[x86] Fix two independent miscompiles in the process of getting the same
test case to actually generate correct code.
The primary miscompile fixed here is that we weren't correctly handling
in-place elements in one half of a single-input v8i16 shuffle when
moving a dword of elements from that half to the other half. Some times,
we would clobber the in-place elements in forming the dword to move
across halves.
The fix to this involves forcibly marking the in-place inputs even when
there is no need to gather them into a dword, and to much more carefully
re-arrange the elements when grouping them into a dword to move across
halves. With these two changes we would generate correct shuffles for
the test case, but found another miscompile. There are also some random
perturbations of the generated shuffle pattern in SSE2. It looks like
a wash; more instructions in some cases fewer in others.
The second miscompile would corrupt the results into nonsense. This is
a buggy pattern in one of the added DAG combines. Mapping elements
through a PSHUFD when pairing redundant half-shuffles is *much* harder
than this code makes it out to be -- it requires reasoning about *all*
of where the input is used in the PSHUFD, not just one part of where it
is used. Plus, we can't combine a half shuffle *into* a PSHUFD but the
code didn't guard against it. I think this was just a bad idea and I've
just removed that aspect of the combine. No tests regress as
a consequence so seems OK.
[x86] Switch to a formulation of a for loop that is much more obviously
not corrupting the mask by mutating it more times than intended. No
functionality changed (the results were non-overlapping so the old
version "worked" but was non-obvious).
[dfsan] Try not to create too many additional basic blocks in functions which
already have a large number of blocks. Works around a performance issue with
the greedy register allocator.
Matt Arsenault [Wed, 6 Aug 2014 00:29:49 +0000 (00:29 +0000)]
R600: Increase nearby load scheduling threshold.
This partially fixes weird looking load scheduling
in memcpy test. The load clustering doesn't seem
particularly smart, but this method seems to be partially
deprecated so it might not be worth trying to fix.
Matt Arsenault [Wed, 6 Aug 2014 00:29:43 +0000 (00:29 +0000)]
R600/SI: Implement areLoadsFromSameBasePtr
This currently has a noticable effect on the kernel argument loads.
LDS and global loads are more problematic, I think because of how copies
are currently inserted to ensure that the address is a VGPR.
David Blaikie [Wed, 6 Aug 2014 00:21:25 +0000 (00:21 +0000)]
DebugInfo: Assert that any CU for which debug_loc lists are emitted, has at least one range.
This was coming in weird debug info that had variables (and hence
debug_locs) but was in GMLT mode (because it was missing the 13th field
of the compile_unit metadata) so no ranges were constructed. We should
always have at least one range for any CU with a debug_loc in it -
because the range should cover the debug_loc.
The assertion just ensures that the "!= 1" range case inside the
subsequent loop doesn't get entered for the case where there are no
ranges at all, which should never reach here in the first place.
David Blaikie [Tue, 5 Aug 2014 23:57:31 +0000 (23:57 +0000)]
DebugInfo: Fix a bunch of tests that, owing to their compile_unit metadata not including a 13th field, had some subtle behavior.
Without the 13th field, the "emission kind" field defaults to 0 (which
is not equal to either of the values of the emission kind enum (1 ==
full debug info, 2 == line tables only)).
In this particular instance, the comparison with "FullDebugInfo" was
done when adding elements to the ranges list - so for these test cases
no values were added to the ranges list.
This got weirder when emitting debug_loc entries as the addresses should
be relative to the range of the CU if the CU has only one range (the
reasonable assumption is that if we're emitting debug_loc lists for a CU
that CU has at least one range - but due to the above situation, it has
zero) so the ranges were emitted relative to the start of the section
rather than relative to the start of the CU's singular range.
Fix these tests by accounting for the difference in the description of
debug_loc entries (in some cases making the test ignorant to these
differences, in others adding the extra label difference expression,
etc) or the presence/absence of high/low_pc on the CU, and add the 13th
field to their CUs to enable proper "full debug info" emission here.
In a future commit I'll fix up a bunch of other test cases that are not
so rigorously depending on this behavior, but still doing similarly
weird things due to the missing 13th field.
David Blaikie [Tue, 5 Aug 2014 23:14:16 +0000 (23:14 +0000)]
DebugInfo: Move the reference to the CU from the location list entry to the list itself, since it is constant across an entire list.
This simplifies construction and usage while making the data structure
smaller. It was a holdover from the days when we didn't have a separate
DebugLocList and all we had was a flat list of DebugLocEntries.
Re-apply r214881: Fix return sequence on armv4 thumb
This reverts r214893, re-applying r214881 with the test case relaxed a bit to
satiate the build bots.
POP on armv4t cannot be used to change thumb state (unilke later non-m-class
architectures), therefore we need a different return sequence that uses 'bx'
instead:
POP {r3}
ADD sp, #offset
BX r3
This patch also fixes an issue where the return value in r3 would get clobbered
for functions that return 128 bits of data. In that case, we generate this
sequence instead:
MOV ip, r3
POP {r3}
ADD sp, #offset
MOV lr, r3
MOV r3, ip
BX lr
Lang Hames [Tue, 5 Aug 2014 20:51:46 +0000 (20:51 +0000)]
[MCJIT] Make llvm-rtdyld check RuntimeDyld's error state when running in -verify
mode.
This will cause -verify mode to report failure when RuntimeDyld encounters an
internal error (e.g. overflows in relocation computations). Previously we had
let these errors slip past unreported.
Bill Schmidt [Tue, 5 Aug 2014 20:47:25 +0000 (20:47 +0000)]
[PowerPC] Swap arguments and adjust shift count for vsldoi on little endian
Commits r213915 and r214718 fix recognition of shuffle masks for vmrg*
and vpku*um instructions for a little-endian target, by swapping the
input arguments. The vsldoi instruction requires similar treatment,
and also needs its shift count adjusted for little endian.
Reviewed by Ulrich Weigand.
This is a bug fix candidate for release 3.5 (and hopefully the last of
those for PowerPC).
This is mostly a cleanup, but it changes a fairly old behavior.
Every "real" LTO user was already disabling the silly internalize pass
and creating the internalize pass itself. The difference with this
patch is for "opt -std-link-opts" and the C api.
Now to get a usable behavior out of opt one doesn't need the funny
looking command line:
[x86] Fix a crasher due to shuffles which cancel each other out and add
a test case.
We also miscompile this test case which is showing a serious flaw in the
single-input v8i16 shuffle code. I've left the specific instruction
checks FIXME-ed out until I can address the bug in the single-input
code, but I wanted to separate out a significant functionality change to
produce correct code from a very simple and targeted crasher fix.
The miscompile problem stems from keeping track of inputs by value
rather than by index. As a consequence of doing this, we can't reliably
update those inputs because they might swap and we can't detect this
without copying the mask.
The blend code now uses indices for the input lists and this seems
strictly better. It also should make it easier to sort things and do
other cleanups. I think the time has come to simplify The Great Lambda
here.
BitcodeReader: Fix non-determinism in use-list order
`BasicBlockFwdRefs` (and `BlockAddrFwdRefs` before it) was being emptied
in a non-deterministic order. When predicting use-list order I've
worked around this another way, but even when parsing lazily (and we
can't recreate use-list order) use-lists should be deterministic.
Make them so by using a side-queue of functions with forward-referenced
blocks that gets visited in order.
Philip Reames [Tue, 5 Aug 2014 17:48:20 +0000 (17:48 +0000)]
Remove dead zero store to calloc initialized memory
Optimize the following IR:
%1 = tail call noalias i8* @calloc(i64 1, i64 4)
%2 = bitcast i8* %1 to i32*
; This store is dead and should be removed
store i32 0, i32* %2, align 4
Memory returned by calloc is guaranteed to be zero initialized. If the value being stored is the constant zero (and the store is not otherwise observable across threads), we can delete the store. If the store is to an out of bounds address, it is undefined and thus also removable.
Sanjay Patel [Tue, 5 Aug 2014 17:35:22 +0000 (17:35 +0000)]
Optimize vector fabs of bitcasted constant integer values.
Allow vector fabs operations on bitcasted constant integer values to be optimized
in the same way that we already optimize scalar fabs.
So for code like this:
%bitcast = bitcast i64 18446744069414584320 to <2 x float> ; 0xFFFF_FFFF_0000_0000
%fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %bitcast)
%ret = bitcast <2 x float> %fabs to i64
Instead of generating something like this:
movabsq (constant pool loadi of mask for sign bits)
vmovq (move from integer register to vector/fp register)
vandps (mask off sign bits)
vmovq (move vector/fp register back to integer return register)
We should generate:
mov (put constant value in return register)
I have also removed a redundant clause in the first 'if' statement:
N0.getOperand(0).getValueType().isInteger()
is the same thing as:
IntVT.isInteger()
Testcases for x86 and ARM added to existing files that deal with vector fabs.
One existing testcase for x86 removed because it is no longer ideal.
For more background, please see:
http://reviews.llvm.org/D4770
Adam Nemet [Tue, 5 Aug 2014 17:23:04 +0000 (17:23 +0000)]
[AVX512] Add masking variant and intrinsics for valignd/q
This is similar to what I did with the two-source permutation recently. (It's
almost too similar so that we should consider generating the masking variants
with some tablegen help.)
Both encoding and intrinsic tests are added as well. For the latter, this is
what the IR that the intrinsic test on the clang side generates.
Adam Nemet [Tue, 5 Aug 2014 17:22:59 +0000 (17:22 +0000)]
[X86] Add lowering to VALIGN
This was currently part of lowering to PALIGNR with some special-casing to
make interlane shifting work. Since AVX512F has interlane alignr (valignd/q)
and AVX512BW has vpalignr we need to support both of these *at the same time*,
e.g. for SKX.
This patch breaks out the common code and then add support to check both of
these lowering options from LowerVECTOR_SHUFFLE.
I also added some FIXMEs where I think the AVX512BW and AVX512VL additions
should probably go.
Adam Nemet [Tue, 5 Aug 2014 17:22:55 +0000 (17:22 +0000)]
[X86] Separate DAG node for valign and palignr
They have different semantics (valign is interlane while palingr is intralane)
and palingr is still needed even in the AVX512 context. According to the
latest spec AVX512BW provides these.
POP on armv4t cannot be used to change thumb state (unilke later non-m-class
architectures), therefore we need a different return sequence that uses 'bx'
instead:
POP {r3}
ADD sp, #offset
BX r3
This patch also fixes an issue where the return value in r3 would get clobbered
for functions that return 128 bits of data. In that case, we generate this
sequence instead:
MOV ip, r3
POP {r3}
ADD sp, #offset
MOV lr, r3
MOV r3, ip
BX lr
David Blaikie [Tue, 5 Aug 2014 16:20:25 +0000 (16:20 +0000)]
Improve test for merged global debug info by using llvm-dwarfdump.
It's a bit of a tradeoff, since llvm-dwarfdump doesn't print the name of
the global symbol being used as an address in the addressing mode, but
this avoids the dependence on hardcoded set labels that keep changing
(5+ commits over the last few years that each update the set label as it
changes due to other, unrelated differences in output). This could've,
instead, been changed to match the set name then match the name in the
string pool but that would present other issues (needing to skip over
the sets that weren't of interest, etc) and checking that the addresses
(granted, without relocations applied - so it's not the whole story)
match in the two variable location descriptions seems sufficient and
fairly stable here.
There are a few similar other tests with similar label dependence that
I'll update soonish.
James Molloy [Tue, 5 Aug 2014 12:30:34 +0000 (12:30 +0000)]
Teach the SLP Vectorizer that keeping some values live over a callsite can have a cost.
Some types, such as 128-bit vector types on AArch64, don't have any callee-saved registers. So if a value needs to stay live over a callsite, it must be spilled and refilled. This cost is now taken into account.
[x86] Fix a crash and wrong-code bug in the new vector lowering all
found by a single test reduced out of a failure on llvm-stress.
The start of the problem (and the crash) came when we tried to use
a find of a non-used slot in the move-to half of the move-mask as the
target for two bad-half inputs. While if lucky this will be the first of
a pair of slots which we can place the bad-half inputs into, it isn't
actually guaranteed. This really isn't surprising, not sure what I was
thinking. The correct way to find the two unused slots is to look for
one of the *used* slots. We know it isn't that pair, and we can use some
modular arithmetic to find the other pair by masking off the odd bit and
adding 2 modulo 4. With this, we reliably found a viable pair of slots
for the bad-half inputs.
Sadly, that wasn't enough. We also had a wrong code bug that surfaced
when I reduced the test case for this where we would use the same slot
twice for the two bad inputs. This is because both of the bad inputs
could be in odd slots originally and thus the mod-2 mapping would
actually be the same. The whole point of the weird indexing into the
pair of empty slots was to try to leverage when the end result needed
the two bad-half inputs to be paired in a dword and pre-pair them in the
correct orrientation. This is less important with the powerful combining
we're now doing, and also easier and more reliable to achieve be noting
that we add the bad-half inputs in order. Thus, if they are in a dword
pair, the low part of that will be the first input in the sequence.
Always putting that in the low element will just do the right thing in
addition to computing the correct result.
[FastIsel][AArch64] Fix previous commit r214844 (Don't perform sign-/zero-extension for function arguments that have already been sign-/zero-extended.)
The original code would fail for unsupported value types like i1, i8, and i16.
This fix changes the code to only create a sub-register copy for i64 value types
and all other types (i1/i8/i16/i32) just use the source register without any
modifications.
getRegClassFor() is now guarded by the i64 value type check, that guarantees
that we always request a register for a valid value type.
[FastISel][AArch64] Implement the FastLowerArguments hook.
This implements basic argument lowering for AArch64 in FastISel. It only
handles a small subset of the C calling convention. It supports simple
arguments that can be passed in GPR and FPR registers.
This should cover most of the trivial cases without falling back to
SelectionDAG.
Have MachineFunction cache a pointer to the subtarget to make lookups
shorter/easier and have the DAG use that to do the same lookup. This
can be used in the future for TargetMachine based caching lookups from
the MachineFunction easily.
Update the MIPS subtarget switching machinery to update this pointer
at the same time it runs.
MachineCombiner Pass for selecting faster instruction
sequence on AArch64
Re-commit of r214669 without changes to test cases
LLVM::CodeGen/AArch64/arm64-neon-mul-div.ll and
LLVM:: CodeGen/AArch64/dp-3source.ll
This resolves the reported compfails of the original commit.
Renato Golin [Mon, 4 Aug 2014 23:21:56 +0000 (23:21 +0000)]
Allow CP10/CP11 operations on ARMv5/v6
Those registers are VFP/NEON and vector instructions should be used instead,
but old cores rely on those co-processors to enable VFP unwinding. This change
was prompted by the libc++abi's unwinding routine and is also present in many
legacy low-level bare-metal code that we ought to compile/assemble.
Fixing bug PR20025 and allowing PR20529 to proceed with a fix in libc++abi.
Bill Schmidt [Mon, 4 Aug 2014 23:21:01 +0000 (23:21 +0000)]
[PPC64LE] Fix wrong IR for vec_sld and vec_vsldoi
My original LE implementation of the vsldoi instruction, with its
altivec.h interfaces vec_sld and vec_vsldoi, produces incorrect
shufflevector operations in the LLVM IR. Correct code is generated
because the back end handles the incorrect shufflevector in a
consistent manner.
This patch and a companion patch for Clang correct this problem by
removing the fixup from altivec.h and the corresponding fixup from the
PowerPC back end. Several test cases are also modified to reflect the
now-correct LLVM IR.
[FastISel][AArch64] Fix shift lowering for i8 and i16 value types.
This fix changes the parameters #r and #s that are passed to the UBFM/SBFM
instruction to get the zero/sign-extension for free.
The original problem was that the shift left would use the 32-bit shift even for
i8/i16 value types, which could leave the upper bits set with "garbage" values.
The arithmetic shift right on the other side would use the wrong MSB as sign-bit
to determine what bits to shift into the value.
Justin Bogner [Mon, 4 Aug 2014 21:49:15 +0000 (21:49 +0000)]
IR: Fix up doxygen comment for LLVMContext::diagnose
This comment was referring to the DiagnosticSeverity with RS_
prefixes, but they're actually DS_. I've also modernized the comment
style since I was changing it anyway.
[SDAG] Fix a really, really terrible bug in the DAG combiner.
This code is completely wrong. It is also dead, as if it were to *ever*
run, it would crash. Fortunately, after my work to the combiner, it is
at least *possible* to reach the code, and llvm-stress has found a test
case. Thanks to Patrick for reporting.
It would be really good if anyone who remembers how this code works and
what it was intended to do could add some more obvious test coverage
instead of my completely contrived and reduced test case. My test case
was so brittle I left a bread crumb comment in it to help the next
person to stumble on it and not know what it was actually testing for.
Reimplement the temporary non-const getSubtargetImpl routine so
that we can avoid implementing it on every target. Thanks to Richard
Smith for the suggestions!
Reid Kleckner [Mon, 4 Aug 2014 21:05:27 +0000 (21:05 +0000)]
Fix failure to invoke exception handler on Win64
When the last instruction prior to a function epilogue is a call, we
need to emit a nop so that the return address is not in the epilogue IP
range. This is consistent with MSVC's behavior, and may be a workaround
for a bug in the Win64 unwinder.
Add a sentence that all entries should include an email address.
Add one for Greg Clayton, Peter Collingbourne, Tobias Grosser and
Jakob Olesen based on recent commits.
David Blaikie [Mon, 4 Aug 2014 19:30:08 +0000 (19:30 +0000)]
Reapply "DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself."
Originally reverted in r213432 with flakey failures on an ASan self-host
build. After reduction it seems to be the same issue fixed in r213805
(ArgPromo + DebugInfo: Handle updating debug info over multiple
applications of argument promotion) and r213952 (by having
LiveDebugVariables strip dbg_value intrinsics in functions that are not
described by debug info). Though I cannot explain why this failure was
flakey...
Move the R600 intrinsic support back to the target machine - there's
nothing subtarget dependent about the intrinsic support in any
backend as far as I can tell.
Justin Bogner [Mon, 4 Aug 2014 17:36:41 +0000 (17:36 +0000)]
Path: Stop claiming path::const_iterator is bidirectional
path::const_iterator claims that it's a bidirectional iterator, but it
doesn't satisfy all of the contracts for a bidirectional iterator.
For example, n3376 24.2.5 p6 says "If a and b are both dereferenceable,
then a == b if and only if *a and *b are bound to the same object",
but this doesn't work with how we stash and recreate Components.
This means that our use of reverse_iterator on this type is invalid
and leads to many of the valgrind errors we're hitting, as explained
by Tilmann Scheller here:
Instead, we admit that path::const_iterator is only an input_iterator,
and implement a second input_iterator for path::reverse_iterator (by
changing const_iterator::operator-- to reverse_iterator::operator++).
All of the uses of this just traverse once over the path in one
direction or the other anyway.
Add a dummy subtarget to the CPP backend target machine. This will
allow us to forward all of the standard TargetMachine calls to the
subtarget and still return null as we were before.
projects / llvm / log