Retrying after upstream changes.
Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search which only checks for parallel stores
through the chain subgraph. This is cleaner as the separation of
non-interfering loads/stores from the store-merging logic.
Whem merging stores, search up the chain through a single load, and
finds all possible stores by looking down from through a load and a
TokenFactor to all stores visited. This improves the quality of the
output SelectionDAG and generally the output CodeGen (with some
exceptions).
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies the the chain aggregation in the merged stores across
code paths
3. Re-add the Store node to the worklist after calling
SimplifyDemandedBits.
4. Increase GatherAllAliasesMaxDepth from 6 to 18. That number is
arbitrary, but seemed sufficient to not cause regressions in
tests.
This finishes the change Matt Arsenault started in r246307 and
jyknight's original patch.
Many tests required some changes as memory operations are now
reorderable. Some tests relying on the order were changed to use
volatile memory operations
Noteworthy tests:
CodeGen/AArch64/argument-blocks.ll -
It's not entirely clear what the test_varargs_stackalign test is
supposed to be asserting, but the new code looks right.
CodeGen/AArch64/arm64-memset-inline.lli -
CodeGen/AArch64/arm64-stur.ll -
CodeGen/ARM/memset-inline.ll -
The backend now generates *worse* code due to store merging
succeeding, as we do do a 16-byte constant-zero store efficiently.
CodeGen/AArch64/merge-store.ll -
Improved, but there still seems to be an extraneous vector insert
from an element to itself?
CodeGen/PowerPC/ppc64-align-long-double.ll -
Worse code emitted in this case, due to the improved store->load
forwarding.
CodeGen/X86/dag-merge-fast-accesses.ll -
CodeGen/X86/MergeConsecutiveStores.ll -
CodeGen/X86/stores-merging.ll -
CodeGen/Mips/load-store-left-right.ll -
Restored correct merging of non-aligned stores
CodeGen/AMDGPU/promote-alloca-stored-pointer-value.ll -
Improved. Correctly merges buffer_store_dword calls
CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll -
Improved. Sidesteps loading a stored value and
merges two stores
CodeGen/X86/pr18023.ll -
This test has been removed, as it was asserting incorrect
behavior. Non-volatile stores *CAN* be moved past volatile loads,
and now are.
CodeGen/X86/vector-idiv.ll -
CodeGen/X86/vector-lzcnt-128.ll -
It's basically impossible to tell what these tests are actually
testing. But, looks like the code got better due to the memory
operations being recognized as non-aliasing.
CodeGen/X86/win32-eh.ll -
Both loads of the securitycookie are now merged.
CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot-compute.ll -
This test appears to work but no longer exhibits the spill behavior.
Reviewers: arsenm, hfinkel, tstellarAMD, jyknight, nhaehnle
Subscribers: wdng, nhaehnle, nemanjai, arsenm, weimingz, niravd, RKSimon, aemerson, qcolombet, dsanders, resistor, tstellarAMD, t.p.northover, spatel
Differential Revision: https://reviews.llvm.org/D14834
llvm-svn: 284151
Reverts r283938 to reinstate r283867 with a fix.
The original change had an ArrayRef referring to a destroyed temporary
initializer list. Use plain C arrays instead.
llvm-svn: 283942
The high registers are not allocatable in Thumb1 functions, but they
could still be used by inline assembly, so we need to save and restore
the callee-saved high registers (r8-r11) in the prologue and epilogue.
This is complicated by the fact that the Thumb1 push and pop
instructions cannot access these registers. Therefore, we have to move
them down into low registers before pushing, and move them back after
popping into low registers.
In most functions, we will have low registers that are also being
pushed/popped, which we can use as the temporary registers for
saving/restoring the high registers. However, this is not guaranteed, so
we may need to push some extra low registers to ensure that the high
registers can be saved/restored. For correctness, it would be sufficient
to use just one low register, but if we have enough low registers
available then we only need one push/pop instruction, rather than one
per high register.
We can also use the argument/return registers when they are not live,
and the link register when saving (but not restoring), reducing the
number of extra registers we need to push.
There are still a few extreme edge cases where we need two push/pop
instructions, because not enough low registers can be made live in the
prologue or epilogue.
In addition to the regression tests included here, I've also tested this
using a script to generate functions which clobber different
combinations of registers, have different numbers of argument and return
registers (including variadic arguments), allocate different fixed sized
objects on the stack, and do or don't use variable sized allocas and the
__builtin_return_address intrinsic (all of which affect the available
registers in the prologue and epilogue). I ran these functions in a test
harness which verifies that all of the callee-saved registers are
correctly preserved.
Differential Revision: https://reviews.llvm.org/D24228
llvm-svn: 283867
Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search which only checks for parallel stores
through the chain subgraph. This is cleaner as the separation of
non-interfering loads/stores from the store-merging logic.
Whem merging stores, search up the chain through a single load, and
finds all possible stores by looking down from through a load and a
TokenFactor to all stores visited. This improves the quality of the
output SelectionDAG and generally the output CodeGen (with some
exceptions).
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies the the chain aggregation in the merged stores across
code paths
3. Re-add the Store node to the worklist after calling
SimplifyDemandedBits.
4. Increase GatherAllAliasesMaxDepth from 6 to 18. That number is
arbitrary, but seemed sufficient to not cause regressions in
tests.
This finishes the change Matt Arsenault started in r246307 and
jyknight's original patch.
Many tests required some changes as memory operations are now
reorderable. Some tests relying on the order were changed to use
volatile memory operations
Noteworthy tests:
CodeGen/AArch64/argument-blocks.ll -
It's not entirely clear what the test_varargs_stackalign test is
supposed to be asserting, but the new code looks right.
CodeGen/AArch64/arm64-memset-inline.lli -
CodeGen/AArch64/arm64-stur.ll -
CodeGen/ARM/memset-inline.ll -
The backend now generates *worse* code due to store merging
succeeding, as we do do a 16-byte constant-zero store efficiently.
CodeGen/AArch64/merge-store.ll -
Improved, but there still seems to be an extraneous vector insert
from an element to itself?
CodeGen/PowerPC/ppc64-align-long-double.ll -
Worse code emitted in this case, due to the improved store->load
forwarding.
CodeGen/X86/dag-merge-fast-accesses.ll -
CodeGen/X86/MergeConsecutiveStores.ll -
CodeGen/X86/stores-merging.ll -
CodeGen/Mips/load-store-left-right.ll -
Restored correct merging of non-aligned stores
CodeGen/AMDGPU/promote-alloca-stored-pointer-value.ll -
Improved. Correctly merges buffer_store_dword calls
CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll -
Improved. Sidesteps loading a stored value and merges two stores
CodeGen/X86/pr18023.ll -
This test has been removed, as it was asserting incorrect
behavior. Non-volatile stores *CAN* be moved past volatile loads,
and now are.
CodeGen/X86/vector-idiv.ll -
CodeGen/X86/vector-lzcnt-128.ll -
It's basically impossible to tell what these tests are actually
testing. But, looks like the code got better due to the memory
operations being recognized as non-aliasing.
CodeGen/X86/win32-eh.ll -
Both loads of the securitycookie are now merged.
CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot-compute.ll -
This test appears to work but no longer exhibits the spill
behavior.
Reviewers: arsenm, hfinkel, tstellarAMD, nhaehnle, jyknight
Subscribers: wdng, nhaehnle, nemanjai, arsenm, weimingz, niravd, RKSimon, aemerson, qcolombet, resistor, tstellarAMD, t.p.northover, spatel
Differential Revision: https://reviews.llvm.org/D14834
llvm-svn: 282600
For the common pattern (CMPZ (AND x, #bitmask), #0), we can do some more efficient instruction selection if the bitmask is one consecutive sequence of set bits (32 - clz(bm) - ctz(bm) == popcount(bm)).
1) If the bitmask touches the LSB, then we can remove all the upper bits and set the flags by doing one LSLS.
2) If the bitmask touches the MSB, then we can remove all the lower bits and set the flags with one LSRS.
3) If the bitmask has popcount == 1 (only one set bit), we can shift that bit into the sign bit with one LSLS and change the condition query from NE/EQ to MI/PL (we could also implement this by shifting into the carry bit and branching on BCC/BCS).
4) Otherwise, we can emit a sequence of LSLS+LSRS to remove the upper and lower zero bits of the mask.
1-3 require only one 16-bit instruction and can elide the CMP. 4 requires two 16-bit instructions but can elide the CMP and doesn't require materializing a complex immediate, so is also a win.
llvm-svn: 281323
For the common pattern (CMPZ (AND x, #bitmask), #0), we can do some more efficient instruction selection if the bitmask is one consecutive sequence of set bits (32 - clz(bm) - ctz(bm) == popcount(bm)).
1) If the bitmask touches the LSB, then we can remove all the upper bits and set the flags by doing one LSLS.
2) If the bitmask touches the MSB, then we can remove all the lower bits and set the flags with one LSRS.
3) If the bitmask has popcount == 1 (only one set bit), we can shift that bit into the sign bit with one LSLS and change the condition query from NE/EQ to MI/PL (we could also implement this by shifting into the carry bit and branching on BCC/BCS).
4) Otherwise, we can emit a sequence of LSLS+LSRS to remove the upper and lower zero bits of the mask.
1-3 require only one 16-bit instruction and can elide the CMP. 4 requires two 16-bit instructions but can elide the CMP and doesn't require materializing a complex immediate, so is also a win.
llvm-svn: 281215
The CMPZ #0 disappears during peepholing, leaving just a tADDi3, tADDi8 or t2ADDri. This avoids having to materialize the expensive negative constant in Thumb-1, and allows a shrinking from a 32-bit CMN to a 16-bit ADDS in Thumb-2.
llvm-svn: 281040
This avoids us doing a completely unneeded "cmp r0, #0" after a flag-setting instruction if we only care about the Z or C flags.
Add LSL/LSR to the whitelist while we're here and add testing. This code could really do with a spring clean.
llvm-svn: 281027
There is not an official documented ABI for frame pointers in Thumb2,
but we should try to emit something which is useful.
We use r7 as the frame pointer for Thumb code, which currently means
that if a function needs to save a high register (r8-r11), it will get
pushed to the stack between the frame pointer (r7) and link register
(r14). This means that while a stack unwinder can follow the chain of
frame pointers up the stack, it cannot know the offset to lr, so does
not know which functions correspond to the stack frames.
To fix this, we need to push the callee-saved registers in two batches,
with the first push saving the low registers, fp and lr, and the second
push saving the high registers. This is already implemented, but
previously only used for iOS. This patch turns it on for all Thumb2
targets when frame pointers are required by the ABI, and the frame
pointer is r7 (Windows uses r11, so this isn't a problem there). If
frame pointer elimination is enabled we still emit a single push/pop
even if we need a frame pointer for other reasons, to avoid increasing
code size.
We must also ensure that lr is pushed to the stack when using a frame
pointer, so that we end up with a complete frame record. Situations that
could cause this were rare, because we already push lr in most
situations so that we can return using the pop instruction.
Differential Revision: https://reviews.llvm.org/D23516
llvm-svn: 279506
The ppc64 multistage bot fails on this.
This reverts commit r279124.
Also Revert "CodeGen: Add/Factor out LiveRegUnits class; NFCI" because it depends on the previous change
This reverts commit r279171.
llvm-svn: 279199
Re-apply r276044 with off-by-1 instruction fix for the reload placement.
This is a variant of scavengeRegister() that works for
enterBasicBlockEnd()/backward(). The benefit of the backward mode is
that it is not affected by incomplete kill flags.
This patch also changes
PrologEpilogInserter::doScavengeFrameVirtualRegs() to use the register
scavenger in backwards mode.
Differential Revision: http://reviews.llvm.org/D21885
llvm-svn: 279124
Summary:
When performing cmp for EQ/NE and the operand is sign extended, we can
avoid the truncaton if the bits to be tested are no less than origianl
bits.
Reviewers: eli.friedman
Subscribers: eli.friedman, aemerson, nemanjai, t.p.northover, llvm-commits
Differential Revision: https://reviews.llvm.org/D22933
llvm-svn: 277252
Reverting this commit for now as it seems to be causing failures on
test-suite tests on the clang-ppc64le-linux-lnt bot.
This reverts commit r276044.
llvm-svn: 276068
This is a variant of scavengeRegister() that works for
enterBasicBlockEnd()/backward(). The benefit of the backward mode is
that it is not affected by incomplete kill flags.
This patch also changes
PrologEpilogInserter::doScavengeFrameVirtualRegs() to use the register
scavenger in backwards mode.
Differential Revision: http://reviews.llvm.org/D21885
llvm-svn: 276044
Thumb-1 doesn't have post-inc or pre-inc load or store instructions. However the LDM/STM instructions with writeback can function as post-inc load/store:
ldm r0!, {r1} @ load from r0 into r1 and increment r0 by 4
Obviously, this only works if the post increment is 4.
llvm-svn: 275540
The important thing I was missing was ensuring newly added constants were kept in topological order. Repositioning the node is correct if the constant is newly added (so it has no topological ordering) but wrong if it already existed - positioning it next in the worklist would break the topological ordering.
Original commit message:
[Thumb] Select a BIC instead of AND if the immediate can be encoded more optimally negated
If an immediate is only used in an AND node, it is possible that the immediate can be more optimally materialized when negated. If this is the case, we can negate the immediate and use a BIC instead;
int i(int a) {
return a & 0xfffffeec;
}
Used to produce:
ldr r1, [CONSTPOOL]
ands r0, r1
CONSTPOOL: 0xfffffeec
And now produces:
movs r1, #255
adds r1, #20 ; Less costly immediate generation
bics r0, r1
llvm-svn: 274543
We were using DAG->getConstant instead of DAG->getTargetConstant. This meant that we could inadvertently increase the use count of a constant if stars aligned, which it did in this testcase. Increasing the use count of the constant could cause ISel to fall over (because DAGToDAG lowering assumed the constant had only one use!)
Original commit message:
[Thumb] Select a BIC instead of AND if the immediate can be encoded more optimally negated
If an immediate is only used in an AND node, it is possible that the immediate can be more optimally materialized when negated. If this is the case, we can negate the immediate and use a BIC instead;
int i(int a) {
return a & 0xfffffeec;
}
Used to produce:
ldr r1, [CONSTPOOL]
ands r0, r1
CONSTPOOL: 0xfffffeec
And now produces:
movs r1, #255
adds r1, #20 ; Less costly immediate generation
bics r0, r1
llvm-svn: 274510
We can only generate immediates up to #510 with a MOV+ADD, not #511, because there's no such instruction as add #256.
Found by Oliver Stannard and csmith!
llvm-svn: 272665
If an immediate is only used in an AND node, it is possible that the immediate can be more optimally materialized when negated. If this is the case, we can negate the immediate and use a BIC instead;
int i(int a) {
return a & 0xfffffeec;
}
Used to produce:
ldr r1, [CONSTPOOL]
ands r0, r1
CONSTPOOL: 0xfffffeec
And now produces:
movs r1, #255
adds r1, #20 ; Less costly immediate generation
bics r0, r1
llvm-svn: 272251
I'm really not sure why we were in the first place, it's the linker's job to
convert between BL/BLX as necessary. Even worse, using BLX left Thumb calls
that could be locally resolved completely unencodable since all offsets to BLX
are multiples of 4.
rdar://26182344
llvm-svn: 269101
Currently each Function points to a DISubprogram and DISubprogram has a
scope field. For member functions the scope is a DICompositeType. DIScopes
point to the DICompileUnit to facilitate type uniquing.
Distinct DISubprograms (with isDefinition: true) are not part of the type
hierarchy and cannot be uniqued. This change removes the subprograms
list from DICompileUnit and instead adds a pointer to the owning compile
unit to distinct DISubprograms. This would make it easy for ThinLTO to
strip unneeded DISubprograms and their transitively referenced debug info.
Motivation
----------
Materializing DISubprograms is currently the most expensive operation when
doing a ThinLTO build of clang.
We want the DISubprogram to be stored in a separate Bitcode block (or the
same block as the function body) so we can avoid having to expensively
deserialize all DISubprograms together with the global metadata. If a
function has been inlined into another subprogram we need to store a
reference the block containing the inlined subprogram.
Attached to https://llvm.org/bugs/show_bug.cgi?id=27284 is a python script
that updates LLVM IR testcases to the new format.
http://reviews.llvm.org/D19034
<rdar://problem/25256815>
llvm-svn: 266446
This mostly cosmetic patch moves the DebugEmissionKind enum from DIBuilder
into DICompileUnit. DIBuilder is not the right place for this enum to live
in — a metadata consumer should not have to include DIBuilder.h.
I also added a Verifier check that checks that the emission kind of a
DICompileUnit is actually legal.
http://reviews.llvm.org/D18612
<rdar://problem/25427165>
llvm-svn: 265077
For historic reasons, the behavior of .align differs between targets.
Fortunately, there are alternatives, .p2align and .balign, which make the
interpretation of the parameter explicit, and which behave consistently across
targets.
This patch teaches MC to use .p2align instead of .align, so that people reading
code for multiple architectures don't have to remember which way each platform
does its .align directive.
Differential Revision: http://reviews.llvm.org/D16549
llvm-svn: 258750
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
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
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
Note, this was reviewed (and more details are in) http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
These intrinsics currently have an explicit alignment argument which is
required to be a constant integer. It represents the alignment of the
source and dest, and so must be the minimum of those.
This change allows source and dest to each have their own alignments
by using the alignment attribute on their arguments. The alignment
argument itself is removed.
There are a few places in the code for which the code needs to be
checked by an expert as to whether using only src/dest alignment is
safe. For those places, they currently take the minimum of src/dest
alignments which matches the current behaviour.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 500, i32 8, i1 false)
will now read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %dest, i8* align 8 %src, i32 500, i1 false)
For out of tree owners, I was able to strip alignment from calls using sed by replacing:
(call.*llvm\.memset.*)i32\ [0-9]*\,\ i1 false\)
with:
$1i1 false)
and similarly for memmove and memcpy.
I then added back in alignment to test cases which needed it.
A similar commit will be made to clang which actually has many differences in alignment as now
IRBuilder can generate different source/dest alignments on calls.
In IRBuilder itself, a new argument was added. Instead of calling:
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, /* isVolatile */ false)
you now call
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, SrcAlign, /* isVolatile */ false)
There is a temporary class (IntegerAlignment) which takes the source alignment and rejects
implicit conversion from bool. This is to prevent isVolatile here from passing its default
parameter to the source alignment.
Note, changes in future can now be made to codegen. I didn't change anything here, but this
change should enable better memcpy code sequences.
Reviewed by Hal Finkel.
llvm-svn: 253511
It turns out we decide whether to use SjLj exceptions or some alternative in
two separate places in the backend, and they disagreed with each other. This
led to inconsistent code and is generally a terrible idea.
So make them consistent and add an assert that they *do* match (unfortunately
MCAsmInfo isn't available in opt, so it can't be used to initialise the CodeGen
version directly).
llvm-svn: 253502
We were previously codegen'ing memcpy as regular load/store operations and
hoping that the register allocator would allocate registers in ascending order
so that we could apply an LDM/STM combine after register allocation. According
to the commit that first introduced this code (r37179), we planned to teach the
register allocator to allocate the registers in ascending order. This never got
implemented, and up to now we've been stuck with very poor codegen.
A much simpler approach for achieving better codegen is to create MEMCPY pseudo
instructions, attach scratch virtual registers to them and then, post register
allocation, expand the MEMCPYs into LDM/STM pairs using the scratch registers.
The register allocator will have picked arbitrary registers which we sort when
expanding the MEMCPY. This approach also avoids the need to repeatedly calculate
offsets which ultimately ought to be eliminated pre-RA in order to decrease
register pressure.
Fixes PR9199 and PR23768.
[This is based on Peter Collingbourne's r238473 which was reverted.]
Differential Revision: http://reviews.llvm.org/D13239
Change-Id: I727543c2e94136e0f80b8e22d5642d7b9ee5b458
Author: Peter Collingbourne <peter@pcc.me.uk>
llvm-svn: 249322
As a follow-up to r246098, require `DISubprogram` definitions
(`isDefinition: true`) to be 'distinct'. Specifically, add an assembler
check, a verifier check, and bitcode upgrading logic to combat testcase
bitrot after the `DIBuilder` change.
While working on the testcases, I realized that
test/Linker/subprogram-linkonce-weak-odr.ll isn't relevant anymore. Its
purpose was to check for a corner case in PR22792 where two subprogram
definitions match exactly and share the same metadata node. The new
verifier check, requiring that subprogram definitions are 'distinct',
precludes that possibility.
I updated almost all the IR with the following script:
git grep -l -E -e '= !DISubprogram\(.* isDefinition: true' |
grep -v test/Bitcode |
xargs sed -i '' -e 's/= \(!DISubprogram(.*, isDefinition: true\)/= distinct \1/'
Likely some variant of would work for out-of-tree testcases.
llvm-svn: 246327
Since r241097, `DIBuilder` has only created distinct `DICompileUnit`s.
The backend is liable to start relying on that (if it hasn't already),
so make uniquable `DICompileUnit`s illegal and automatically upgrade old
bitcode. This is a nice cleanup, since we can remove an unnecessary
`DenseSet` (and the associated uniquing info) from `LLVMContextImpl`.
Almost all the testcases were updated with this script:
git grep -e '= !DICompileUnit' -l -- test |
grep -v test/Bitcode |
xargs sed -i '' -e 's,= !DICompileUnit,= distinct !DICompileUnit,'
I imagine something similar should work for out-of-tree testcases.
llvm-svn: 243885
Remove the fake `DW_TAG_auto_variable` and `DW_TAG_arg_variable` tags,
using `DW_TAG_variable` in their place Stop exposing the `tag:` field at
all in the assembly format for `DILocalVariable`.
Most of the testcase updates were generated by the following sed script:
find test/ -name "*.ll" -o -name "*.mir" |
xargs grep -l 'DILocalVariable' |
xargs sed -i '' \
-e 's/tag: DW_TAG_arg_variable, //' \
-e 's/tag: DW_TAG_auto_variable, //'
There were only a handful of tests in `test/Assembly` that I needed to
update by hand.
(Note: a follow-up could change `DILocalVariable::DILocalVariable()` to
set the tag to `DW_TAG_formal_parameter` instead of `DW_TAG_variable`
(as appropriate), instead of having that logic magically in the backend
in `DbgVariable`. I've added a FIXME to that effect.)
llvm-svn: 243774
This commit defines subtarget feature strict-align and uses it instead of
cl::opt -arm-strict-align to decide whether strict alignment should be
forced. Also, remove the logic that was checking the OS and architecture
as clang is now responsible for setting strict-align based on the command
line options specified and the target architecute and OS.
rdar://problem/21529937
http://reviews.llvm.org/D11470
llvm-svn: 243493
Nirav Dave