The A64 instruction set includes a generic register syntax for accessing
implementation-defined system registers. The syntax for these registers is:
S<op0>_<op1>_<CRn>_<CRm>_<op2>
The encoding space permitted for implementation-defined system registers
is:
op0 op1 CRn CRm op2
11 xxx 1x11 xxxx xxx
The full encoding space can now be accessed:
op0 op1 CRn CRm op2
xx xxx xxxx xxxx xxx
This is useful to anyone needing to write assembly code supporting new
system registers before the assembler has learned the official names for
them.
llvm-svn: 218753
Summary: The natual vector cast node (similar to bitcast) AArch64ISD::NVCAST
was introduced in r217159 and r217138. This patch adds a missing cast from
v2f32 to v1i64 which is causing some compilation failures. Also added test
cases to cover various modimm types and BUILD_VECTORs with i64 elements.
llvm-svn: 218751
The Cortex-M7 has 3 options for its FPU: none, FPv5-SP-D16 and
FPv5-DP-D16. FPv5 has the same instructions as FP-ARMv8, so it can be
modelled using the same target feature, and all double-precision
operations are already disabled by the fp-only-sp target features.
llvm-svn: 218747
doesn't generate lazy binding stub for a function whose address is taken in
the program.
Differential Revision: http://reviews.llvm.org/D5067
llvm-svn: 218744
in exposing the scalar value to the broadcast DAG fragment so that we
can catch even reloads and fold them into the broadcast.
This is somewhat magical I'm afraid but seems to work. It is also what
the old lowering did, and I've switched an old test to run both
lowerings demonstrating that we get the same result.
Unlike the old code, I'm not lowering f32 or f64 scalars through this
path when we only have AVX1. The target patterns include pretty heinous
code to re-cast those as shuffles when the scalar happens to not be
spilled because AVX1 provides no broadcast mechanism from registers
what-so-ever. This is terribly brittle. I'd much rather go through our
generic lowering code to get this. If needed, we can add a peephole to
get even more opportunities to broadcast-from-spill-slots that are
exposed post-RA, but my suspicion is this just doesn't matter that much.
llvm-svn: 218734
the same speed as pshufd but we can fold loads into the pmovzx
instructions.
This fixes some regressions that came up in the regression test suite
for the new vector shuffle lowering.
llvm-svn: 218733
This allows proper disambiguation of unbounded arrays and arrays of zero
bound ("struct foo { int x[]; };" and "struct foo { int x[0]; }"). GCC
instead produces an upper bound of -1 in the latter situation, but count
seems tidier. This way lower_bound is provided if it's not the language
default and count is provided if the count is known, otherwise it's
omitted. Simple.
If someone wants to look at rdar://problem/12566646 and see if this
change is acceptable to that bug/fix, that might be helpful (see the
empty-and-one-elem-array.ll test case which cites that radar).
llvm-svn: 218726
VPBROADCAST.
This has the somewhat expected pervasive impact. I don't know why
I forgot about this. Everything seems good with lots of significant
improvements in the tests.
llvm-svn: 218724
In special cases select instructions can be eliminated by
replacing them with a cheaper bitwise operation even when the
select result is used outside its home block. The instances implemented
are patterns like
%x=icmp.eq
%y=select %x,%r, null
%z=icmp.eq|neq %y, null
br %z,true, false
==> %x=icmp.ne
%y=icmp.eq %r,null
%z=or %x,%y
br %z,true,false
The optimization is integrated into the instruction
combiner and performed only when all uses of the select result can
be replaced by the select operand proper. For this dominator information
is used and dominance is now a required analysis pass in the combiner.
The optimization itself is iterative. The critical step is to replace the
select result with the non-constant select operand. So the select becomes
local and the combiner iteratively works out simpler code pattern and
eventually eliminates the select.
rdar://17853760
llvm-svn: 218721
Two related things:
1. Fixes a bug when calculating the offset in GetLinearExpression. The code
previously used zext to extend the offset, so negative offsets were converted
to large positive ones.
2. Enhance aliasGEP to deduce that, if the difference between two GEP
allocations is positive and all the variables that govern the offset are also
positive (i.e. the offset is strictly after the higher base pointer), then
locations that fit in the gap between the two base pointers are NoAlias.
Patch by Nick White!
llvm-svn: 218714
Summary:
This patch adds a threshold that controls the number of bonus instructions
allowed for folding branches with common destination. The original code allows
at most one bonus instruction. With this patch, users can customize the
threshold to allow multiple bonus instructions. The default threshold is still
1, so that the code behaves the same as before when users do not specify this
threshold.
The motivation of this change is that tuning this threshold significantly (up
to 25%) improves the performance of some CUDA programs in our internal code
base. In general, branch instructions are very expensive for GPU programs.
Therefore, it is sometimes worth trading more arithmetic computation for a more
straightened control flow. Here's a reduced example:
__global__ void foo(int a, int b, int c, int d, int e, int n,
const int *input, int *output) {
int sum = 0;
for (int i = 0; i < n; ++i)
sum += (((i ^ a) > b) && (((i | c ) ^ d) > e)) ? 0 : input[i];
*output = sum;
}
The select statement in the loop body translates to two branch instructions "if
((i ^ a) > b)" and "if (((i | c) ^ d) > e)" which share a common destination.
With the default threshold, SimplifyCFG is unable to fold them, because
computing the condition of the second branch "(i | c) ^ d > e" requires two
bonus instructions. With the threshold increased, SimplifyCFG can fold the two
branches so that the loop body contains only one branch, making the code
conceptually look like:
sum += (((i ^ a) > b) & (((i | c ) ^ d) > e)) ? 0 : input[i];
Increasing the threshold significantly improves the performance of this
particular example. In the configuration where both conditions are guaranteed
to be true, increasing the threshold from 1 to 2 improves the performance by
18.24%. Even in the configuration where the first condition is false and the
second condition is true, which favors shortcuts, increasing the threshold from
1 to 2 still improves the performance by 4.35%.
We are still looking for a good threshold and maybe a better cost model than
just counting the number of bonus instructions. However, according to the above
numbers, we think it is at least worth adding a threshold to enable more
experiments and tuning. Let me know what you think. Thanks!
Test Plan: Added one test case to check the threshold is in effect
Reviewers: nadav, eliben, meheff, resistor, hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, llvm-commits
Differential Revision: http://reviews.llvm.org/D5529
llvm-svn: 218711
cases.
While clearly we don't need the AVX vector width, these ISA extensions
often cause us to select different instructions and we should cover them
even with the narrow vector width.
Also, while here, nuke the stress_test2 contents. There is no reason to
try to FileCheck this entire body when it is mostly a test for
successfully surviving the code generator.
llvm-svn: 218710
shuffle tests to match that used in the script I posted and now used
consistently in 128-bit tests.
Nothing interesting changing here, just using the label name as the
FileCheck label and a slightly more general comment marker consumption
strategy.
llvm-svn: 218709
updating script so that they are more thorough and consistent.
Specific fixes here include:
- Actually test VEX-encoded AVX mnemonics.
- Actually use an SSE 4.1 run to test SSE 4.1 features!
- Correctly check instructions sequences from the start of the function.
- Elide the shuffle operands and comment designator in a consistent way.
- Test all of the architectures instead of just the ones I was motivated
to manually author.
I've gone back through and fixed up any egregious issues I spotted. Let
me know if I missed something you really dislike.
One downside to this is that we're now not as diligently using FileCheck
variables for registers. I would be much more concerned with this if we
had larger register usage, but there just aren't that interesting of
register choices here and most of the registers are constrained by the
ABI. Ultimately, I don't think this is likely to be the maintenance
burden for these tests and updating them again should be staright
forward.
llvm-svn: 218707
r218129 omits DW_TAG_subprograms which have no inlined subroutines when
emitting -gmlt data. This makes -gmlt very low cost for -O0 builds.
Darwin's dsymutil reasonably considers a CU empty if it has no
subprograms (which occurs with the above optimization in -O0 programs
without any force_inline function calls) and drops the line table, CU,
and everything in this situation, making backtraces impossible.
Until dsymutil is modified to account for this, disable this
optimization on Darwin to preserve the desired functionality.
(see r218545, which should be reverted after this patch, for other
discussion/details)
Footnote:
In the long term, it doesn't look like this scheme (of simplified debug
info to describe inlining to enable backtracing) is tenable, it is far
too size inefficient for optimized code (the DW_TAG_inlined_subprograms,
even once compressed, are nearly twice as large as the line table
itself (also compressed)) and we'll be considering things like Cary's
two level line table proposal to encode all this information directly in
the line table.
llvm-svn: 218702
Note: This version fixed an issue with the TBZ/TBNZ instructions that were
generated in FastISel. The issue was that the 64bit version of TBZ (TBZX)
automagically sets the upper bit of the immediate field that is used to specify
the bit we want to test. To test for any of the lower 32bits we have to first
extract the subregister and use the 32bit version of the TBZ instruction (TBZW).
Original commit message:
Teach selectBranch to fold bit test and branch into a single instruction (TBZ or
TBNZ).
llvm-svn: 218693
No tests for omod since nothing uses it yet, but
this should get rid of the remaining annoying trailing
zeros after some instructions.
llvm-svn: 218692
This commit adds a test which checks that the functions defined in header files will get associated with the header files rather than the source files in the reports.
Differential Revision: http://reviews.llvm.org/D5489
llvm-svn: 218673
This commit fixes llvm-cov's function coverage metric by using the number of executed functions instead of the number of fully covered functions.
Differential Revision: http://reviews.llvm.org/D5196
llvm-svn: 218672
Fixed lowering of this intrinsics in case when mask is v2i1 and v4i1.
Now cmp intrinsics lower in the following way:
(i8 (int_x86_avx512_mask_pcmpeq_q_128
(v2i64 %a), (v2i64 %b), (i8 %mask))) ->
(i8 (bitcast
(v8i1 (insert_subvector undef,
(v2i1 (and (PCMPEQM %a, %b),
(extract_subvector
(v8i1 (bitcast %mask)), 0))), 0))))
llvm-svn: 218669
a flawed direction and causing miscompiles. Read on for details.
Fundamentally, the premise of this patch series was to map
VECTOR_SHUFFLE DAG nodes into VSELECT DAG nodes for all blends because
we are going to *have* to lower to VSELECT nodes for some blends to
trigger the instruction selection patterns of variable blend
instructions. This doesn't actually work out so well.
In order to match performance with the existing VECTOR_SHUFFLE
lowering code, we would need to re-slice the blend in order to fit it
into either the integer or floating point blends available on the ISA.
When coming from VECTOR_SHUFFLE (or other vNi1 style VSELECT sources)
this works well because the X86 backend ensures that these types of
operands to VSELECT get sign extended into '-1' and '0' for true and
false, allowing us to re-slice the bits in whatever granularity without
changing semantics.
However, if the VSELECT condition comes from some other source, for
example code lowering vector comparisons, it will likely only have the
required bit set -- the high bit. We can't blindly slice up this style
of VSELECT. Reid found some code using Halide that triggers this and I'm
hopeful to eventually get a test case, but I don't need it to understand
why this is A Bad Idea.
There is another aspect that makes this approach flawed. When in
VECTOR_SHUFFLE form, we have very distilled information that represents
the *constant* blend mask. Converting back to a VSELECT form actually
can lose this information, and so I think now that it is better to treat
this as VECTOR_SHUFFLE until the very last moment and only use VSELECT
nodes for instruction selection purposes.
My plan is to:
1) Clean up and formalize the target pre-legalization DAG combine that
converts a VSELECT with a constant condition operand into
a VECTOR_SHUFFLE.
2) Remove any fancy lowering from VSELECT during *legalization* relying
entirely on the DAG combine to catch cases where we can match to an
immediate-controlled blend instruction.
One additional step that I'm not planning on but would be interested in
others' opinions on: we could add an X86ISD::VSELECT or X86ISD::BLENDV
which encodes a fully legalized VSELECT node. Then it would be easy to
write isel patterns only in terms of this to ensure VECTOR_SHUFFLE
legalization only ever forms the fully legalized construct and we can't
cycle between it and VSELECT combining.
llvm-svn: 218658
The sign-/zero-extension of the loaded value can be performed by the memory
instruction for free. If the result of the load has only one use and the use is
a sign-/zero-extend, then we emit the proper load instruction. The extend is
only a register copy and will be optimized away later on.
Other instructions that consume the sign-/zero-extended value are also made
aware of this fact, so they don't fold the extend too.
This fixes rdar://problem/18495928.
llvm-svn: 218653
map, this makes sure that we can compile the same code for two different
ABIs (hard and soft float) in the same module.
Update one testcase accordingly (and fix some confusing naming) and
add a new testcase as well with the ordering swapped which would
highlight the problem.
llvm-svn: 218632
This patch improves the target-specific cost model to better handle signed
division by a power of two. The immediate result is that this enables the SLP
vectorizer to do a better job.
http://reviews.llvm.org/D5469
PR20714
llvm-svn: 218607
Runtime unrolling will create a prologue to execute the extra
iterations which is can't divided by the unroll factor. It
generates an if-then-else sequence to jump into a factor -1
times unrolled loop body, like
extraiters = tripcount % loopfactor
if (extraiters == 0) jump Loop:
if (extraiters == loopfactor) jump L1
if (extraiters == loopfactor-1) jump L2
...
L1: LoopBody;
L2: LoopBody;
...
if tripcount < loopfactor jump End
Loop:
...
End:
It means if the unroll factor is 4, the loop body will be 7
times unrolled, 3 are in loop prologue, and 4 are in the loop.
This commit is to use a loop to execute the extra iterations
in prologue, like
extraiters = tripcount % loopfactor
if (extraiters == 0) jump Loop:
else jump Prol
Prol: LoopBody;
extraiters -= 1 // Omitted if unroll factor is 2.
if (extraiters != 0) jump Prol: // Omitted if unroll factor is 2.
if (tripcount < loopfactor) jump End
Loop:
...
End:
Then when unroll factor is 4, the loop body will be copied by
only 5 times, 1 in the prologue loop, 4 in the original loop.
And if the unroll factor is 2, new loop won't be created, just
as the original solution.
llvm-svn: 218604
nodes, and rely exclusively on its logic. This removes a ton of
duplication from the blend lowering and centralizes it in one place.
One downside is that it requires a bunch of hacks to make this work with
the current legalization framework. We have to manually speculate one
aspect of legalizing VSELECT nodes to get everything to work nicely
because the existing legalization framework isn't *actually* bottom-up.
The other grossness is that we somewhat duplicate the analysis of
constant blends. I'm on the fence here. If reviewers thing this would
look better with VSELECT when it has constant operands dumping over tho
VECTOR_SHUFFLE, we could go that way. But it would be a substantial
change because currently all of the actual blend instructions are
matched via patterns in the TD files based around VSELECT nodes (despite
them not being perfect fits for that). Suggestions welcome, but at least
this removes the rampant duplication in the backend.
llvm-svn: 218600
X86 target-specific DAG combining that tried to convert VSELECT nodes
into VECTOR_SHUFFLE nodes that it "knew" would lower into
immediate-controlled blend nodes.
Turns out, we have perfectly good lowering of all these VSELECT nodes,
and indeed that lowering already knows how to handle lowering through
BLENDI to immediate-controlled blend nodes. The code just wasn't getting
used much because this thing forced the world to go through the vector
shuffle lowering. Yuck.
This also exposes that I was too aggressive in avoiding domain crossing
in v218588 with that lowering -- when the other option is to expand into
two 128-bit vectors, it is worth domain crossing. Restore that behavior
now that we have nice tests covering it.
The test updates here fall into two camps. One is where previously we
ended up with an unsigned encoding of the blend operand and now we get
a signed encoding. In most of those places there were elaborate comments
explaining exactly what these operands really mean. Rather than that,
just switch these tests to use the nicely decoded comments that make it
obvious that the final shuffle matches.
The other updates are just removing pointless domain crossing by
blending integers with PBLENDW rather than BLENDPS.
llvm-svn: 218589
AVX-512.
There is no interesting logic yet. Everything ends up eventually
delegating to the generic code to split the vector and shuffle the
halves. Interestingly, that logic does a significantly better job of
lowering all of these types than the generic vector expansion code does.
Mostly, it lets most of the cases fall back to nice AVX2 code rather
than all the way back to SSE code paths.
Step 2 of basic AVX-512 support in the new vector shuffle lowering. Next
up will be to incrementally add direct support for the basic instruction
set to each type (adding tests first).
llvm-svn: 218585
vectors.
Someone will need to build the AVX512 lowering, which should follow
AVX1 and AVX2 *very* closely for AVX512F and AVX512BW resp. I've added
a dummy test which is a port of the v8f32 and v8i32 tests from AVX and
AVX2 to v8f64 and v8i64 tests for AVX512F and AVX512BW. Hopefully this
is enough information for someone to implement proper lowering here. If
not, I'll be happy to help, but right now the AVX-512 support isn't
a priority for me.
llvm-svn: 218583
lowerings.
This was hopelessly broken. First, the x86 backend wants '-1' to be the
element value representing true in a boolean vector, and second the
operand order for VSELECT is backwards from the actual x86 instructions.
To make matters worse, the backend is just using '-1' as the true value
to get the high bit to be set. It doesn't actually symbolically map the
'-1' to anything. But on x86 this isn't quite how it works: there *only*
the high bit is relevant. As a consequence weird non-'-1' values like
0x80 actually "work" once you flip the operands to be backwards.
Anyways, thanks to Hal for helping me sort out what these *should* be.
llvm-svn: 218582
new vector shuffle target DAG combines -- it helps to actually test for
the value you want rather than just using an integer in a boolean
context.
Have I mentioned that I loathe implicit conversions recently? :: sigh ::
llvm-svn: 218576
of widening masks.
We can't widen a zeroing mask unless both elements that would be merged
are either zeroed or undef. This is the only way to widen a mask if it
has a zeroed element.
Also clean up the code here by ordering the checks in a more logical way
and by using the symoblic values for undef and zero. I'm actually torn
on using the symbolic values because the existing code is littered with
the assumption that -1 is undef, and moreover that entries '< 0' are the
special entries. While that works with the values given to these
constants, using the symbolic constants actually makes it a bit more
opaque why this is the case.
llvm-svn: 218575
If there is a store followed by a store with the same value to the same location, then the store is dead/noop. It can be removed.
This problem is found in spec2006-197.parser.
For example,
stur w10, [x11, #-4]
stur w10, [x11, #-4]
Then one of the two stur instructions can be removed.
Patch by David Xu!
llvm-svn: 218569
and in the target shuffle combining when trying to widen vector
elements.
Previously only one of these was correct, and we didn't correctly
propagate zeroing target shuffle masks (which have a different sentinel
value from undef in non- target shuffle masks now). This isn't just
a missed optimization, this caused us to drop zeroing shuffles on the
floor and miscompile code. The added test case is one example of that.
There are other fixes to the test suite as a consequence of this as well
as restoring the undef elements in some of the masks that were lost when
I brought sanity to the actual *value* of the undef and zero sentinels.
I've also just cleaned up some of the PSHUFD and PSHUFLW and PSHUFHW
combining code, but that code really needs to go. It was a nice initial
attempt, but it isn't very principled and the recursive shuffle combiner
is much more powerful.
llvm-svn: 218562
to significantly more sane sentinels. Notably, everywhere else in the
backend's representation of shuffles uses '-1' to represent undef. The
target shuffle masks really shouldn't diverge from that, especially as
in a few places they are manipulated by shared code.
This causes us to lose some undef lanes in various test masks. I want to
get these back, but technically it isn't invalid and there are a *lot*
of bugs here so I want to try to establish a saner baseline for fixing
some of the bugs by aligning the specific senitnel values used.
llvm-svn: 218561
This is purely refactoring. No functional changes intended. PowerPC is the only target
that is currently using this interface.
The ultimate goal is to allow targets other than PowerPC (certainly X86 and Aarch64) to turn this:
z = y / sqrt(x)
into:
z = y * rsqrte(x)
And:
z = y / x
into:
z = y * rcpe(x)
using whatever HW magic they can use. See http://llvm.org/bugs/show_bug.cgi?id=20900 .
There is one hook in TargetLowering to get the target-specific opcode for an estimate instruction
along with the number of refinement steps needed to make the estimate usable.
Differential Revision: http://reviews.llvm.org/D5484
llvm-svn: 218553
Users of getSectionContents shouldn't try to pass in BSS or virtual
sections. In all instances, this is a bug in the code calling this
routine.
N.B. Some COFF implementations (like CL) will mark their BSS sections as
taking space on disk. This would confuse COFFObjectFile into thinking
the section is larger than the file.
llvm-svn: 218549
So in fully linked images when a call is made through a stub it now gets a
comment like the following in the disassembly:
callq 0x100000f6c ## symbol stub for: _printf
indicating the call is to a symbol stub and which symbol it is for. This is
done for branch reference types and seeing if the branch target is in a stub
section and if so using the indirect symbol table entry for that stub and
using that symbol table entries symbol name.
llvm-svn: 218546
that managed to elude all of my fuzz testing historically. =/
Something changed to allow this code path to actually be exercised and
it was doing bad things. It is especially heavily exercised by the
patterns that emerge when doing AVX shuffles that end up lowered through
the 128-bit code path.
llvm-svn: 218540
This has weird operand requirements so it's worthwhile
to have very strict checks for its operands.
Add different combinations of SGPR operands.
llvm-svn: 218535
Instead of moving the first SGPR that is different than the first,
legalize the operand that requires the fewest moves if one
SGPR is used for multiple operands.
This saves extra moves and is also required for some instructions
which require that the same operand be used for multiple operands.
llvm-svn: 218532
Disable the SGPR usage restriction parts of the DAG legalizeOperands.
It now should only be doing immediate folding until it can be replaced
later. The real legalization work is now done by the other
SIInstrInfo::legalizeOperands
llvm-svn: 218531
e.g. v_cndmask_b32 requires the condition operand be an SGPR.
If one of the source operands were an SGPR, that would be considered
the one SGPR use and the condition operand would be illegally moved.
llvm-svn: 218529
No test since the current SIISelLowering::legalizeOperands
effectively hides this, and the general uses seem to only fire
on SALU instructions which don't have modifiers between
the operands.
When trying to use legalizeOperands immediately after
instruction selection, it now sees a lot more patterns
it did not see before which break on this.
llvm-svn: 218527
The annotation instructions are dropped during codegen and have no
impact on size. In some cases, the annotations were preventing the
unroller from unrolling a loop because the annotation calls were
pushing the cost over the unrolling threshold.
Differential Revision: http://reviews.llvm.org/D5335
llvm-svn: 218525
layer of tie-breaking sorting, it really helps to check that you're in
a tie first. =] Otherwise the whole thing cycles infinitely. Test case
added, another one found through fuzz testing.
llvm-svn: 218523
AVX support.
New test cases included. Note that none of the existing test cases
covered these buggy code paths. =/ Also, it is clear from this that
SHUFPS and SHUFPD are the most bug prone shuffle instructions in x86. =[
These were all detected by fuzz-testing. (I <3 fuzz testing.)
llvm-svn: 218522
This patch makes the ARM backend transform 3 operand instructions such as
'adds/subs' to the 2 operand version of the same instruction if the first
two register operands are the same.
Example: 'adds r0, r0, #1' will is transformed to 'adds r0, #1'.
Currently for some instructions such as 'adds' if you try to assemble
'adds r0, r0, #8' for thumb v6m the assembler would throw an error message
because the immediate cannot be encoded using 3 bits.
The backend should be smart enough to transform the instruction to
'adds r0, #8', which allows for larger immediate constants.
Patch by Ranjeet Singh.
llvm-svn: 218521
based on the Function. This is currently used to implement
mips16 support in the mips backend via the existing module
pass resetting the subtarget.
Things to note:
a) This involved running resetTargetOptions before creating a
new subtarget so that code generation options like soft-float
could be recognized when creating the new subtarget. This is
to deal with initialization code in isel lowering that only
paid attention to the initial value.
b) Many of the existing testcases weren't using the soft-float
feature correctly. I've corrected these based on the check
values assuming that was the desired behavior.
c) The mips port now pays attention to the target-cpu and
target-features strings when generating code for a particular
function. I've removed these from one function where the
requested cpu and features didn't match the check lines in
the testcase.
llvm-svn: 218492
Machine Sink uses loop depth information to select between successors BBs to
sink machine instructions into, where BBs within smaller loop depths are
preferable. This patch adds support for choosing between successors by using
profile information from BlockFrequencyInfo instead, whenever the information
is available.
Tested it under SPEC2006 train (average of 30 runs for each program); ~1.5%
execution speedup in average on x86-64 darwin.
<rdar://problem/18021659>
llvm-svn: 218472
Summary:
I originally tried doing this specifically for X86 in the backend in D5091,
but it was rather brittle and generally running too late to be general.
Furthermore, other targets may want to implement similar optimizations.
So I reimplemented it at the IR-level, fitting it into AtomicExpandPass
as it interacts with that pass (which could not be cleanly done before
at the backend level).
This optimization relies on a new target hook, which is only used by X86
for now, as the correctness of the optimization on other targets remains
an open question. If it is found correct on other targets, it should be
trivial to enable for them.
Details of the optimization are discussed in D5091.
Test Plan: make check-all + a new test
Reviewers: jfb
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5422
llvm-svn: 218455
These instructions do not indicate they are extendable or the
number of bits in the extendable operand. Rename to match
architected names. Add a testcase for the intrinsics.
llvm-svn: 218453
Summary:
The N32/N64 ABI's require that structs passed in registers are laid out
such that spilling the register with 'sd' places the struct at the lowest
address. For little endian this is trivial but for big-endian it requires
that structs are shifted into the upper bits of the register.
We also require that structs passed in registers have the 'inreg'
attribute for big-endian N32/N64 to work correctly. This is because the
tablegen-erated calling convention implementation only has access to the
lowered form of struct arguments (one or more integers of up to 64-bits
each) and is unable to determine the original type.
Reviewers: vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5286
llvm-svn: 218451
On ARM NEON, VAND with immediate (16/32 bits) is an alias to VBIC ~imm with
the same type size. Adding that logic to the parser, and generating VBIC
instructions from VAND asm files.
This patch also fixes the validation routines for NEON splat immediates which
were wrong.
Fixes PR20702.
llvm-svn: 218450
v4f64 and v8f32 shuffles when they are lane-crossing. We have fully
general lane-crossing permutation functions in AVX2 that make this easy.
Part of this also changes exactly when and how these vectors are split
up when we don't have AVX2. This isn't always a win but it usually is
a win, so on the balance I think its better. The primary regressions are
all things that just need to be fixed anyways such as modeling when
a blend can be completely accomplished via VINSERTF128, etc.
Also, this highlights one of the few remaining big features: we do
a really poor job of inserting elements into AVX registers efficiently.
This completes almost all of the big tricks I have in mind for AVX2. The
only things left that I plan to add:
1) element insertion smarts
2) palignr and other fairly specialized lowerings when they happen to
apply
llvm-svn: 218449
256-bit vectors with lane-crossing.
Rather than immediately decomposing to 128-bit vectors, try flipping the
256-bit vector lanes, shuffling them and blending them together. This
reduces our worst case shuffle by a pretty significant margin across the
board.
llvm-svn: 218446
The Thumb2 BXJ instruction (Branch and Exchange Jazelle) is not
defined for v7M or v8A. It is defined for all other Thumb2-supporting
architectures (v6T2, v7A and v7R).
llvm-svn: 218445
lowering where it only used the mask of the low 128-bit lane rather than
the entire mask.
This allows the new lowering to correctly match the unpack patterns for
v8i32 vectors.
For reference, the reason that we check for the the entire mask rather
than checking the repeated mask is because the repeated masks don't
abide by all of the invariants of normal masks. As a consequence, it is
safer to use the full mask with functions like the generic equivalence
test.
llvm-svn: 218442
lowering.
This completes the basic AVX2 feature support, but there are still some
improvements I'd like to do to really get the last mile of performance
here.
llvm-svn: 218440
I made a mistake in the previous commit and produced the wrong pattern.
Fix that. Also make one more shuffle pattern byte-based rather than
word-based, and add two more blend patterns.
llvm-svn: 218439
shuffles rather than word shuffles.
As you might guess, these were built starting from the word shuffle test
cases and I failed to properly port a bunch of them and left them as
widened word shuffle test cases. We still have a couple of tests that
check our ability to widen shuffles, but now we will test the actual
byte shuffle quite a bit better.
llvm-svn: 218438
Nico Rieck added support for this 32-bit COFF relocation some time ago
for Win64 stuff. It appears that as an oversight, the assembly output
used "foo"@IMGREL32 instead of "foo"@IMGREL, which is what we can parse.
Sadly, there were actually tests that took in IMGREL and put out
IMGREL32, and we didn't notice the inconsistency. Oh well. Now LLVM can
assemble it's own output with slightly more fidelity.
llvm-svn: 218437
missing test cases for it.
Unsurprisingly, without test cases, there were bugs here. Surprisingly,
this bug wasn't caught at compile time. Yep, there is an X86ISD::BLENDV.
It isn't wired to anything. Oops. I'll fix than next.
llvm-svn: 218434
If we have multiple coverage counts for the same segment, we need to
add them up rather than arbitrarily choosing one. This fixes that and
adds a test with template instantiations to exercise it.
llvm-svn: 218432
lowering.
This also implements the fancy blend lowering for v16i16 using AVX2 and
teaches the X86 backend to print shuffle masks for 256-bit PSHUFB
and PBLENDW instructions. It also makes the mask decoding correct for
PBLENDW instructions. The yaks, they are legion.
Tests are updated accordingly. There are some missing tests for the
VBLENDVB lowering, but I'll add those in a follow-up as this commit has
accumulated enough cruft already.
llvm-svn: 218430
get the literal string “Hello world” printed as a comment on the instruction
that loads the pointer to it. For now this is just for x86_64. So for object
files with relocation entries it produces things like:
leaq L_.str(%rip), %rax ## literal pool for: "Hello world\n"
and similar for fully linked images like executables:
leaq 0x4f(%rip), %rax ## literal pool for: "Hello world\n"
Also to allow testing against darwin’s otool(1), I hooked up the existing
-no-show-raw-insn option to the Mach-O parser code, added the new Mach-O
only -full-leading-addr option to match otool(1)'s printing of addresses and
also added the new -print-imm-hex option.
llvm-svn: 218423
For biendian targets like ARM and AArch64, it is useful to have the
output of the llvm-dwarfdump and llvm-objdump report the endianness
used when the object files were generated.
Patch by Charlie Turner.
llvm-svn: 218408
This change fixes the ARM and AArch64 relocation visitors in
RelocVisitor. They were unconditionally assuming the object data are
little-endian. Tests have been added to ensure that the
llvm-dwarfdump utility does not crash when processing big-endian
object files.
Patch by Charlie Turner.
llvm-svn: 218407
This change replaces the brittle if/else chain of string comparisons
with a switch statement on the detected target triple, removing the
need for testing arbitrary architecture names returned from
getFileFormatName, whose primary purpose seems to be for display
(user-interface) purposes. The visitor now takes a reference to the
object file, rather than its arbitrary file format name to figure out
whether the file is a 32 or 64-bit object file and what the detected
target triple is.
A set of tests have been added to help show that the refactoring processes
relocations for the same targets as the original code.
Patch by Charlie Turner.
llvm-svn: 218406
Use the same environment when invoking llvm-config from lit.cfg as
will be used when running tests, so that ASAN_OPTIONS, INCLUDE, etc.
are present.
llvm-svn: 218403
This reverts commit faac033f7364bb4226e22c8079c221c96af10d02.
The test depends on all targets to be enabled in llc in order to pass,
and needs to be rewritten/refactored to not have that dependency.
llvm-svn: 218393
For biendian targets like ARM and AArch64, it is useful to have the
output of the llvm-dwarfdump and llvm-objdump report the endianness
used when the object files were generated.
Patch by Charlie Turner.
llvm-svn: 218391
This change fixes the ARM and AArch64 relocation visitors in
RelocVisitor. They were unconditionally assuming the object data are
little-endian. Tests have been added to ensure that the
llvm-dwarfdump utility does not crash when processing big-endian
object files.
Patch by Charlie Turner.
llvm-svn: 218389
This change replaces the brittle if/else chain of string comparisons
with a switch statement on the detected target triple, removing the
need for testing arbitrary architecture names returned from
getFileFormatName, whose primary purpose seems to be for display
(user-interface) purposes. The visitor now takes a reference to the
object file, rather than its arbitrary file format name to figure out
whether the file is a 32 or 64-bit object file and what the detected
target triple is.
A set of tests have been added to help show that the refactoring processes
relocations for the same targets as the original code.
Patch by Charlie Turner.
llvm-svn: 218388
The doFinalization method checks that the LoopToAliasSetMap is
empty. LICM populates that map as it runs through the loop nest,
deleting the entries for child loops as it goes. However, if a child
loop is deleted by another pass (e.g. unrolling) then the loop will
never be deleted from the map because LICM walks the loop nest to
find entries it can delete.
The fix is to delete the loop from the map and free the alias set
when the loop is deleted from the loop nest.
Differential Revision: http://reviews.llvm.org/D5305
llvm-svn: 218387
If it's safe to clobber the condition flags, we can do a few extra things:
it's then possible to reset the base register writeback using a SUBS, so
we can try to merge even if the base register isn't dead after the merged
instruction.
This is effectively a (heavily bug-fixed) rewrite of r208992.
llvm-svn: 218386
v7M only allows the 16-bit encoding of the 'cps' (Change Processor
State) instruction, and does not have the 32-bit encoding which is
valid from v6T2 onwards.
llvm-svn: 218382
pool data being loaded into a vector register.
The comments take the form of:
# ymm0 = [a,b,c,d,...]
# xmm1 = <x,y,z...>
The []s are used for generic sequential data and the <>s are used for
specifically ConstantVector loads. Undef elements are printed as the
letter 'u', integers in decimal, and floating point values as floating
point values. Suggestions on improving the formatting or other aspects
of the display are very welcome.
My primary use case for this is to be able to FileCheck test masks
passed to vector shuffle instructions in-register. It isn't fantastic
for that (no decoding special zeroing semantics or other tricks), but it
at least puts the mask onto an instruction line that could reasonably be
checked. I've updated many of the new vector shuffle lowering tests to
leverage this in their test cases so that we're actually checking the
shuffle masks remain as expected.
Before implementing this, I tried a *bunch* of different approaches.
I looked into teaching the MCInstLower code to scan up the basic block
and find a definition of a register used in a shuffle instruction and
then decode that, but this seems incredibly brittle and complex.
I talked to Hal a lot about the "right" way to do this: attach the raw
shuffle mask to the instruction itself in some form of unencoded
operands, and then use that to emit the comments. I still think that's
the optimal solution here, but it proved to be beyond what I'm up for
here. In particular, it seems likely best done by completing the
plumbing of metadata through these layers and attaching the shuffle mask
in metadata which could have fully automatic dropping when encoding an
actual instruction.
llvm-svn: 218377
the native AVX2 instructions.
Note that the test case is really frustrating here because VPERMD
requires the mask to be in the register input and we don't produce
a comment looking through that to the constant pool. I'm going to
attempt to improve this in a subsequent commit, but not sure if I will
succeed.
llvm-svn: 218347
detection. It was incorrectly handling undef lanes by actually treating
an undef lane in the first 128-bit lane as a *numeric* shuffle value.
Fortunately, this almost always DTRT and disabled detecting repeated
patterns. But not always. =/ This patch introduces a much more
principled approach and fixes the miscompiles I spotted by inspection
previously.
llvm-svn: 218346
This testcase was not testing what it meant: because there were only two checks for
dmb {{ish}} in the second function, it could have missed a bug where one of the three
required dmb {{ish}} became dmb {{ishst}}. As I was fixing it, I also added
CHECK-LABELs to make it a bit less brittle.
llvm-svn: 218341
shuffles using the AVX2 instructions. This is the first step of cutting
in real AVX2 support.
Note that I have spotted at least one bug in the test cases already, but
I suspect it was already present and just is getting surfaced. Will
investigate next.
llvm-svn: 218338
Rather than slurping in and splatting out the whole ctor list, preserve
the existing array entries without trying to understand them. Only
remove the entries that we know we can optimize away. This way we don't
need to wire through priority and comdats or anything else we might add.
Fixes a linker issue where the .init_array or .ctors entry would point
to discarded initialization code if the comdat group from the TU with
the faulty global_ctors entry was dropped.
llvm-svn: 218337
e.g., add w1, w2, w3, lsl #(2 - 1)
This sort of thing comes up in pre-processed assembly playing macro games.
Still validate that it's an assembly time constant. The early exit error check
was just a bit overzealous and disallowed a left paren.
rdar://18430542
llvm-svn: 218336
add VPBLENDD to the InstPrinter's comment generation so we get nice
comments everywhere.
Now that we have the nice comments, I can see the bug introduced by
a silly typo in the commit that enabled VPBLENDD, and have fixed it. Yay
tests that are easy to inspect.
llvm-svn: 218335
Summary:
AtomicExpand already had logic for expanding wide loads and stores on LL/SC
architectures, and for expanding wide stores on CmpXchg architectures, but
not for wide loads on CmpXchg architectures. This patch fills this hole,
and makes use of this new feature in the X86 backend.
Only one functionnal change: we now lose the SynchScope attribute.
It is regrettable, but I have another patch that I will submit soon that will
solve this for all of AtomicExpand (it seemed better to split it apart as it
is a different concern).
Test Plan: make check-all (lots of tests for this functionality already exist)
Reviewers: jfb
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5404
llvm-svn: 218332
Summary:
This patch makes use of AtomicExpandPass in Power for inserting fences around
atomic as part of an effort to remove fence insertion from SelectionDAGBuilder.
As a big bonus, it lets us use sync 1 (lightweight sync, often used by the mnemonic
lwsync) instead of sync 0 (heavyweight sync) in many cases.
I also added a test, as there was no test for the barriers emitted by the Power
backend for atomic loads and stores.
Test Plan: new test + make check-all
Reviewers: jfb
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5180
llvm-svn: 218331
VPBLENDD where appropriate even on 128-bit vectors.
According to Agner's tables, this instruction is significantly higher
throughput (can execute on any port) on Haswell chips so we should
aggressively try to form it when available.
Sadly, this loses our delightful shuffle comments. I'll add those back
for VPBLENDD next.
llvm-svn: 218322
undef in the shuffle mask. This shows up when we're printing comments
during lowering and we still have an IR-level constant hanging around
that models undef.
A nice consequence of this is *much* prettier test cases where the undef
lanes actually show up as undef rather than as a particular set of
values. This also allows us to print shuffle comments in cases that use
undef such as the recently added variable VPERMILPS lowering. Now those
test cases have nice shuffle comments attached with their details.
The shuffle lowering for PSHUFB has been augmented to use undef, and the
shuffle combining has been augmented to comprehend it.
llvm-svn: 218301
trick that I missed.
VPERMILPS has a non-immediate memory operand mode that allows it to do
asymetric shuffles in the two 128-bit lanes. Use this rather than two
shuffles and a blend.
However, it turns out the variable shuffle path to VPERMILPS (and
VPERMILPD, although that one offers no functional differenc from the
immediate operand other than variability) wasn't even plumbed through
codegen. Do such plumbing so that we can reasonably emit
a variable-masked VPERMILP instruction. Also plumb basic comment parsing
and printing through so that the tests are reasonable.
There are still a few tests which don't show the shuffle pattern. These
are tests with undef lanes. I'll teach the shuffle decoding and printing
to handle undef mask entries in a follow-up. I've looked at the masks
and they seem reasonable.
llvm-svn: 218300
This includes constants, attributes, and some additional instructions not covered by previous tests.
Work was done by lama.saba@intel.com.
llvm-svn: 218297
We manage to generate all of the matching instructions (and a lot more) via
the reciprocal optimization function - even if we completely remove the square
root optimization. With CHECK_NEXT, we assure that we're executing the
expected square root optimization paths and not generating extra insts.
llvm-svn: 218284
Shift-left immediate with sign-/zero-extensions also works for boolean values.
Update the assert and the test cases to reflect that fact.
This should fix a bug found by Chad.
llvm-svn: 218275
These are just test cases, no actual code yet. This establishes the
baseline fallback strategy we're starting from on AVX2 and the expected
lowering we use on AVX1.
Also, these test cases are very much generated. I've manually crafted
the specific pattern set that I'm hoping will be useful at exercising
the lowering code, but I've not (and could not) manually verify *all* of
these. I've spot checked and they seem legit to me.
As with the rest of vector shuffling, at a certain point the only really
useful way to check the correctness of this stuff is through fuzz
testing.
llvm-svn: 218267
We generate broadcast instructions on CPUs with AVX2 to load some constant splat vectors.
This patch should preserve all existing behavior with regular optimization levels,
but also use splats whenever possible when optimizing for *size* on any CPU with AVX or AVX2.
The tradeoff is up to 5 extra instruction bytes for the broadcast instruction to save
at least 8 bytes (up to 31 bytes) of constant pool data.
Differential Revision: http://reviews.llvm.org/D5347
llvm-svn: 218263
This reverts commit r218254.
The global_atomics.ll test fails with asserts disabled. For some reason,
the compiler fails to produce the atomic no return variants.
llvm-svn: 218257
Summary:
Update segmented-stacks*.ll tests with x32 target case and make
corresponding changes to make them pass.
Test Plan: tests updated with x32 target
Reviewers: nadav, rafael, dschuff
Subscribers: llvm-commits, zinovy.nis
Differential Revision: http://reviews.llvm.org/D5245
llvm-svn: 218247
Summary: getSubroutineName is currently only used by llvm-symbolizer, thus add a binary test containing a cross-cu inlining example.
Reviewers: samsonov, dblaikie
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5394
llvm-svn: 218245
The PSHUFB mask decode routine used to assert if the mask index was out of
range (<0 or greater than the size of the vector). The problem is, we can
legitimately have a PSHUFB with a large index using intrinsics. The
instruction only uses the least significant 4 bits. This change removes the
assert and masks the index to match the instruction behaviour.
llvm-svn: 218242
We currently emit an error when trying to assemble a file with more
than one section using DWARF2 debug info. This should be a warning
instead, as the resulting file will still be usable, but with a
degraded debug illusion.
llvm-svn: 218241
a more sane approach to AVX2 support.
Fundamentally, there is no useful way to lower integer vectors in AVX.
None. We always end up with a VINSERTF128 in the end, so we might as
well eagerly switch to the floating point domain and do everything
there. This cleans up lots of weird and unlikely to be correct
differences between integer and floating point shuffles when we only
have AVX1.
The other nice consequence is that by doing things this way we will make
it much easier to write the integer lowering routines as we won't need
to duplicate the logic to check for AVX vs. AVX2 in each one -- if we
actually try to lower a 256-bit vector as an integer vector, we have
AVX2 and can rely on it. I think this will make the code much simpler
and more comprehensible.
Currently, I've disabled *all* support for AVX2 so that we always fall
back to AVX. This keeps everything working rather than asserting. That
will go away with the subsequent series of patches that provide
a baseline AVX2 implementation.
Please note, I'm going to implement AVX2 *without access to hardware*.
That means I cannot correctness test this path. I will be relying on
those with access to AVX2 hardware to do correctness testing and fix
bugs here, but as a courtesy I'm trying to sketch out the framework for
the new-style vector shuffle lowering in the context of the AVX2 ISA.
llvm-svn: 218228
input v8f32 shuffles which are not 128-bit lane crossing but have
different shuffle patterns in the low and high lanes. This removes most
of the extract/insert traffic that was unnecessary and is particularly
good at lowering cases where only one of the two lanes is shuffled at
all.
I've also added a collection of test cases with undef lanes because this
lowering is somewhat more sensitive to undef lanes than others.
llvm-svn: 218226
in the high and low 128-bit lanes of a v8f32 vector.
No functionality change yet, but wanted to set up the baseline for my
next patch which will make these quite a bit better. =]
llvm-svn: 218224
lowering when it can use a symmetric SHUFPS across both 128-bit lanes.
This required making the SHUFPS lowering tolerant of other vector types,
and adjusting our canonicalization to canonicalize harder.
This is the last of the clever uses of symmetry I've thought of for
v8f32. The rest of the tricks I'm aware of here are to work around
assymetry in the mask.
llvm-svn: 218216
of a single element into a zero vector for v4f64 and v4i64 in AVX.
Ironically, there is less to see here because xor+blend is so crazy fast
that we can't really beat that to zero the high 128-bit lane.
llvm-svn: 218214
UNPCKHPS with AVX vectors by recognizing those patterns when they are
repeated for both 128-bit lanes.
With this, we now generate the exact same (really nice) code for
Quentin's avx_test_case.ll which was the most significant regression
reported for the new shuffle lowering. In fact, I'm out of specific test
cases for AVX lowering, the rest were AVX2 I think. However, there are
a bunch of pretty obvious remaining things to improve with AVX...
llvm-svn: 218213
important bits of cleverness: to detect and lower repeated shuffle
patterns between the two 128-bit lanes with a single instruction.
This patch just teaches it how to lower single-input shuffles that fit
this model using VPERMILPS. =] There is more that needs to happen here.
llvm-svn: 218211
generating the test cases to format things more consistently and
actually catch all the operand sequences that should be elided in favor
of the asm comments. No actual changes here.
llvm-svn: 218210
VBLENDPD over using VSHUFPD. While the 256-bit variant of VBLENDPD slows
down to the same speed as VSHUFPD on Sandy Bridge CPUs, it has twice the
reciprocal throughput on Ivy Bridge CPUs much like it does everywhere
for 128-bits. There isn't a downside, so just eagerly use this
instruction when it suffices.
llvm-svn: 218208
This expands the integer cases to cover the fact that AVX2 moves their
lane-crossing shuffles into the integer domain. It also adds proper
support for AVX2 run lines and the "ALL" group when it doesn't matter.
llvm-svn: 218206
actual support for complex AVX shuffling tricks. We can do independent
blends of the low and high 128-bit lanes of an avx vector, so shuffle
the inputs into place and then do the blend at 256 bits. This will in
many cases remove one blend instruction.
The next step is to permute the low and high halves in-place rather than
extracting them and re-inserting them.
llvm-svn: 218202
link.exe:
Fuzz testing has shown that COMMON symbols with size > 32 will always
have an alignment of at least 32 and all symbols with size < 32 will
have an alignment of at least the largest power of 2 less than the size
of the symbol.
binutils:
The BFD linker essentially work like the link.exe behavior but with
alignment 4 instead of 32. The BFD linker also supports an extension to
COFF which adds an -aligncomm argument to the .drectve section which
permits specifying a precise alignment for a variable but MC currently
doesn't support editing .drectve in this way.
With all of this in mind, we decide to play a little trick: we can
ensure that the alignment will be respected by bumping the size of the
global to it's alignment.
llvm-svn: 218201
under AVX.
This really just documents the current state of the world. I'm going to
try to flesh it out to cover any test cases I plan to improve prior to
improving them so that the delta made by changes is actually visible to
code reviewers.
This is made easier by the fact that I now have a script to automate the
process of producing test cases including the check lines. =]
llvm-svn: 218199
single-input shuffles with doubles. This allows them to fold memory
operands into the shuffle, etc. This is just the analog to the v4f32
case in my prior commit.
llvm-svn: 218193
instruction for single-vector floating point shuffles. This in turn
allows the shuffles to fold a load into the instruction which is one of
the common regressions hit with the new shuffle lowering.
llvm-svn: 218190
We had a few bugs:
- We were considering the GVKind instead of just looking at the section
characteristics
- We would never print out 'y' when a section was meant to be unreadable
- We would never print out 's' when a section was meant to be shared
- We translated IMAGE_SCN_MEM_DISCARDABLE to 'n' when it should've meant
IMAGE_SCN_LNK_REMOVE
llvm-svn: 218189
duplication of check lines. The idea is to have broad sets of
compilation modes that will frequently diverge without having to always
and immediately explode to the precise ISA feature set.
While this already helps due to VEX encoded differences, it will help
much more as I teach the new shuffle lowering about more of the new VEX
encoded instructions which can still be used to implement 128-bit
shuffles.
llvm-svn: 218188
A problem with our old behavior becomes observable under x86-64 COFF
when we need a read-only GV which has an initializer which is referenced
using a relocation: we would mark the section as writable. Marking the
section as writable interferes with section merging.
This fixes PR21009.
llvm-svn: 218179
tricky case of single-element insertion into the zero lane of a zero
vector.
We can't just use the same pattern here as we do in every other vector
type because the general insertion logic can handle insertion into the
non-zero lane of the vector. However, in SSE4.1 with v4f32 vectors we
have INSERTPS that is a much better choice than the generic one for such
lowerings. But INSERTPS can do lots of other lowerings as well so
factoring its logic into the general insertion logic doesn't work very
well. We also can't just extract the core common part of the general
insertion logic that is faster (forming VZEXT_MOVL synthetic nodes that
lower to MOVSS when they can) because VZEXT_MOVL is often *faster* than
a blend while INSERTPS is slower! So instead we do a restrictive
condition on attempting to use the generic insertion logic to narrow it
to those cases where VZEXT_MOVL won't need a shuffle afterward and thus
will do better than INSERTPS. Then we try blending. Then we go back to
INSERTPS.
This still doesn't generate perfect code for some silly reasons that can
be fixed by tweaking the td files for lowering VZEXT_MOVL to use
XORPS+BLENDPS when available rather than XORPS+MOVSS when the input ends
up in a register rather than a load from memory -- BLENDPSrr has twice
the reciprocal throughput of MOVSSrr. Don't you love this ISA?
llvm-svn: 218177
floating point types and use it for both v2f64 and v2i64 single-element
insertion lowering.
This fixes the last non-AVX performance regression test case I've gotten
of for the new vector shuffle lowering. There is obvious analogous
lowering for v4f32 that I'll add in a follow-up patch (because with
INSERTPS, v4f32 requires special treatment). After that, its AVX stuff.
llvm-svn: 218175
When looking through sign/zero-extensions the code would always assume there is
such an extension instruction and use the wrong operand for the address.
There was also a minor issue in the handling of 'AND' instructions. I
accidentially used a 'cast' instead of a 'dyn_cast'.
llvm-svn: 218161
lowering to support both anyext and zext and to custom lower for many
different microarchitectures.
Using this allows us to get *exactly* the right code for zext and anyext
shuffles in all the vector sizes. For v16i8, the improvement is *huge*.
The new SSE2 test case added I refused to add before this because it was
sooooo muny instructions.
llvm-svn: 218143
To reduce the size of -gmlt data, skip the subprograms without any
inlined subroutines. Since we've now got the ability to make these
determinations in the backend (funnily enough - we added the flag so we
wouldn't produce ranges under -gmlt, but with this change we use the
flag, but go back to producing ranges under -gmlt).
Instead, just produce CU ranges to inform the consumer which parts of
the code are described by this CU's line table. Tools could inspect the
line table directly to compute the range, but the CU ranges only seem to
be about 0.5% of object/executable size, so I'm not too worried about
teaching llvm-symbolizer that trick just yet - it's certainly a possible
piece of future work.
Update an llvm-symbolizer test just to demonstrate that this schema is
acceptable there (if it wasn't, the compiler-rt tests would catch this,
but good to have an in-llvm-tree test for llvm-symbolizer's behavior
here)
Building the clang binary with -gmlt with this patch reduces the total
size of object files by 5.1% (5.56% without ranges) without compression
and the executable by 4.37% (4.75% without ranges).
llvm-svn: 218129
The heuristic used by DAGCombine to form FMAs checks that the FMUL has only one
use, but this is overly-conservative on some systems. Specifically, if the FMA
and the FADD have the same latency (and the FMA does not compete for resources
with the FMUL any more than the FADD does), there is no need for the
restriction, and furthermore, forming the FMA leaving the FMUL can still allow
for higher overall throughput and decreased critical-path length.
Here we add a new TLI callback, enableAggressiveFMAFusion, false by default, to
elide the hasOneUse check. This is enabled for PowerPC by default, as most
PowerPC systems will benefit.
Patch by Olivier Sallenave, thanks!
llvm-svn: 218120
to undef lanes as well as defined widenable lanes. This dramatically
improves the lowering we use for undef-shuffles in a zext-ish pattern
for SSE2.
llvm-svn: 218115
Not sure why I only did SSSE3 here. Also, I've left out some of the SSE2
ones because the shuffles are so absurd it's not worth transcribing
them. Will try to fix them to be sane and then check them.
llvm-svn: 218114
shuffles that are zext-ing.
Not a lot to see here; the undef lane variant is better handled with
pshufd, but this improves the actual zext pattern.
llvm-svn: 218112
to the new vector shuffle lowering code.
This allows us to emit PMOVZX variants consistently for patterns where
it is a viable lowering. This instruction is both fast and allows us to
fold loads into it. This only hooks the new lowering up for i16 and i8
element widths, mostly so I could manage the change to the tests. I'll
add the i32 one next, although it is significantly less interesting.
One thing to note is that we already had some tests for these patterns
but those tests had far less horrible instructions. The problem is that
those tests weren't checking the strict start and end of the instruction
sequence. =[ As a consequence something changed in the lowering making
us generate *TERRIBLE* code for these patterns in SSE2 through SSSE3.
I've consolidated all of the tests and spelled out the madness that we
currently emit for these shuffles. I'm going to try to figure out what
has gone wrong here.
llvm-svn: 218102
With this optimization, we will not always insert zext for values crossing
basic blocks, but insert sext if the users of a value crossing basic block
has preference of sign predicate.
llvm-svn: 218101
Tom Coxon