This allows forming more 'not' ops, so we get improvements for ISAs that have and-not.
Follow-up to:
https://reviews.llvm.org/rL300725
llvm-svn: 300763
Check the legality of ISD::[US]MULO to see whether
Intrinsic::[us]mul_with_overflow will legalize into a function call (and, thus,
will use the CTR register). Fixes PR32485.
Patch by Tim Neumann!
Differential Revision: https://reviews.llvm.org/D31790
llvm-svn: 299910
The new codepath has been in the tree for years, and there isn't any
reason to use two codepaths here.
Differential Revision: https://reviews.llvm.org/D30596
llvm-svn: 299723
This is a generic combine enabled via target hook to reduce icmp logic as discussed in:
https://bugs.llvm.org/show_bug.cgi?id=32401
It's likely that other targets will want to enable this hook for scalar transforms,
and there are probably other patterns that can use bitwise logic to reduce comparisons.
Note that we are missing an IR canonicalization for these patterns, and we will probably
prefer the pair-of-compares form in IR (shorter, more likely to fold).
Differential Revision: https://reviews.llvm.org/D31483
llvm-svn: 299542
The code already allowed vector types in via "isInteger" (which might want
a more specific name), so use splat-friendly constant predicates to match
those types.
llvm-svn: 299304
(and (setlt X, 0), (setlt Y, 0)) --> (setlt (and X, Y), 0)
We have 7 similar folds, but this one got away. The fact that the
x86 test with a branch didn't change is probably a separate bug. We
may also be missing this and the related folds in instcombine.
llvm-svn: 299252
These are the same tests added for x86 with r299238,
but PPC doesn't specify all branches as cheap, so we
see different patterns in tests with branches.
llvm-svn: 299244
Now alternatively to the TargetOption.AllowFPOpFusion global flag, FMUL->FADD
can also use the per operation FMF to allow fusion.
The idea here is not to port everything to the new scheme (e.g. fused
multiply-and-sub will be ported later) but that this work all the way from
clang.
The transformation is conditionalized on *both* the FADD and the FMUL having
the FMF contract flag.
Differential Revision: https://reviews.llvm.org/D31169
llvm-svn: 299096
In PPCBoolRetToInt bool value is changed to i32 type. On ppc64 it may introduce an extra zero extension for the return value. This patch changes the integer type to i64 to avoid the zero extension on ppc64.
This patch fixed PR32442.
Differential Revision: https://reviews.llvm.org/D31407
llvm-svn: 298955
Summary:
For the following CFG:
A->B
B->C
A->C
If there is another edge B->D, then ABC should not be considered as triangle.
Reviewers: davidxl, iteratee
Reviewed By: iteratee
Subscribers: nemanjai, llvm-commits
Differential Revision: https://reviews.llvm.org/D31310
llvm-svn: 298661
Summary: Add tests for all atomic operations for powerpc64le, so that all changes can be easily examined.
Reviewers: kbarton, hfinkel, echristo
Subscribers: mehdi_amini, nemanjai, llvm-commits
Differential Revision: https://reviews.llvm.org/D31285
llvm-svn: 298614
I had ajusted the test case before when testing a chain of length 2, and then
reverted it with rL296845 when I switched to 3 triangles. After running
benchmarks and examining generated code at length 2 I forgot to put the test
back.
llvm-svn: 298000
mfvrd and mffprd are both alias to mfvrsd.
This patch enables correct parsing of the aliases, but we still emit a mfvrsd.
Committing on behalf of brunoalr (Bruno Rosa).
Differential Revision: https://reviews.llvm.org/D29177
llvm-svn: 297849
Recommiting with compiler time improvements
Recommitting after fixup of 32-bit aliasing sign offset bug in DAGCombiner.
* Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search and chain alias analysis 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.
When 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 the output
Codegen (save perhaps for some ARM cases where we correctly constructs
wider loads, but then promotes them to float operations which appear
but requires more expensive constant generation).
Some minor peephole optimizations to deal with improved SubDAG shapes (listed below)
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies 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 seems sufficient to not cause regressions in
tests.
5. Remove Chain dependencies of Memory operations on CopyfromReg
nodes as these are captured by data dependence
6. Forward loads-store values through tokenfactors containing
{CopyToReg,CopyFromReg} Values.
7. Peephole to convert buildvector of extract_vector_elt to
extract_subvector if possible (see
CodeGen/AArch64/store-merge.ll)
8. Store merging for the ARM target is restricted to 32-bit as
some in some contexts invalid 64-bit operations are being
generated. This can be removed once appropriate checks are
added.
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, improving load-store forwarding. One test in
particular is worth noting:
CodeGen/PowerPC/ppc64-align-long-double.ll - Improved load-store
forwarding converts a load-store pair into a parallel store and
a memory-realized bitcast of the same value. However, because we
lose the sharing of the explicit and implicit store values we
must create another local store. A similar transformation
happens before SelectionDAG as well.
Reviewers: arsenm, hfinkel, tstellarAMD, jyknight, nhaehnle
llvm-svn: 297695
After inspection, it's an UB in our code base. Someone cast a var-arg
function pointer to a non-var-arg one. :/
Re-commit r296771 to continue testing on the patch.
Sorry for the trouble!
llvm-svn: 297256
This reverts commit r296771.
We found some wide spread test failures internally. I'm working on a
testcase. Politely revert the patch in the mean time. :)
llvm-svn: 297124
select Cond, C +/- 1, C --> add(ext Cond), C -- with a target hook.
This is part of the ongoing process to obsolete D24480. The motivation is to
canonicalize to select IR in InstCombine whenever possible, so we need to have a way to
undo that easily in codegen.
PowerPC is an obvious winner for this kind of transform because it has fast and complete
bit-twiddling abilities but generally lousy conditional execution perf (although this might
have changed in recent implementations).
x86 also sees some wins, but the effect is limited because these transforms already mostly
exist in its target-specific combineSelectOfTwoConstants(). The fact that we see any x86
changes just shows that that code is a mess of special-case holes. We may be able to remove
some of that logic now.
My guess is that other targets will want to enable this hook for most cases. The likely
follow-ups would be to add value type and/or the constants themselves as parameters for the
hook. As the tests in select_const.ll show, we can transform any select-of-constants to
math/logic, but the general transform for any 2 constants needs one more instruction
(multiply or 'and').
ARM is one target that I think may not want this for most cases. I see infinite loops there
because it wants to use selects to enable conditionally executed instructions.
Differential Revision: https://reviews.llvm.org/D30537
llvm-svn: 296977
This patch causes compile times for some patterns to explode. I have
a (large, unreduced) test case that slows down by more than 20x and
several test cases slow down by 2x. I'm sending some of the test cases
directly to Nirav and following up with more details in the review log,
but this should unblock anyone else hitting this.
llvm-svn: 296862
For chains of triangles with small join blocks that can be tail duplicated, a
simple calculation of probabilities is insufficient. Tail duplication
can be profitable in 3 different ways for these cases:
1) The post-dominators marked 50% are actually taken 56% (This shrinks with
longer chains)
2) The chains are statically correlated. Branch probabilities have a very
U-shaped distribution.
[http://nrs.harvard.edu/urn-3:HUL.InstRepos:24015805]
If the branches in a chain are likely to be from the same side of the
distribution as their predecessor, but are independent at runtime, this
transformation is profitable. (Because the cost of being wrong is a small
fixed cost, unlike the standard triangle layout where the cost of being
wrong scales with the # of triangles.)
3) The chains are dynamically correlated. If the probability that a previous
branch was taken positively influences whether the next branch will be
taken
We believe that 2 and 3 are common enough to justify the small margin in 1.
The code pre-scans a function's CFG to identify this pattern and marks the edges
so that the standard layout algorithm can use the computed results.
llvm-svn: 296845
This patch fixes pr32063.
Current code in PPCTargetLowering::PerformDAGCombine can transform
bswap
store
into a single PPCISD::STBRX instruction. but it doesn't consider the case that the operand size of bswap may be larger than store size. When it occurs, we need 2 modifications,
1 For the last operand of PPCISD::STBRX, we should not use DAG.getValueType(N->getOperand(1).getValueType()), instead we should use cast<StoreSDNode>(N)->getMemoryVT().
2 Before PPCISD::STBRX, we need to shift the original operand of bswap to the right side.
Differential Revision: https://reviews.llvm.org/D30362
llvm-svn: 296811
This patch reduces the stack frame size by not allocating the parameter area if
it is not required. In the current implementation LowerFormalArguments_64SVR4
already handles the parameter area, but LowerCall_64SVR4 does not
(when calculating the stack frame size). What this patch does is make
LowerCall_64SVR4 consistent with LowerFormalArguments_64SVR4.
Committing on behalf of Hiroshi Inoue.
Differential Revision: https://reviews.llvm.org/D29881
llvm-svn: 296771
This is part of the ongoing attempt to improve select codegen for all targets and select
canonicalization in IR (see D24480 for more background). The transform is a subset of what
is done in InstCombine's FoldOpIntoSelect().
I first noticed a regression in the x86 avx512-insert-extract.ll tests with a patch that
hopes to convert more selects to basic math ops. This appears to be a general missing DAG
transform though, so I added tests for all standard binops in rL296621
(PowerPC was chosen semi-randomly; it has scripted FileCheck support, but so do ARM and x86).
The poor output for "sel_constants_shl_constant" is tracked with:
https://bugs.llvm.org/show_bug.cgi?id=32105
Differential Revision: https://reviews.llvm.org/D30502
llvm-svn: 296699
This patch adds a MachineSSA pass that coalesces blocks that branch
on the same condition.
Committing on behalf of Lei Huang.
Differential Revision: https://reviews.llvm.org/D28249
llvm-svn: 296670
Resubmit r295336 after the bug with non-zero offset patterns on BE targets is fixed (r296336).
Support {a|s}ext, {a|z|s}ext load nodes as a part of load combine patters.
Reviewed By: filcab
Differential Revision: https://reviews.llvm.org/D29591
llvm-svn: 296651
Recommiting after fixup of 32-bit aliasing sign offset bug in DAGCombiner.
* Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search and chain alias analysis 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.
When 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 the output
Codegen (save perhaps for some ARM cases where we correctly constructs
wider loads, but then promotes them to float operations which appear
but requires more expensive constant generation).
Some minor peephole optimizations to deal with improved SubDAG shapes (listed below)
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies 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 seems sufficient to not cause regressions in
tests.
5. Remove Chain dependencies of Memory operations on CopyfromReg
nodes as these are captured by data dependence
6. Forward loads-store values through tokenfactors containing
{CopyToReg,CopyFromReg} Values.
7. Peephole to convert buildvector of extract_vector_elt to
extract_subvector if possible (see
CodeGen/AArch64/store-merge.ll)
8. Store merging for the ARM target is restricted to 32-bit as
some in some contexts invalid 64-bit operations are being
generated. This can be removed once appropriate checks are
added.
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, improving load-store forwarding. One test in
particular is worth noting:
CodeGen/PowerPC/ppc64-align-long-double.ll - Improved load-store
forwarding converts a load-store pair into a parallel store and
a memory-realized bitcast of the same value. However, because we
lose the sharing of the explicit and implicit store values we
must create another local store. A similar transformation
happens before SelectionDAG as well.
Reviewers: arsenm, hfinkel, tstellarAMD, jyknight, nhaehnle
llvm-svn: 296476
Splitting critical edges when one of the source edges is an indirectbr
is hard in general (because it requires changing the memory the indirectbr
reads). But if a block only has a single indirectbr predecessor (which is
the common case), we can simulate splitting that edge by splitting
the destination block, and retargeting the *direct* branches.
This is motivated by the use of computed gotos in python 2.7: PyEval_EvalFrame()
ends up using an indirect branch with ~100 successors, and passing a constant to
each of those. Since MachineSink can't break indirect critical edges on demand
(and doing this in MIR doesn't look feasible), this causes us to emit about ~100
defs of registers containing constants, which we in the predecessor block, where
only one of those constants is used in each successor. So, at each computed goto,
we needlessly spill about a 100 constants to stack. The end result is that a
clang-compiled python interpreter can be about ~2.5x slower on a simple python
reduction loop than a gcc-compiled interpreter.
Differential Revision: https://reviews.llvm.org/D29916
llvm-svn: 296416
Recommiting after fixup of 32-bit aliasing sign offset bug in DAGCombiner.
* Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search and chain alias analysis 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.
When 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 the output
Codegen (save perhaps for some ARM cases where we correctly constructs
wider loads, but then promotes them to float operations which appear
but requires more expensive constant generation).
Some minor peephole optimizations to deal with improved SubDAG shapes (listed below)
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies 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 seems sufficient to not cause regressions in
tests.
5. Remove Chain dependencies of Memory operations on CopyfromReg
nodes as these are captured by data dependence
6. Forward loads-store values through tokenfactors containing
{CopyToReg,CopyFromReg} Values.
7. Peephole to convert buildvector of extract_vector_elt to
extract_subvector if possible (see
CodeGen/AArch64/store-merge.ll)
8. Store merging for the ARM target is restricted to 32-bit as
some in some contexts invalid 64-bit operations are being
generated. This can be removed once appropriate checks are
added.
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, improving load-store forwarding. One test in
particular is worth noting:
CodeGen/PowerPC/ppc64-align-long-double.ll - Improved load-store
forwarding converts a load-store pair into a parallel store and
a memory-realized bitcast of the same value. However, because we
lose the sharing of the explicit and implicit store values we
must create another local store. A similar transformation
happens before SelectionDAG as well.
Reviewers: arsenm, hfinkel, tstellarAMD, jyknight, nhaehnle
llvm-svn: 296252
Splitting critical edges when one of the source edges is an indirectbr
is hard in general (because it requires changing the memory the indirectbr
reads). But if a block only has a single indirectbr predecessor (which is
the common case), we can simulate splitting that edge by splitting
the destination block, and retargeting the *direct* branches.
This is motivated by the use of computed gotos in python 2.7: PyEval_EvalFrame()
ends up using an indirect branch with ~100 successors, and passing a constant to
each of those. Since MachineSink can't break indirect critical edges on demand
(and doing this in MIR doesn't look feasible), this causes us to emit about ~100
defs of registers containing constants, which we in the predecessor block, where
only one of those constants is used in each successor. So, at each computed goto,
we needlessly spill about a 100 constants to stack. The end result is that a
clang-compiled python interpreter can be about ~2.5x slower on a simple python
reduction loop than a gcc-compiled interpreter.
Differential Revision: https://reviews.llvm.org/D29916
llvm-svn: 296149
Provide a 64-bit pattern to use SUBFIC for subtracting from a 16-bit immediate.
The corresponding pattern already exists for 32-bit integers.
Committing on behalf of Hiroshi Inoue.
Differential Revision: https://reviews.llvm.org/D29387
llvm-svn: 296144
Emit clrrdi (extended mnemonic for rldicr) for AND-ing with masks that
clear bits from the right hand size.
Committing on behalf of Hiroshi Inoue.
Differential Revision: https://reviews.llvm.org/D29388
llvm-svn: 296143
The motivation for filling out these select-of-constants cases goes back to D24480,
where we discussed removing an IR fold from add(zext) --> select. And that goes back to:
https://reviews.llvm.org/rL75531https://reviews.llvm.org/rL159230
The idea is that we should always canonicalize patterns like this to a select-of-constants
in IR because that's the smallest IR and the best for value tracking. Note that we currently
do the opposite in some cases (like the cases in *this* patch). Ie, the proposed folds in
this patch already exist in InstCombine today:
https://github.com/llvm-mirror/llvm/blob/master/lib/Transforms/InstCombine/InstCombineSelect.cpp#L1151
As this patch shows, most targets generate better machine code for simple ext/add/not ops
rather than a select of constants. So the follow-up steps to make this less of a patchwork
of special-case folds and missing IR canonicalization:
1. Have DAGCombiner convert any select of constants into ext/add/not ops.
2 Have InstCombine canonicalize in the other direction (create more selects).
Differential Revision: https://reviews.llvm.org/D30180
llvm-svn: 296137
Splitting critical edges when one of the source edges is an indirectbr
is hard in general (because it requires changing the memory the indirectbr
reads). But if a block only has a single indirectbr predecessor (which is
the common case), we can simulate splitting that edge by splitting
the destination block, and retargeting the *direct* branches.
This is motivated by the use of computed gotos in python 2.7: PyEval_EvalFrame()
ends up using an indirect branch with ~100 successors, and passing a constant to
each of those. Since MachineSink can't break indirect critical edges on demand
(and doing this in MIR doesn't look feasible), this causes us to emit about ~100
defs of registers containing constants, which we in the predecessor block, where
only one of those constants is used in each successor. So, at each computed goto,
we needlessly spill about a 100 constants to stack. The end result is that a
clang-compiled python interpreter can be about ~2.5x slower on a simple python
reduction loop than a gcc-compiled interpreter.
Differential Revision: https://reviews.llvm.org/D29916
llvm-svn: 296060
Resubmit -r295314 with PowerPC and AMDGPU tests updated.
Support {a|s}ext, {a|z|s}ext load nodes as a part of load combine patters.
Reviewed By: filcab
Differential Revision: https://reviews.llvm.org/D29591
llvm-svn: 295336
Lay out trellis-shaped CFGs optimally.
A trellis of the shape below:
A B
|\ /|
| \ / |
| X |
| / \ |
|/ \|
C D
would be laid out A; B->C ; D by the current layout algorithm. Now we identify
trellises and lay them out either A->C; B->D or A->D; B->C. This scales with an
increasing number of predecessors. A trellis is a a group of 2 or more
predecessor blocks that all have the same successors.
because of this we can tail duplicate to extend existing trellises.
As an example consider the following CFG:
B D F H
/ \ / \ / \ / \
A---C---E---G---Ret
Where A,C,E,G are all small (Currently 2 instructions).
The CFG preserving layout is then A,B,C,D,E,F,G,H,Ret.
The current code will copy C into B, E into D and G into F and yield the layout
A,C,B(C),E,D(E),F(G),G,H,ret
define void @straight_test(i32 %tag) {
entry:
br label %test1
test1: ; A
%tagbit1 = and i32 %tag, 1
%tagbit1eq0 = icmp eq i32 %tagbit1, 0
br i1 %tagbit1eq0, label %test2, label %optional1
optional1: ; B
call void @a()
br label %test2
test2: ; C
%tagbit2 = and i32 %tag, 2
%tagbit2eq0 = icmp eq i32 %tagbit2, 0
br i1 %tagbit2eq0, label %test3, label %optional2
optional2: ; D
call void @b()
br label %test3
test3: ; E
%tagbit3 = and i32 %tag, 4
%tagbit3eq0 = icmp eq i32 %tagbit3, 0
br i1 %tagbit3eq0, label %test4, label %optional3
optional3: ; F
call void @c()
br label %test4
test4: ; G
%tagbit4 = and i32 %tag, 8
%tagbit4eq0 = icmp eq i32 %tagbit4, 0
br i1 %tagbit4eq0, label %exit, label %optional4
optional4: ; H
call void @d()
br label %exit
exit:
ret void
}
here is the layout after D27742:
straight_test: # @straight_test
; ... Prologue elided
; BB#0: # %entry ; A (merged with test1)
; ... More prologue elided
mr 30, 3
andi. 3, 30, 1
bc 12, 1, .LBB0_2
; BB#1: # %test2 ; C
rlwinm. 3, 30, 0, 30, 30
beq 0, .LBB0_3
b .LBB0_4
.LBB0_2: # %optional1 ; B (copy of C)
bl a
nop
rlwinm. 3, 30, 0, 30, 30
bne 0, .LBB0_4
.LBB0_3: # %test3 ; E
rlwinm. 3, 30, 0, 29, 29
beq 0, .LBB0_5
b .LBB0_6
.LBB0_4: # %optional2 ; D (copy of E)
bl b
nop
rlwinm. 3, 30, 0, 29, 29
bne 0, .LBB0_6
.LBB0_5: # %test4 ; G
rlwinm. 3, 30, 0, 28, 28
beq 0, .LBB0_8
b .LBB0_7
.LBB0_6: # %optional3 ; F (copy of G)
bl c
nop
rlwinm. 3, 30, 0, 28, 28
beq 0, .LBB0_8
.LBB0_7: # %optional4 ; H
bl d
nop
.LBB0_8: # %exit ; Ret
ld 30, 96(1) # 8-byte Folded Reload
addi 1, 1, 112
ld 0, 16(1)
mtlr 0
blr
The tail-duplication has produced some benefit, but it has also produced a
trellis which is not laid out optimally. With this patch, we improve the layouts
of such trellises, and decrease the cost calculation for tail-duplication
accordingly.
This patch produces the layout A,C,E,G,B,D,F,H,Ret. This layout does have
back edges, which is a negative, but it has a bigger compensating
positive, which is that it handles the case where there are long strings
of skipped blocks much better than the original layout. Both layouts
handle runs of executed blocks equally well. Branch prediction also
improves if there is any correlation between subsequent optional blocks.
Here is the resulting concrete layout:
straight_test: # @straight_test
; BB#0: # %entry ; A (merged with test1)
mr 30, 3
andi. 3, 30, 1
bc 12, 1, .LBB0_4
; BB#1: # %test2 ; C
rlwinm. 3, 30, 0, 30, 30
bne 0, .LBB0_5
.LBB0_2: # %test3 ; E
rlwinm. 3, 30, 0, 29, 29
bne 0, .LBB0_6
.LBB0_3: # %test4 ; G
rlwinm. 3, 30, 0, 28, 28
bne 0, .LBB0_7
b .LBB0_8
.LBB0_4: # %optional1 ; B (Copy of C)
bl a
nop
rlwinm. 3, 30, 0, 30, 30
beq 0, .LBB0_2
.LBB0_5: # %optional2 ; D (Copy of E)
bl b
nop
rlwinm. 3, 30, 0, 29, 29
beq 0, .LBB0_3
.LBB0_6: # %optional3 ; F (Copy of G)
bl c
nop
rlwinm. 3, 30, 0, 28, 28
beq 0, .LBB0_8
.LBB0_7: # %optional4 ; H
bl d
nop
.LBB0_8: # %exit
Differential Revision: https://reviews.llvm.org/D28522
llvm-svn: 295223
Sanjay Patel