4v@toc@l(3)
lbz 5, b4v@toc@l+1(3)
lbz 6, b4v@toc@l+2(3)
lbz 7, b4v@toc@l+3(3)
Saving a register and an addition.
Larger alignments allow larger structures/arrays to be optimized.
llvm-svn: 255319
Previously in the conversion cost table there are no entries for integer-integer
conversions on SSE2. This will result in imprecise costs for certain vectorized
operations. This patch adds those entries for SSE2 and SSE4.1. The cost numbers
are counted from the result of running llc on the new test case in this patch.
Differential revision: http://reviews.llvm.org/D15132
llvm-svn: 255315
This was causing bad code gen and assembly that won't assemble, as
mixed altivec and vsx code would end up with a vsx high register
assigned to an altivec instruction, which won't work. Constraining the
classes allows the optimization to proceed.
llvm-svn: 255299
This patch corresponds to review:
http://reviews.llvm.org/D15286
LLVM IR frequently contains bitcast operations between floating point and
integer values of the same width. Doing this through memory operations is
quite expensive on PPC. This patch allows the use of direct register moves
between FPRs and GPRs for lowering bitcasts.
llvm-svn: 255246
SystemZ needs to do its scheduling after branch relaxation, which can
only happen after block placement, and therefore the standard
PostRAScheduler point in the pass sequence is too early.
TargetMachine::targetSchedulesPostRAScheduling() is a new method that
signals on returning true that target will insert the final scheduling
pass on its own.
Reviewed by Hal Finkel
llvm-svn: 255234
ISD::FCOPYSIGN permits its operands to have differing types, and DAGCombiner
uses this. Add some def : Pat rules to expand this out into an explicit
conversion and a normal copysign operation.
llvm-svn: 255220
Summary:
This allows us to remove the END_OF_TEXT_LABEL hack we had been using
and simplifies the fixups used to compute the address of constant
arrays.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15257
llvm-svn: 255204
without a frame pointer when unwind may happen.
This is a workaround for a bug in the way we emit the CFI directives for
frameless unwind information. See PR25614.
llvm-svn: 255175
Reinteroduce the code for moving ARGUMENTS back to the top of the basic block.
While the ARGUMENTS physical register prevents sinking and scheduling from
moving them, it does not appear to be sufficient to prevent SelectionDAG from
moving them down in the initial schedule. This patch introduces a patch that
moves them back to the top immediately after SelectionDAG runs.
This is still hopefully a temporary solution. http://reviews.llvm.org/D14750 is
one alternative, though the review has not been favorable, and proposed
alternatives are longer-term and have other downsides.
This fixes the main outstanding -verify-machineinstrs failures, so it adds
-verify-machineinstrs to several tests.
Differential Revision: http://reviews.llvm.org/D15377
llvm-svn: 255125
We mutated the DAG, which invalidated the node we were trying to use
as a base register. Sometimes we got away with it, but other times the
node really did get deleted before it was finished with.
Should fix PR25733
llvm-svn: 255120
The bots are now running the torture tests properly. Bin all failures from the GCC C torture tests so that we can tackle failures and make the tree go red on regressions.
llvm-svn: 255111
Summary:
Although the multiclass for i32 selects might seem redundant as it has
only one instantiation, we will use it to replace the correspondent
patterns in Mips64r6InstrInfo.td in follow-up commits.
Reviewers: dsanders
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D14612
llvm-svn: 255110
Commited patch was intended to implement LH, LHE, LHU and LHUE instructions.
After commit test-suite failed with error message in the form of:
fatal error: error in backend: Cannot select: t124: i32,ch = load<LD2[%d](tbaa=<0x94acc48>), sext from i16> t0, t2, undef:i32
For that reason I decided to revert commit r254897 and make new patch which besides implementation and standard regression tests will also have dedicated tests (CodeGen) for the above error.
llvm-svn: 255109
Otherwise, we think that most types that look like they'd fit in a
legal vector type are legal (so, basically, *any* vector type with a
size between 33 and 128 bits, I think, since we use pow2 alignment;
e.g., v2i25, v3f32, ...).
DataLayout::getTypeAllocSize rounds up based on alignment.
When checking for target intrinsic legality, that's not what we want:
if rounding makes a difference, the type isn't legal, and the
target intrinsics shouldn't be used, as they are always assumed legal.
One could make the argument that alloc size is ultimately the most
relevant here, since we're dealing with LD/ST intrinsics. That's only
true if we did legalize them though; that's a problem for another day.
Use DataLayout::getTypeSizeInBits instead of getTypeAllocSizeInBits.
Type::getSizeInBits can't be used because that'd gratuitously break
pointer vector support.
Some of these uses are currently fine, because we only hit them when
the type is already known legal (e.g., r114454). Update them for
consistency. It's faster to avoid the rounding anyway!
llvm-svn: 255089
Summary:
This fixes failure when trying to select
insertelement <4 x half> undef, half %a, i64 0
which gets transformed to a scalar_to_vector node.
The accompanying v4 and v8 tests fail instruction selection without this
patch.
Reviewers: ab, jmolloy
Subscribers: srhines, llvm-commits
Differential Revision: http://reviews.llvm.org/D15322
llvm-svn: 255072
On AVX and AVX2, BROADCAST instructions can load a scalar into all elements of a target vector.
This patch improves the lowering of 'splat' shuffles of a loaded vector into a broadcast - currently the lowering only works for cases where we are splatting the zero'th element, which is now generalised to any element.
Fix for PR23022
Differential Revision: http://reviews.llvm.org/D15310
llvm-svn: 255061
Summary:
Before ARMv5T, Thumb1 code could not pop PC, as described at D14357 and D14986;
so we need the special fixup in the epilogue.
Reviewers: jroelofs, qcolombet
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D15126
llvm-svn: 255047
AND/BIC instructions do accept SP/PC, so the register class should be
more generic (rGPR -> GPR) to cope with that case. Adding more tests.
llvm-svn: 255034
Summary:
We don't check the size operand on ext/dext*/ins/dins* yet because the
permitted range depends on the pos argument and we can't check that using
this mechanism.
The bug was that dextu/dinsu accepted 0..31 in the pos operand instead of 32..63.
Reviewers: vkalintiris
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D15190
llvm-svn: 255015
ARMv8.2-A adds 16-bit floating point versions of all existing SIMD
floating-point instructions. This is an optional extension, so all of
these instructions require the FeatureFullFP16 subtarget feature.
Note that VFP without SIMD is not a valid combination for any version of
ARMv8-A, but I have ensured that these instructions all depend on both
FeatureNEON and FeatureFullFP16 for consistency.
The ".2h" vector type specifier is now legal (for the scalar pairwise
reduction instructions), so some unrelated tests have been modified as
different error messages are emitted. This is not a problem as the
invalid operands are still caught.
llvm-svn: 255010
FP logic instructions are supported in DQ extension on AVX-512 target.
I use integer operations instead.
Added tests.
I also enabled FABS in this patch in order to check ANDPS.
The operations are FOR, FXOR, FAND, FANDN.
The instructions, that supported for 512-bit vector under DQ are:
VORPS/PD, VXORPS/PD, VANDPS/PD, FANDNPS/PD.
Differential Revision: http://reviews.llvm.org/D15110
llvm-svn: 254913
Summary: This reverts r254234, and adds a simple fix for the annoying case of use-after-free.
Reviewers: rengolin
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D15236
llvm-svn: 254912
Patterns were missing for KNL target for <8 x i32>, <8 x float> masked load/store.
This intrinsic comes with all legal types:
<8 x float> @llvm.masked.load.v8f32(<8 x float>* %addr, i32 align, <8 x i1> %mask, <8 x float> %passThru),
but still requires lowering, because VMASKMOVPS, VMASKMOVDQU32 work with 512-bit vectors only.
All data operands should be widened to 512-bit vector.
The mask operand should be widened to v16i1 with zeroes.
Differential Revision: http://reviews.llvm.org/D15265
llvm-svn: 254909
According to x86 spec, loopz and loopnz should be supported for Intel syntax, where loopz is equivalent to loope and loopnz is equivalent to loopne.
Differential Revision: http://reviews.llvm.org/D15148
llvm-svn: 254877
This removes the code path that generate "synchronous" (only correct at call site) CFA.
We will probably want to re-introduce it once we are capable of emitting different
.eh_frame and .debug_frame sections.
Differential Revision: http://reviews.llvm.org/D14948
llvm-svn: 254874
This patch introduces a codegen-only instruction currently named br_unless,
which makes it convenient to implement ReverseBranchCondition and re-enable
the MachineBlockPlacement pass. Then in a late pass, it lowers br_unless
back into br_if.
Differential Revision: http://reviews.llvm.org/D14995
llvm-svn: 254826
Add physical register defs to instructions used from stackified
instructions to prevent them from being scheduled into the middle of
a stack sequence. This is a conservative measure which may be loosened
in the future.
Differential Revision: http://reviews.llvm.org/D15252
llvm-svn: 254811
This is just prototype for load/store for i32 types. I'll add them to
the rest of the types if we like this direction.
Differential Revision: http://reviews.llvm.org/D15197
llvm-svn: 254807
When the notion of target specific memory intrinsics was introduced to EarlyCSE, the commit confused the notions of volatile and simple memory access. Since I'm about to start working on this area, cleanup the naming so that patches aren't horribly confusing. Note that the actual implementation was always bailing if the load or store wasn't simple.
Reminder:
- "volatile" - C++ volatile, can't remove any memory operations, but in principal unordered
- "ordered" - imposes ordering constraints on other nearby memory operations
- "atomic" - can't be split or sheared. In LLVM terms, all "ordered" operations are also atomic so the predicate "isAtomic" is often used.
- "simple" - a load which is none of the above. These are normal loads and what most of the optimizer works with.
llvm-svn: 254805
Full varargs support will depend on prologue/epilogue support, but this patch
gets us started with most of the basic infrastructure.
Differential Revision: http://reviews.llvm.org/D15231
llvm-svn: 254799
These instructions are not supported by all CPUs in 64-bit mode. Emitting them
causes Chromium to crash on start-up for users with such chips.
(GCC puts these instructions behind -msahf on 64-bit for the same reason.)
This patch adds FeatureLAHFSAHF, enables it by default for 32-bit targets
and modern CPUs, and changes X86InstrInfo::copyPhysReg back to the lowering
from before r244503 when the instructions are not available.
Differential Revision: http://reviews.llvm.org/D15240
llvm-svn: 254793
This commit adds a new target-independent calling convention for C++ TLS
access functions. It aims to minimize overhead in the caller by perserving as
many registers as possible.
The target-specific implementation for X86-64 is defined as following:
Arguments are passed as for the default C calling convention
The same applies for the return value(s)
The callee preserves all GPRs - except RAX and RDI
The access function makes C-style TLS function calls in the entry and exit
block, C-style TLS functions save a lot more registers than normal calls.
The added calling convention ties into the existing implementation of the
C-style TLS functions, so we can't simply use existing calling conventions
such as preserve_mostcc.
rdar://9001553
llvm-svn: 254737
Since BuildMI() automatically adds the implicit operands for a new instruction,
adding the old instructions CC operand resulted in that there were two CC imp-def
operands, where only one was marked as dead. This caused buildSchedGraph() to
miss dependencies on the CC reg.
Review by Ulrich Weigand
llvm-svn: 254714
Add new x86 pass which replaces address calculations in load or store instructions with def register of existing LEA (must be in the same basic block), if the LEA calculates address that differs only by a displacement. Works only with -Os or -Oz.
Differential Revision: http://reviews.llvm.org/D13294
llvm-svn: 254712
with its source instead of forcing the values on GPRs.
This improves the lowering of vector code when such bitcasts happen in the
middle of vector computations.
rdar://problem/23691584
llvm-svn: 254684
Summary:
computeRegisterLiveness and analyzePhysReg are currently getting
confused about liveness in some cases, breaking copyPhysReg's
calculation of whether AX is dead in some cases. Work around this issue
temporarily by assuming that AX is always live.
See detail in: https://llvm.org/bugs/show_bug.cgi?id=25033#c7
And associated bugs PR24535 PR25033 PR24991 PR24992 PR25201.
This workaround makes the code correct but slightly inefficient, but it
seems to confuse the machine instr verifier which now things EAX was
undefined in some cases where it's being conservatively saved /
restored.
Reviewers: majnemer, sanjoy
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15198
llvm-svn: 254680
When a block has no terminator instructions, getFirstTerminator() returns
end(), which can't be used in dominance checks. Check dominance for phi
operands separately.
Also, remove some bits from WebAssemblyRegStackify.cpp that were causing
trouble on the same testcase; they were left behind from an earlier
experiment.
Differential Revision: http://reviews.llvm.org/D15210
llvm-svn: 254662
Almost all these changes are conditioned and only apply to the new
x86-64 f128 type configuration, which will be enabled in a follow up
patch. They are required together to make new f128 work. If there is
any error, we should fix or revert them as a whole.
These changes should have no impact to current configurations.
* Relax type legalization checks to accept new f128 type configuration,
whose TypeAction is TypeSoftenFloat, not TypeLegal, but also has
TLI.isTypeLegal true.
* Relax GetSoftenedFloat to return in some cases f128 type SDValue,
which is TLI.isTypeLegal but not "softened" to i128 node.
* Allow customized FABS, FNEG, FCOPYSIGN on new f128 type configuration,
to generate optimized bitwise operators for libm functions.
* Enhance related Lower* functions to handle f128 type.
* Enhance DAGTypeLegalizer::run, SoftenFloatResult, and related functions
to keep new f128 type in register, and convert f128 operators to library calls.
* Fix Combiner, Emitter, Legalizer routines that did not handle f128 type.
* Add ExpandConstant to handle i128 constants, ExpandNode
to handle ISD::Constant node.
* Add one more parameter to getCommonSubClass and firstCommonClass,
to guarantee that returned common sub class will contain the specified
simple value type.
This extra parameter is used by EmitCopyFromReg in InstrEmitter.cpp.
* Fix infinite loop in getTypeLegalizationCost when f128 is the value type.
* Fix printOperand to handle null operand.
* Enhance ISD::BITCAST node to handle f128 constant.
* Expand new f128 type for BR_CC, SELECT_CC, SELECT, SETCC nodes.
* Enhance X86AsmPrinter to emit f128 values in comments.
Differential Revision: http://reviews.llvm.org/D15134
llvm-svn: 254653
Summary:
These ADJCALLSTACK markers don't generate code, but they keep dynamic
alloca code that calls chkstk out of the prologue.
This slightly pessimizes inalloca calls by preventing some register copy
coalescing, but I can live with that.
Reviewers: qcolombet
Subscribers: hans, llvm-commits
Differential Revision: http://reviews.llvm.org/D15200
llvm-svn: 254645
In the case of a conditional branch without a preceding cmp we used to emit
a "and; cmp; b.eq/b.ne" sequence, use tbz/tbnz instead.
Differential Revision: http://reviews.llvm.org/D15122
llvm-svn: 254621
Currently "<type> ptr <reg name>" treated as <reg name> in MS inline asm, ignoring the "<type> ptr" completely and possibly ignoring the intention of the user.
Fixed llvm to produce an error when encountering "<type> ptr <reg name>" operands.
For example: andpd xmm1,xmmword ptr xmm1 --> andpd xmm1, xmm1
though andpd has 2 possible matching formats - andpd xmm, xmm/m128
Patch by: ziv.izhar@intel.com
Differential Revision: http://reviews.llvm.org/D14607
llvm-svn: 254607
Summary: This is done only when targeting HSA.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D13807
llvm-svn: 254587
This call should in fact be made by RegScavenger::enterBasicBlock()
called below. The first call does nothing except for triggering UB,
indicated by UBSan (passing nullptr to memset()).
llvm-svn: 254548
std::hex is not used anywhere in LLVM code base except for this place,
and it has a known undefined behavior (at least in libstdc++ 4.9.3):
https://llvm.org/bugs/show_bug.cgi?id=18156, which fires in UBSan
bootstrap of LLVM.
llvm-svn: 254547
The ARM ARM is clear that 128-bit loads are only guaranteed to have been atomic
if there has been a corresponding successful stxp. It's less clear for AArch32, so
I'm leaving that alone for now.
llvm-svn: 254524
Summary: Only global or readonly segment variables should appear in object files.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15111
llvm-svn: 254519
|9B DD /7| FSTSW m2byte| Valid Valid Store FPU status word at m2byteafter checking for pending unmasked floating-point exceptions.|
|9B DF E0| FSTSW AX| Valid Valid Store FPU status word in AX register after checking for pending unmasked floating-point exceptions.|
|DD /7 |FNSTSW *m2byte| Valid Valid Store FPU status word at m2bytewithout checking for pending unmasked floating-point exceptions.|
|DF E0 |FNSTSW *AX| Valid Valid Store FPU status word in AX register without checking for pending unmasked floating-point exceptions|
m2byte is word register, and therefor instruction operand need to be change from f32mem to i16mem.
Differential Revision: http://reviews.llvm.org/D14953
llvm-svn: 254512
On FMA targets, we can avoid having to load a constant to negate a float/double multiply by instead using a FNMSUB (-(X*Y)-0)
Fix for PR24366
Differential Revision: http://reviews.llvm.org/D14909
llvm-svn: 254495
I checked and updated the cost of AVX-512 conversion operations. Added cost of conversion operations in DQ mode.
Conversion of illegal types that requires vector split is not calculated right now (like for other X86 targets).
Differential Revision: http://reviews.llvm.org/D15074
llvm-svn: 254494
We mustn't introduce a shift of exactly 64-bits for any inputs, since that's an
UNDEF value (and worse, it's not what you want with the natural Arch64
implementation).
The generated code is pretty horrific, but I couldn't come up with an obviously
better alternative (if the amount is constant EXTR could help). Turns out
128-bit shifts are just nasty.
rdar://22491037
llvm-svn: 254475
The values in this field are compared against getAvailableFeatures()
which returns an uint64_t. This was causing problems in an internal
branch.
llvm-svn: 254462
Don't use commuteInstruction, and don't commute if
doing so will not improve legality. Skip the more
complex checks for literal operands and constant bus restrictions,
which are not a concern for VOP2 instructions because src1
does not accept SGPRs or constants and few implicitly
read vcc.
This gets called quite a few times and the
attempts at commuting are a significant fraction
of the time spent in SIFixSGPRCopies, so it's
somewhat worthwhile to optimize. With this patch and others
leading up to it, this reduces the compile time of SIFixSGPRCopies
on some of the LuxMark 2 kernels from ~8ms to ~5ms on my system.
llvm-svn: 254452
Summary:
This had been broken for a very long time, but nobody noticed until
D14357 enabled shrink-wrapping by default.
Reviewers: jroelofs, qcolombet
Subscribers: tyomitch, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D14986
llvm-svn: 254444
Summary:
When not useful bits, BitWidth becomes 0 and APInt will not be happy.
See https://llvm.org/bugs/show_bug.cgi?id=25571
We can just mark the operand as IMPLICIT_DEF is none bits of it is used.
Reviewers: t.p.northover, jmolloy
Subscribers: gberry, jmolloy, mgrang, aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D14803
llvm-svn: 254440
The cost for scalarized operations is computed as N * (scalar operation
cost + 1 extractelement + 1 insertelement). This partially fixes
inflating the cost of scalarized operations since every operation is
scalarized and free. I don't think we want any cost asociated with
scalarization, but for now insertelement is still counted. I'm not sure
if we should pretend that insertelement is also free, or add a way
to compute a custom scalarization cost.
llvm-svn: 254438
Summary:
This makes the assembly output look nicer and there is no reason to
have custom strings for these.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D14671
llvm-svn: 254426
The @llvm.get.dynamic.area.offset.* intrinsic family is used to get the offset
from native stack pointer to the address of the most recent dynamic alloca on
the caller's stack. These intrinsics are intendend for use in combination with
@llvm.stacksave and @llvm.restore to get a pointer to the most recent dynamic
alloca. This is useful, for example, for AddressSanitizer's stack unpoisoning
routines.
Patch by Max Ostapenko.
Differential Revision: http://reviews.llvm.org/D14983
llvm-svn: 254404
The Statistical Profiling Extension is an optional extension to
ARMv8.2-A. Since it is an optional extension, I have added the
FeatureSPE subtarget feature to control it. The assembler-visible parts
of this extension are the new "psb csync" instruction, which is
equivalent to "hint #17", and a number of system registers.
Differential Revision: http://reviews.llvm.org/D15021
llvm-svn: 254401
Add ARMv8.2-A to TargetParser, so that it can be used by the clang
command-line options and the .arch directive.
Most testing of this will be done in clang, checking that the
command-line options that this enables work.
Differential Revision: http://reviews.llvm.org/D15037
llvm-svn: 254400
This adds subtarget features for ARMv8.2-A, which builds on (and
requires the features from) ARMv8.1-A. Most assembler-visible features
of ARMv8.2-A are system instructions, and are all required parts of the
architecture, so just depend on the HasV8_2aOps subtarget feature.
There is also one large, optional feature, which adds 16-bit floating
point versions of all existing floating-point instructions (VFP and
SIMD), this is represented by the FeatureFullFP16 subtarget feature.
Differential Revision: http://reviews.llvm.org/D15036
llvm-svn: 254399
Not sure how to test this. I noticed by inspection in the isel tables where the same pattern tried to produce DIV and DIVR or SUB and SUBR.
llvm-svn: 254388
(This is the second attempt to submit this patch. The first caused two assertion
failures and was reverted. See https://llvm.org/bugs/show_bug.cgi?id=25687)
The patch in http://reviews.llvm.org/D13745 is broken into four parts:
1. New interfaces without functional changes (http://reviews.llvm.org/D13908).
2. Use new interfaces in SelectionDAG, while in other passes treat probabilities
as weights (http://reviews.llvm.org/D14361).
3. Use new interfaces in all other passes.
4. Remove old interfaces.
This patch is 3+4 above. In this patch, MBB won't provide weight-based
interfaces any more, which are totally replaced by probability-based ones.
The interface addSuccessor() is redesigned so that the default probability is
unknown. We allow unknown probabilities but don't allow using it together
with known probabilities in successor list. That is to say, we either have a
list of successors with all known probabilities, or all unknown
probabilities. In the latter case, we assume each successor has 1/N
probability where N is the number of successors. An assertion checks if the
user is attempting to add a successor with the disallowed mixed use as stated
above. This can help us catch many misuses.
All uses of weight-based interfaces are now updated to use probability-based
ones.
Differential revision: http://reviews.llvm.org/D14973
llvm-svn: 254377
and the follow-up r254356: "Fix a bug in MachineBlockPlacement that may cause assertion failure during BranchProbability construction."
Asserts were firing in Chromium builds. See PR25687.
llvm-svn: 254366
The patch in http://reviews.llvm.org/D13745 is broken into four parts:
1. New interfaces without functional changes (http://reviews.llvm.org/D13908).
2. Use new interfaces in SelectionDAG, while in other passes treat probabilities
as weights (http://reviews.llvm.org/D14361).
3. Use new interfaces in all other passes.
4. Remove old interfaces.
This patch is 3+4 above. In this patch, MBB won't provide weight-based
interfaces any more, which are totally replaced by probability-based ones.
The interface addSuccessor() is redesigned so that the default probability is
unknown. We allow unknown probabilities but don't allow using it together
with known probabilities in successor list. That is to say, we either have a
list of successors with all known probabilities, or all unknown
probabilities. In the latter case, we assume each successor has 1/N
probability where N is the number of successors. An assertion checks if the
user is attempting to add a successor with the disallowed mixed use as stated
above. This can help us catch many misuses.
All uses of weight-based interfaces are now updated to use probability-based
ones.
Differential revision: http://reviews.llvm.org/D14973
llvm-svn: 254348
We currently output FMA instructions on targets which support both FMA4 + FMA (i.e. later Bulldozer CPUS bdver2/bdver3/bdver4).
This patch flips this so FMA4 is preferred; this is for several reasons:
1 - FMA4 is non-destructive reducing the need for mov instructions.
2 - Its more straighforward to commute and fold inputs (although the recent work on FMA has reduced this difference).
3 - All supported targets have FMA4 performance equal or better to FMA - Piledriver (bdver2) in particular has half the throughput when executing FMA instructions.
Its looks like no future AMD processor lines will support FMA4 after the Bulldozer series so we're not causing problems for later CPUs.
Differential Revision: http://reviews.llvm.org/D14997
llvm-svn: 254339
If we know we have stack objects, we reserve the registers
that the private buffer resource and wave offset are passed
and use them directly.
If not, reserve the last 5 SGPRs just in case we need to spill.
After register allocation, try to pick the next available registers
instead of the last SGPRs, and then insert copies from the inputs
to the reserved registers in the progloue.
This also only selectively enables all of the input registers
which are really required instead of always enabling them.
llvm-svn: 254331
It does not work because of emergency stack slots.
This pass was supposed to eliminate dummy registers for the
spill instructions, but the register scavenger can introduce
more during PrologEpilogInserter, so some would end up
left behind if they were needed.
The potential for spilling the scratch resource descriptor
and offset register makes doing something like this
overly complicated. Reserve registers to use for the resource
descriptor and use them directly in eliminateFrameIndex.
Also removes creating another scratch resource descriptor
when directly selecting scratch MUBUF instructions.
The choice of which registers are reserved is temporary.
For now it attempts to pick the next available registers
after the user and system SGPRs.
llvm-svn: 254329
The MachineVerifier wants to check that the register operands of an
instruction belong to the instruction's register class. RIP-relative
control flow instructions violated this by referencing RIP. While this
was fixed for SysV, it was never fixed for Win64.
llvm-svn: 254315
Re-enable shrink wrapping for PPC64 Little Endian.
One minor modification to PPCFrameLowering::findScratchRegister was necessary to handle fall-thru blocks (blocks with no terminator) correctly.
Tested with all LLVM test, clang tests, and the self-hosting build, with no problems found.
PHabricator: http://reviews.llvm.org/D14778
llvm-svn: 254314
Value of offset operand for microMIPS BALC and BC instructions is currently shifted 2 bits, but it should be 1 bit.
Differential Revision: http://reviews.llvm.org/D14770
llvm-svn: 254296
We could already recognise shuffle(FSUB, FADD) -> ADDSUB, this allow us to recognise shuffle(FADD, FSUB) -> ADDSUB by commuting the shuffle mask prior to matching.
llvm-svn: 254259
This patch implements dynamic realignment of stack objects for targets
with a non-realigned stack pointer. Behaviour in FunctionLoweringInfo
is changed so that for a target that has StackRealignable set to
false, over-aligned static allocas are considered to be variable-sized
objects and are handled with DYNAMIC_STACKALLOC nodes.
It would be good to group aligned allocas into a single big alloca as
an optimization, but this is yet todo.
SystemZ benefits from this, due to its stack frame layout.
New tests SystemZ/alloca-03.ll for aligned allocas, and
SystemZ/alloca-04.ll for "no-realign-stack" attribute on functions.
Review and help from Ulrich Weigand and Hal Finkel.
llvm-svn: 254227
Summary:
Since this build attribute corresponds to a whole module, and
different functions in a module may differ in the optimizations
enabled for them, this attribute is emitted after all functions,
and only in the case that the optimization goals for all
functions match.
Reviewers: logan, hans
Subscribers: aemerson, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D14934
llvm-svn: 254201
ARMv8.2-A adds 16-bit floating point versions of all existing VFP
floating-point instructions. This is an optional extension, so all of
these instructions require the FeatureFullFP16 subtarget feature.
Most of these instructions are the same as the 32- and 64-bit versions,
but with the type field (bits 23-22) set to 0b11. Previously the top bit
of the size field was always 0, so the instruction classes only provided
a 1-bit size field, which I have widened to 2 bits.
Differential Revision: http://reviews.llvm.org/D15014
llvm-svn: 254198
Summary:
The bugs were:
* append, prepend, and balign were not tested
* balign takes a uimm2 not a uimm5.
* drotr32 was correctly implemented with a uimm5 but the tests expected
'52' to be valid.
* li/la were implemented with a uimm5 instead of simm32. simm32 isn't
completely correct either but I'll fix that when I get to simm32.
A notable omission are some of the shift instructions. Several of these
have been implemented using a single uimm6 instruction (rather than two
uimm5 instructions and a CodeGen-only uimm6 pseudo). These will be updated
in the uimm6 patch.
Reviewers: vkalintiris
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D14712
llvm-svn: 254164
ARMv8.2-A adds new variants of the "at" (address translate) system
instruction, which take the PSTATE.PAN bit (added in ARMv8.1-A). These
are a required part of ARMv8.2-A, so no additional subtarget features
are required.
Differential Revision: http://reviews.llvm.org/D15018
llvm-svn: 254159
Building on r253865 the crash is not limited to signed overflows.
Disable custom handling of unsigned 32-bit and 64-bit integer divide.
Add test cases for both 32-bit and 64-bit unsigned integer overflow.
llvm-svn: 254158
ARMv8.2-A adds a new PSTATE bit, PSTATE.UAO, which allows the LDTR/STTR
instructions to behave the same as LDR/STR with respect to execute-only
pages at higher privilege levels. New variants of the MSR/MRS
instructions are added to allow reading and writing this bit. It is a
required part of ARMv8.2-A, so no additional subtarget features are
required.
Differential Revision: http://reviews.llvm.org/D15020
llvm-svn: 254157
ARMv8.2-A adds the "dc cvap" instruction, which is a system instruction
that cleans caches to the point of persistence (for systems that have
persistent memory). It is a required part of ARMv8.2-A, so no additional
subtarget features are required.
Differential Revision: http://reviews.llvm.org/D15016
llvm-svn: 254156
ARMv8.2-A adds a new ID register, ID_A64MMFR2_EL1, which behaves in the
same way as ID_A64MMFR0_EL1 and ID_A64MMFR1_EL1. It is a required part
of ARMv8.2-A, so no additional subtarget features are required.
Differential Revision: http://reviews.llvm.org/D15017
llvm-svn: 254155
This adds subtarget features for ARMv8.2-A, which builds on (and
requires the features from) ARMv8.1-A. Most assembler-visible features
of ARMv8.2-A are system instructions, and are all required parts of the
architecture, so just depend on the HasV8_2aOps subtarget feature. There
is also one large, optional feature, which adds 16-bit floating point
versions of all existing floating-point instructions (VFP and SIMD),
this is represented by the FeatureFullFP16 subtarget feature.
Differential Revision: http://reviews.llvm.org/D15013
llvm-svn: 254154
generated for _mm_losd_s{s,d}() intrinsics and used in scalar FMAs generated
for FMA intrinsics _mm_f{madd,msub,nmadd,nmsub}_s{s,d}().
Reviewer: David Kreitzer
Differential Revision: http://reviews.llvm.org/D14762
llvm-svn: 254140
Summary:
This returns a pointer to the dispatch packet, which can be used to load
information about the kernel dispach.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D14898
llvm-svn: 254116
This is a temporary fix to address ICE on 2005-10-21-longlonggtu.ll.
The proper fix will be to use A2_tfrsi, but it will need more work to
teach all users of A2_tfrsi to also expect a floating-point operand.
llvm-svn: 254099
v2: added more tests, moved the SALU->VALU conversion to a separate function
It looks like it's not possible to get subregisters in the S_ABS lowering
code, and I don't feel like guessing without testing what the correct code
would look like.
llvm-svn: 254095
If virtual registers are created late, mappings to WebAssembly
registers need to be added explicitly. This patch adds a function
to do so and teaches WebAssemblyPeephole to use it. This fixes
an out-of-bounds access on the WARegs vector.
llvm-svn: 254094
Summary:
Many target lowerings copy-paste the code to test SDValues for known constants.
This code can instead be shared in SelectionDAG.cpp, and reused in the targets.
Reviewers: MatzeB, andreadb, tstellarAMD
Subscribers: arsenm, jyknight, llvm-commits
Differential Revision: http://reviews.llvm.org/D14945
llvm-svn: 254085
Instead of trying to move ARGUMENT instructions back up to the top after
they've been scheduled or sunk down, use a fake physical register to
create a liveness constraint that prevents ARGUMENT instructions from
moving down in the first place. This is still not entirely ideal, however
it is more robust than letting them move and moving them back.
llvm-svn: 254084
The e500mc does not actually support the mfocrf instruction; update the
processor definitions to reflect that fact.
Patch by Tom Rix (with some test-case cleanup by me).
llvm-svn: 254064
It was wrong order of operands (from intrinsic to DAG node).
I added more strict type specification for instruction selection.
Differential Revision: http://reviews.llvm.org/D14942
llvm-svn: 254059
This caused PR25607 and also caused Chromium to crash on start-up.
(Also had to update test/CodeGen/X86/avx-splat.ll, which was committed
after shrink wrapping was enabled.)
llvm-svn: 254044
X86 needs to use its own FMA opcodes, preventing the standard FNEG(FMA) pattern table recognition method used by other platforms. This patch adds support for lowering FNEG(FMA(X,Y,Z)) into a single suitably negated FMA instruction.
Fix for PR24364
Differential Revision: http://reviews.llvm.org/D14906
llvm-svn: 254016
This patch fixes the following issues:
1. Fix the return type of X86psadbw: it should not be the same type of inputs.
For vNi8 inputs the output should be vMi64, where M = N/8.
2. Fix the return type of int_x86_avx512_psad_bw_512 accordingly.
3. Fix the definiton of PSADBW, VPSADBW, and VPSADBWY accordingly.
4. Adjust the return type when building a DAG node of X86ISD::PSADBW type.
5. Update related tests.
Differential revision: http://reviews.llvm.org/D14897
llvm-svn: 254010
We had duplicated definitions for the same hardware '[v]movq' instructions. For example with SSE:
def MOVZQI2PQIrr : RS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only
[(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 (scalar_to_vector GR64:$src)))))],
IIC_SSE_MOVDQ>;
def MOV64toPQIrr : RS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v2i64 (scalar_to_vector GR64:$src)))],
IIC_SSE_MOVDQ>, Sched<[WriteMove]>;
As shown in the test case and PR25554:
https://llvm.org/bugs/show_bug.cgi?id=25554
This causes us to miss reusing an operand because later passes don't know these 'movq' are the same instruction.
This patch deletes one pair of these defs.
Sadly, this won't fix the original test case in the bug report. Something else is still broken.
Differential Revision: http://reviews.llvm.org/D14941
llvm-svn: 253988
The one regression in the builtin tests is in the read2 test which now
(again) has many extra copies, but this should be solved once the pass
is replaced with a DAG combine.
llvm-svn: 253974
The patch in http://reviews.llvm.org/D13745 is broken into four parts:
1. New interfaces without functional changes.
2. Use new interfaces in SelectionDAG, while in other passes treat probabilities
as weights.
3. Use new interfaces in all other passes.
4. Remove old interfaces.
This the second patch above. In this patch SelectionDAG starts to use
probability-based interfaces in MBB to add successors but other MC passes are
still using weight-based interfaces. Therefore, we need to maintain correct
weight list in MBB even when probability-based interfaces are used. This is
done by updating weight list in probability-based interfaces by treating the
numerator of probabilities as weights. This change affects many test cases
that check successor weight values. I will update those test cases once this
patch looks good to you.
Differential revision: http://reviews.llvm.org/D14361
llvm-svn: 253965
This patch detects the AVG pattern in vectorized code, which is simply
c = (a + b + 1) / 2, where a, b, and c have the same type which are vectors of
either unsigned i8 or unsigned i16. In the IR, i8/i16 will be promoted to
i32 before any arithmetic operations. The following IR shows such an example:
%1 = zext <N x i8> %a to <N x i32>
%2 = zext <N x i8> %b to <N x i32>
%3 = add nuw nsw <N x i32> %1, <i32 1 x N>
%4 = add nuw nsw <N x i32> %3, %2
%5 = lshr <N x i32> %N, <i32 1 x N>
%6 = trunc <N x i32> %5 to <N x i8>
and with this patch it will be converted to a X86ISD::AVG instruction.
The pattern recognition is done when combining instructions just before type
legalization during instruction selection. We do it here because after type
legalization, it is much more difficult to do pattern recognition based
on many instructions that are doing type conversions. Therefore, for
target-specific instructions (like X86ISD::AVG), we need to take care of type
legalization by ourselves. However, as X86ISD::AVG behaves similarly to
ISD::ADD, I am wondering if there is a way to legalize operands and result
types of X86ISD::AVG together with ISD::ADD. It seems that the current design
doesn't support this idea.
Tests are added for SSE2, AVX2, and AVX512BW and both i8 and i16 types of
variant vector sizes.
Differential revision: http://reviews.llvm.org/D14761
llvm-svn: 253952
Caller saved regs differ between SysV and Win64. Use the tail call available set to scavenge from.
Refactor register info to create new helper to get at tail call GPRs. Added a new test case for windows. Fixed up a number of X64 tests since now RCX is preferred over RDX on SysV.
Differential Revision: http://reviews.llvm.org/D14878
llvm-svn: 253927
With the '=' suffix now indicating which operands are output operands, it's
no longer as important to distinguish between a call's inputs and its outputs
using operand ordering, so we can go back to printing them in the normal order.
llvm-svn: 253925
This distinguishes input operands from output operands. This is something of
a syntactic experiment to see whether the mild amount of clutter this adds is
outweighed by the extra information it conveys to the reader.
llvm-svn: 253922
The current approach to using get_local and set_local is to use them
implicitly, as register uses and defs. Introduce new copy instructions
which are themselves no-ops except for the get_local and set_local
that they imply, so that we use get_local and set_local consistently.
llvm-svn: 253905
WebAssembly is currently using labels to end scopes, so for example a
loop scope looks like this:
BB0_0:
loop BB0_1
...
BB0_1:
with BB0_0 being the label of the first block not in the loop. This
requires that the label be printed even when it's only reachable via
fallthrough. To arrange this, insert a no-op LOOP_END instruction in
such cases at the end of the loop.
llvm-svn: 253901
Always starting blocks at the top of their containing loops works, but creates
unnecessarily deep nesting because it makes all blocks in a loop overlap.
Refine the BLOCK placement algorithm to start blocks at nearest common
dominating points instead, which significantly shrinks them and reduces
overlapping.
llvm-svn: 253876
Andrey Turetskiy