This reverts commit r179840 with a fix to test/DebugInfo/two-cus-from-same-file.ll
I'm not sure why that test only failed on ARM & MIPS and not X86 Linux, even
though the debug info was clearly invalid on all of them, but this ought to fix
it.
llvm-svn: 179996
Rather than just splitting the input type and hoping for the best, apply
a bit more cleverness. Just splitting the types until the source is
legal often leads to an illegal result time, which is then widened and a
scalarization step is introduced which leads to truly horrible code
generation. With the loop vectorizer, these sorts of operations are much
more common, and so it's worth extra effort to do them well.
Add a legalization hook for the operands of a TRUNCATE node, which will
be encountered after the result type has been legalized, but if the
operand type is still illegal. If simple splitting of both types
ends up with the result type of each half still being legal, just
do that (v16i16 -> v16i8 on ARM, for example). If, however, that would
result in an illegal result type (v8i32 -> v8i8 on ARM, for example),
we can get more clever with power-two vectors. Specifically,
split the input type, but also widen the result element size, then
concatenate the halves and truncate again. For example on ARM,
To perform a "%res = v8i8 trunc v8i32 %in" we transform to:
%inlo = v4i32 extract_subvector %in, 0
%inhi = v4i32 extract_subvector %in, 4
%lo16 = v4i16 trunc v4i32 %inlo
%hi16 = v4i16 trunc v4i32 %inhi
%in16 = v8i16 concat_vectors v4i16 %lo16, v4i16 %hi16
%res = v8i8 trunc v8i16 %in16
This allows instruction selection to generate three VMOVN instructions
instead of a sequences of moves, stores and loads.
Update the ARMTargetTransformInfo to take this improved legalization
into account.
Consider the simplified IR:
define <16 x i8> @test1(<16 x i32>* %ap) {
%a = load <16 x i32>* %ap
%tmp = trunc <16 x i32> %a to <16 x i8>
ret <16 x i8> %tmp
}
define <8 x i8> @test2(<8 x i32>* %ap) {
%a = load <8 x i32>* %ap
%tmp = trunc <8 x i32> %a to <8 x i8>
ret <8 x i8> %tmp
}
Previously, we would generate the truly hideous:
.syntax unified
.section __TEXT,__text,regular,pure_instructions
.globl _test1
.align 2
_test1: @ @test1
@ BB#0:
push {r7}
mov r7, sp
sub sp, sp, #20
bic sp, sp, #7
add r1, r0, #48
add r2, r0, #32
vld1.64 {d24, d25}, [r0:128]
vld1.64 {d16, d17}, [r1:128]
vld1.64 {d18, d19}, [r2:128]
add r1, r0, #16
vmovn.i32 d22, q8
vld1.64 {d16, d17}, [r1:128]
vmovn.i32 d20, q9
vmovn.i32 d18, q12
vmov.u16 r0, d22[3]
strb r0, [sp, #15]
vmov.u16 r0, d22[2]
strb r0, [sp, #14]
vmov.u16 r0, d22[1]
strb r0, [sp, #13]
vmov.u16 r0, d22[0]
vmovn.i32 d16, q8
strb r0, [sp, #12]
vmov.u16 r0, d20[3]
strb r0, [sp, #11]
vmov.u16 r0, d20[2]
strb r0, [sp, #10]
vmov.u16 r0, d20[1]
strb r0, [sp, #9]
vmov.u16 r0, d20[0]
strb r0, [sp, #8]
vmov.u16 r0, d18[3]
strb r0, [sp, #3]
vmov.u16 r0, d18[2]
strb r0, [sp, #2]
vmov.u16 r0, d18[1]
strb r0, [sp, #1]
vmov.u16 r0, d18[0]
strb r0, [sp]
vmov.u16 r0, d16[3]
strb r0, [sp, #7]
vmov.u16 r0, d16[2]
strb r0, [sp, #6]
vmov.u16 r0, d16[1]
strb r0, [sp, #5]
vmov.u16 r0, d16[0]
strb r0, [sp, #4]
vldmia sp, {d16, d17}
vmov r0, r1, d16
vmov r2, r3, d17
mov sp, r7
pop {r7}
bx lr
.globl _test2
.align 2
_test2: @ @test2
@ BB#0:
push {r7}
mov r7, sp
sub sp, sp, #12
bic sp, sp, #7
vld1.64 {d16, d17}, [r0:128]
add r0, r0, #16
vld1.64 {d20, d21}, [r0:128]
vmovn.i32 d18, q8
vmov.u16 r0, d18[3]
vmovn.i32 d16, q10
strb r0, [sp, #3]
vmov.u16 r0, d18[2]
strb r0, [sp, #2]
vmov.u16 r0, d18[1]
strb r0, [sp, #1]
vmov.u16 r0, d18[0]
strb r0, [sp]
vmov.u16 r0, d16[3]
strb r0, [sp, #7]
vmov.u16 r0, d16[2]
strb r0, [sp, #6]
vmov.u16 r0, d16[1]
strb r0, [sp, #5]
vmov.u16 r0, d16[0]
strb r0, [sp, #4]
ldm sp, {r0, r1}
mov sp, r7
pop {r7}
bx lr
Now, however, we generate the much more straightforward:
.syntax unified
.section __TEXT,__text,regular,pure_instructions
.globl _test1
.align 2
_test1: @ @test1
@ BB#0:
add r1, r0, #48
add r2, r0, #32
vld1.64 {d20, d21}, [r0:128]
vld1.64 {d16, d17}, [r1:128]
add r1, r0, #16
vld1.64 {d18, d19}, [r2:128]
vld1.64 {d22, d23}, [r1:128]
vmovn.i32 d17, q8
vmovn.i32 d16, q9
vmovn.i32 d18, q10
vmovn.i32 d19, q11
vmovn.i16 d17, q8
vmovn.i16 d16, q9
vmov r0, r1, d16
vmov r2, r3, d17
bx lr
.globl _test2
.align 2
_test2: @ @test2
@ BB#0:
vld1.64 {d16, d17}, [r0:128]
add r0, r0, #16
vld1.64 {d18, d19}, [r0:128]
vmovn.i32 d16, q8
vmovn.i32 d17, q9
vmovn.i16 d16, q8
vmov r0, r1, d16
bx lr
llvm-svn: 179989
This is an edge case that can happen if we modify a chain of multiple selects.
Update all operands in that case and remove the assert. PR15805.
llvm-svn: 179982
There is the temptation to make this tranform dependent on target information as
it is not going to be beneficial on all (sub)targets. Therefore, we should
probably do this in MI Early-Ifconversion.
This reverts commit r179957. Original commit message:
"SimplifyCFG: If convert single conditional stores
This transformation will transform a conditional store with a preceeding
uncondtional store to the same location:
a[i] =
may-alias with a[i] load
if (cond)
a[i] = Y
into an unconditional store.
a[i] = X
may-alias with a[i] load
tmp = cond ? Y : X;
a[i] = tmp
We assume that on average the cost of a mispredicted branch is going to be
higher than the cost of a second store to the same location, and that the
secondary benefits of creating a bigger basic block for other optimizations to
work on outway the potential case were the branch would be correctly predicted
and the cost of the executing the second store would be noticably reflected in
performance.
hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With
this change we are on par with gcc's performance (gcc also performs this
transformation). There was a 1.2 % performance improvement on a ARM swift chip.
Other tests in the test-suite+external seem to be mostly uninfluenced in my
experiments:
This optimization was triggered on 41 tests such that the executable was
different before/after the patch. Only 1 out of the 40 tests (dealII) was
reproducable below 100% (by about .4%). Given that hmmer benefits so much I
believe this to be a fair trade off.
I am going to watch performance numbers across the builtbots and will revert
this if anything unexpected comes up."
llvm-svn: 179980
Specifically:
1. Added checks that unwind is being properly added to various instructions.
2. Fixed the declaration/calling of objc_release to have a return type of void.
3. Moved all checks to precede the functions and added checks to ensure that the
checks would only match inside the specific function that we are attempting to
check.
llvm-svn: 179973
With a little help from the frontend, it looks like the standard va_*
intrinsics can do the job.
Also clean up an old bitcast hack in LowerVAARG that dealt with
unaligned double loads. Load SDNodes can specify an alignment now.
Still missing: Calling varargs functions with float arguments.
llvm-svn: 179961
This transformation will transform a conditional store with a preceeding
uncondtional store to the same location:
a[i] =
may-alias with a[i] load
if (cond)
a[i] = Y
into an unconditional store.
a[i] = X
may-alias with a[i] load
tmp = cond ? Y : X;
a[i] = tmp
We assume that on average the cost of a mispredicted branch is going to be
higher than the cost of a second store to the same location, and that the
secondary benefits of creating a bigger basic block for other optimizations to
work on outway the potential case were the branch would be correctly predicted
and the cost of the executing the second store would be noticably reflected in
performance.
hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With
this change we are on par with gcc's performance (gcc also performs this
transformation). There was a 1.2 % performance improvement on a ARM swift chip.
Other tests in the test-suite+external seem to be mostly uninfluenced in my
experiments:
This optimization was triggered on 41 tests such that the executable was
different before/after the patch. Only 1 out of the 40 tests (dealII) was
reproducable below 100% (by about .4%). Given that hmmer benefits so much I
believe this to be a fair trade off.
I am going to watch performance numbers across the builtbots and will revert
this if anything unexpected comes up.
llvm-svn: 179957
Previously, when spilling 64-bit paired registers, an LDMIA with both
a FrameIndex and an offset was produced. This kind of instruction
shouldn't exist, and the extra operand was being confused with the
predicate, causing aborts later on.
This removes the invalid 0-offset from the instruction being
produced.
llvm-svn: 179956
The logic that actually compares the types considers pointers and integers the
same if they are of the same size. This created a strange mismatch between hash
and reality and made the test case for this fail on some platforms (yay,
test cases).
llvm-svn: 179905
When matching a compare with a subtract where the arguments of the compare are
swapped w.r.t. the arguments of the subtract, we need to negate the predicates
(or CR bit indices) of the users. This, however, is not the same as inverting
the predicate (negating LT -> GT, but inverting LT -> GE, for example). The ARM
backend seems to do this correctly, but when I adapted the code for the PPC
backend, I introduced an error in this logic.
Comparison optimization is now enabled again by default.
llvm-svn: 179899
Thanks to Evgeniy Stepanov for reporting this.
It might be a good idea to add a command iterator abstraction to MachO.h, but
this fixes the bug for now.
llvm-svn: 179848
Adding another CU-wide list, in this case of imported_modules (since they
should be relatively rare, it seemed better to add a list where each element
had a "context" value, rather than add a (usually empty) list to every scope).
This takes care of DW_TAG_imported_module, but to fully address PR14606 we'll
need to expand this to cover DW_TAG_imported_declaration too.
llvm-svn: 179836
Many PPC instructions have a so-called 'record form' which stores to a specific
condition register the result of comparing the result of the instruction with
zero (always as a signed comparison). For integer operations on PPC64, this is
always a 64-bit comparison.
This implementation is derived from the implementation in the ARM backend;
there are some differences because PPC condition registers are allocatable
virtual registers (although the record forms always use a specific one), and we
look for a matching subtraction instruction after the compare (but before the
first use) in addition to before it.
llvm-svn: 179802
In X86FastISel::X86SelectStore(), improperly aligned stores are rejected and
handled by the DAG-based ISel. However, X86FastISel::X86SelectLoad() makes
no such requirement. There doesn't appear to be an x86 architectural
correctness issue with allowing potentially unaligned store instructions.
This patch removes this restriction.
Patch by Jim Stichnot.
llvm-svn: 179774
A min/max operation is represented by a select(cmp(lt/le/gt/ge, X, Y), X, Y)
sequence in LLVM. If we see such a sequence we can treat it just as any other
commutative binary instruction and reduce it.
This appears to help bzip2 by about 1.5% on an imac12,2.
radar://12960601
llvm-svn: 179773
This occurs due to an alloca representing a separate ownership from the
original pointer. Thus consider the following pseudo-IR:
objc_retain(%a)
for (...) {
objc_retain(%a)
%block <- %a
F(%block)
objc_release(%block)
}
objc_release(%a)
From the perspective of the optimizer, the %block is a separate
provenance from the original %a. Thus the optimizer pairs up the inner
retain for %a and the outer release from %a, resulting in segfaults.
This is fixed by noting that the signature of a mismatch of
retain/releases inside the for loop is a Use/CanRelease top down with an
None bottom up (since bottom up the Retain-CanRelease-Use-Release
sequence is completed by the inner objc_retain, but top down due to the
differing provenance from the objc_release said sequence is not
completed). In said case in CheckForCFGHazards, we now clear the state
of %a implying that no pairing will occur.
Additionally a test case is included.
rdar://12969722
llvm-svn: 179747
unable to handle cases such as __asm mov eax, 8*-8.
This patch also attempts to simplify the state machine. Further, the error
reporting has been improved. Test cases included, but more will be added to
the clang side shortly.
rdar://13668445
llvm-svn: 179719
for the sdiv/srem/udiv/urem bitcode instructions. This is done for the i8,
i16, and i32 types, as well as i64 for the x86_64 target.
Patch by Jim Stichnoth
llvm-svn: 179715
The reference manual defines only 5 permitted values for the immediate field of the "hint" instruction:
1. nop (imm == 0)
2. yield (imm == 1)
3. wfe (imm == 2)
4. wfi (imm == 3)
5. sev (imm == 4)
Therefore, restrict the permitted values for the "hint" instruction to 0 through 4.
Patch by Mihail Popa <Mihail.Popa@arm.com>
llvm-svn: 179707
This patch allows the Mips assembler to parse and emit nested
expressions as instruction operands. It also extends the
expansion of memory instructions when an offset is given as
an expression.
Contributer: Vladimir Medic
llvm-svn: 179657
If a switch instruction has a case for every possible value of its type,
with the same successor, SimplifyCFG would replace it with an icmp ult,
but the computation of the bound overflows in that case, which inverts
the test.
Patch by Jed Davis!
llvm-svn: 179587
Two return types are not equivalent if one is a pointer and the other is an
integral. This is because we cannot bitcast a pointer to an integral value.
PR15185
llvm-svn: 179569
This patch allows the assembler to recognize $fcc0
as a valid register for conditional move instructions.
Corresponding test cases have been added.
Contributer: Vladimir Medic
llvm-svn: 179567
Instead of emitting config values in a predefined order, the code
emitter will now emit a 32-bit register index followed by the 32-bit
config value.
llvm-svn: 179546
This fixes an ABI bug for non-Darwin PPC64. For the callee-saved condition
registers, the spill location is specified relative to the stack pointer (SP +
8). However, this is not relative to the SP after the new stack frame is
established, but instead relative to the caller's stack pointer (it is stored
into the linkage area of the parent's stack frame).
So, like with the link register, we don't directly spill the CRs with other
callee-saved registers, but just mark them to be spilled during prologue
generation.
In practice, this reverts r179457 for PPC64 (but leaves it in place for PPC32).
llvm-svn: 179500
One performs: (X == 13 | X == 14) -> X-13 <u 2
The other: (A == C1 || A == C2) -> (A & ~(C1 ^ C2)) == C1
The problem is that there are certain values of C1 and C2 that
trigger both transforms but the first one blocks out the second,
this generates suboptimal code.
Reordering the transforms should be better in every case and
allows us to do interesting stuff like turn:
%shr = lshr i32 %X, 4
%and = and i32 %shr, 15
%add = add i32 %and, -14
%tobool = icmp ne i32 %add, 0
into:
%and = and i32 %X, 240
%tobool = icmp ne i32 %and, 224
llvm-svn: 179493
This is basically the same fix in three different places. We use a set to avoid
walking the whole tree of a big ConstantExprs multiple times.
For example: (select cmp, (add big_expr 1), (add big_expr 2))
We don't want to visit big_expr twice here, it may consist of thousands of
nodes.
The testcase exercises this by creating an insanely large ConstantExprs out of
a loop. It's questionable if the optimizer should ever create those, but this
can be triggered with real C code. Fixes PR15714.
llvm-svn: 179458
For functions that need to spill CRs, and have dynamic stack allocations, the
value of the SP during the restore is not what it was during the save, and so
we need to use the FP in these cases (as for all of the other spills and
restores, but the CR restore has a special code path because its reserved slot,
like the link register, is specified directly relative to the adjusted SP).
llvm-svn: 179457
The register allocator expects minimal physreg live ranges. Schedule
physreg copies accordingly. This is slightly tricky when they occur in
the middle of the scheduling region. For now, this is handled by
rescheduling the copy when its associated instruction is
scheduled. Eventually we may instead bundle them, but only if we can
preserve the bundles as parallel copies during regalloc.
llvm-svn: 179449
According to the ARM reference manual, constant offsets are mandatory for pre-indexed addressing modes.
The MC disassembler was not obeying this when the offset is 0.
It was producing instructions like: str r0, [r1]!.
Correct syntax is: str r0, [r1, #0]!.
This change modifies the dumping of operands so that the offset is always printed, regardless of its value, when pre-indexed addressing mode is used.
Patch by Mihail Popa <Mihail.Popa@arm.com>
llvm-svn: 179398
These tests rely specifically on the names of ELF relocations, let alone any
other detail. There's no way they'd work if LLVM was emitting something else by
default.
llvm-svn: 179376
It turns out some platforms (e.g. Windows) lay out their llvm-mc slightly
differently with extra newlines; there was no real reason for the test lines to
be consecutive, so this relaxes the FileCheck.
llvm-svn: 179375
When debugging performance regressions we often ask ourselves if the regression
that we see is due to poor isel/sched/ra or due to some micro-architetural
problem. When comparing two code sequences one good way to rule out front-end
bottlenecks (and other the issues) is to force code alignment. This pass adds
a flag that forces the alignment of all of the basic blocks in the program.
llvm-svn: 179353
Original message:
Print more information about relocations.
With this patch llvm-readobj now prints if a relocation is pcrel, its length,
if it is extern and if it is scattered.
It also refactors the code a bit to use bit fields instead of shifts and
masks all over the place.
llvm-svn: 179345
With this patch llvm-readobj now prints if a relocation is pcrel, its length,
if it is extern and if it is scattered.
It also refactors the code a bit to use bit fields instead of shifts and
masks all over the place.
llvm-svn: 179294
When trying to collapse sequences of insertelement/extractelement
instructions into single shuffle instructions, there is one specific
case where the Instruction Combiner wrongly updates the resulting
Mask of shuffle indexes.
The problem is in function CollectShuffleElments.
If we have a sequence of insert/extract element instructions
like the one below:
%tmp1 = extractelement <4 x float> %LHS, i32 0
%tmp2 = insertelement <4 x float> %RHS, float %tmp1, i32 1
%tmp3 = extractelement <4 x float> %RHS, i32 2
%tmp4 = insertelement <4 x float> %tmp2, float %tmp3, i32 3
Where:
. %RHS will have a mask of [4,5,6,7]
. %LHS will have a mask of [0,1,2,3]
The Mask of shuffle indexes is wrongly computed to [4,1,6,7]
instead of [4,0,6,7].
When analyzing %tmp2 in order to compute the Mask for the
resulting shuffle instruction, the algorithm forgets to update
the mask index at position 1 with the index associated to the
element extracted from %LHS by instruction %tmp1.
Patch by Andrea DiBiagio!
llvm-svn: 179291
As packed comparisons in AVX/SSE produce all 0s or all 1s in each SIMD lane,
vector select could be simplified to AND/OR or removed if one or both values
being selected is all 0s or all 1s.
llvm-svn: 179267
As these two instructions in AVX extension are privileged instructions for
special purpose, it's only expected to be used in inlined assembly.
llvm-svn: 179266
This patch is revised based on patch from Victor Umansky
<victor.umansky@intel.com>. More cases are handled in X86's bool
simplification, i.e.
- SETCC_CARRY
- value is truncated to i1 with AND
As a by-product, PR5443 is also fixed.
llvm-svn: 179265
Add support for the COFF relocation types IMAGE_REL_I386_DIR32NB and
IMAGE_REL_AMD64_ADDR32NB for 32- and 64-bit respectively. These are
similar to normal 4-byte relocations except that they do not include
the base address of the image.
Image-relative relocations are used for debug information (32-bit) and
SEH unwind tables (64-bit).
A new MCSymbolRef variant called 'VK_COFF_IMGREL32' is introduced to
specify such relocations. For AT&T assembly, this variant can be accessed
using the symbol suffix '@imgrel'.
llvm-svn: 179240
In the simple and triangle if-conversion cases, when CopyAndPredicateBlock is
used because the to-be-predicated block has other predecessors, we need to
explicitly remove the old copied block from the successors list. Normally if
conversion relies on TII->AnalyzeBranch combined with BB->CorrectExtraCFGEdges
to cleanup the successors list, but if the predicated block contained an
un-analyzable branch (such as a now-predicated return), then this will fail.
These extra successors were causing a problem on PPC because it was causing
later passes (such as PPCEarlyReturm) to leave dead return-only basic blocks in
the code.
llvm-svn: 179227
to disable following tests for Hexagon that require direct object
generation support.
DebugInfo/dwarf-public-names.ll
DebugInfo/dwarf-version.ll
DebugInfo/member-pointers.ll
DebugInfo/namespace.ll
DebugInfo/two-cus-from-same-file.ll
Fixes bug 15616 - http://llvm.org/bugs/show_bug.cgi?id=15616
llvm-svn: 179209
Mips32 code as Mips16 unless it can't be compiled as Mips 16. For now this
would happen as long as floating point instructions are not needed.
Probably it would also make sense to compile as mips32 if atomic operations
are needed too. There may be other cases too.
A module pass prescans the IR and adds the mips16 or nomips16 attribute
to functions depending on the functions needs.
Mips 16 mode can result in a 40% code compression by utililizing 16 bit
encoding of many instructions.
The hope is for this to replace the traditional gcc way of dealing with
Mips16 code using floating point which involves essentially using soft float
but with a library implemented using mips32 floating point. This gcc
method also requires creating stubs so that Mips32 code can interact with
these Mips 16 functions that have floating point needs. My conjecture is
that in reality this traditional gcc method would never win over this
new method.
I will be implementing the traditional gcc method also. Some of it is already
done but I needed to do the stubs to finish the work and those required
this mips16/32 mixed mode capability.
I have more ideas for to make this new method much better and I think the old
method will just live in llvm for anyone that needs the backward compatibility
but I don't for what reason that would be needed.
llvm-svn: 179185
Summary:
I did a local comparison between using bash and using lit's runner, and
more of the suite passes with lit than passes with bash. Most of the
bash failures have to do with /dev/null, which is nonsensical on
Windows, but the lit runner handles it.
The lit shell runner is also much faster than bash, so I would expect
most Windows devs would want it by default.
The behavior can be overridden on any OS by setting
LIT_USE_INTERNAL_SHELL to 0 or 1 in the environment.
Reviewers: chapuni, ddunbar
CC: llvm-commits, timurrrr
Differential Revision: http://llvm-reviews.chandlerc.com/D559
llvm-svn: 179173
These instructions aren't universally available, but depend on a specific
extension to the normal ARM architecture (rather than, say, v6/v7/...) so a new
feature is appropriate.
This also enables the feature by default on A-class cores which usually have
these extensions, to avoid breaking existing code and act as a sensible
default.
llvm-svn: 179171
Depending on the number of bits set in the writemask.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 179166
Modifier 'D' is to use the second word of a double integer.
We had previously implemented the pure register varient of
the modifier and this patch implements the memory reference.
#include "stdio.h"
int b[8] = {0,1,2,3,4,5,6,7};
void main()
{
int i;
// The first word. Notice, no 'D'
{asm (
"lw %0,%1;"
: "=r" (i)
: "m" (*(b+4))
);}
printf("%d\n",i);
// The second word
{asm (
"lw %0,%D1;"
: "=r" (i)
: "m" (*(b+4))
);}
printf("%d\n",i);
}
llvm-svn: 179135
This enables us to form predicated branches (which are the same conditional
branches we had before) and also a larger set of predicated returns (including
instructions like bdnzlr which is a conditional return and loop-counter
decrement all in one).
At the moment, if conversion does not capture all possible opportunities. A
simple example is provided in early-ret2.ll, where if conversion forms one
predicated return, and then the PPCEarlyReturn pass picks up the other one. So,
at least for now, we'll keep both mechanisms.
llvm-svn: 179134
and mips16 on a per function basis.
Because this patch is somewhat involved I have provide an overview of the
key pieces of it.
The patch is written so as to not change the behavior of the non mixed
mode. We have tested this a lot but it is something new to switch subtargets
so we don't want any chance of regression in the mainline compiler until
we have more confidence in this.
Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1.
For that reason there are derived versions of the register info, frame info,
instruction info and instruction selection classes.
Now we register three separate passes for instruction selection.
One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then
one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and
MipsSEISelDAGToDAG.cpp).
When the ModuleISel pass runs, it determines if there is a need to switch
subtargets and if so, the owning pointers in MipsTargetMachine are
appropriately changed.
When 16Isel or SEIsel is run, they will return immediately without doing
any work if the current subtarget mode does not apply to them.
In addition, MipsAsmPrinter needs to be reset on a function basis.
The pass BasicTargetTransformInfo is substituted with a null pass since the
pass is immutable and really needs to be a function pass for it to be
used with changing subtargets. This will be fixed in a follow on patch.
llvm-svn: 179118
This commit adds the infrastructure for performing bottom-up SLP vectorization (and other optimizations) on parallel computations.
The infrastructure has three potential users:
1. The loop vectorizer needs to be able to vectorize AOS data structures such as (sum += A[i] + A[i+1]).
2. The BB-vectorizer needs this infrastructure for bottom-up SLP vectorization, because bottom-up vectorization is faster to compute.
3. A loop-roller needs to be able to analyze consecutive chains and roll them into a loop, in order to reduce code size. A loop roller does not need to create vector instructions, and this infrastructure separates the chain analysis from the vectorization.
This patch also includes a simple (100 LOC) bottom up SLP vectorizer that uses the infrastructure, and can vectorize this code:
void SAXPY(int *x, int *y, int a, int i) {
x[i] = a * x[i] + y[i];
x[i+1] = a * x[i+1] + y[i+1];
x[i+2] = a * x[i+2] + y[i+2];
x[i+3] = a * x[i+3] + y[i+3];
}
llvm-svn: 179117
therefore not at all) of the pc or statement list. We also don't
need to emit the compilation dir so save so space and time
and don't bother.
Fix up the testcase accordingly and verify that we don't emit
the attributes or the items that they use.
llvm-svn: 179114
This pattern occurs in SROA output due to the way vector arguments are lowered
on ARM.
The testcase from PR15525 now compiles into this, which is better than the code
we got with the old scalarrepl:
_Store:
ldr.w r9, [sp]
vmov d17, r3, r9
vmov d16, r1, r2
vst1.8 {d16, d17}, [r0]
bx lr
Differential Revision: http://llvm-reviews.chandlerc.com/D647
llvm-svn: 179106
On PowerPC, non-vector loads and stores have r+i forms; however, in functions
with large stack frames these were not being used to access slots far from the
stack pointer because such slots were out of range for the signed 16-bit
immediate offset field. This increases register pressure because we need a
separate register for each offset (when the r+r form is used). By enabling
virtual base registers, we can deal with large stack frames without unduly
increasing register pressure.
llvm-svn: 179105
Some translations here are not 1x1 because there are grep|grep
chains that are non-trivial to implement in terms of FileCheck features. I
made an effort for the tests to remain as similar as possible; do let me know
if you notice anything fishy. The good news are that some buggy tests were
fixed (grep | not grep - a bug waiting to happen).
llvm-svn: 179102
The save area is twice as big and there is no struct return slot. The
stack pointer is always 16-byte aligned (after adding the bias).
Also eliminate the stack adjustment instructions around calls when the
function has a reserved stack frame.
llvm-svn: 179083
The costs are overfitted so that I can still use the legalization factor.
For example the following kernel has about half the throughput vectorized than
unvectorized when compiled with SSE2. Before this patch we would vectorize it.
unsigned short A[1024];
double B[1024];
void f() {
int i;
for (i = 0; i < 1024; ++i) {
B[i] = (double) A[i];
}
}
radar://13599001
llvm-svn: 179033
PowerPC has a conditional branch to the link register (return) instruction: BCLR.
This should be used any time when we'd otherwise have a conditional branch to a
return. This adds a small pass, PPCEarlyReturn, which runs just prior to the
branch selection pass (and, importantly, after block placement) to generate
these conditional returns when possible. It will also eliminate unconditional
branches to returns (these happen rarely; most of the time these have already
been tail duplicated by the time PPCEarlyReturn is invoked). This is a nice
optimization for small functions that do not maintain a stack frame.
llvm-svn: 179026
I've managed to convince myself that AArch64's acquire/release
instructions are sufficient to guarantee C++11's required semantics,
even in the sequentially-consistent case.
llvm-svn: 179005
First, we should not cheat: fsel-based lowering of select_cc is a
finite-math-only optimization (the ISA manual, section F.3 of v2.06, makes
this clear, as does a note in our own README).
This also adds fsel-based lowering of EQ and NE condition codes. As it turned
out, fsel generation was covered by a grand total of zero regression test
cases. I've added some test cases to cover the existing behavior (which is now
finite-math only), as well as the new EQ cases.
llvm-svn: 179000
The code in getTypeConversion attempts to promote the element vector type
before it trys to split or widen the vector.
After it failed finding a legal vector type by promoting it would continue using
the promoted vector element type. Thereby missing legal splitted vector types.
For example the type v32i32 that has a legal split of 4 x v3i32 on x86/sse2
would be transformed to: v32i256 and from there on successively split to:
v16i256, v8i256, v1i256 and then finally ends up as an i64 type.
By resetting the vector element type to the original vector element type that
existed before the promotion the code will attempt to split the vector type to
smaller vector widths of the same type.
llvm-svn: 178999
The fix for PR14972 in r177055 introduced a real think-o in the *store*
side, likely because I was much more focused on the load side. While we
can arbitrarily widen (or narrow) a loaded value, we can't arbitrarily
widen a value to be stored, as that changes the width of memory access!
Lock down the code path in the store rewriting which would do this to
only handle the intended circumstance.
All of the existing tests continue to pass, and I've added a test from
the PR.
llvm-svn: 178974
a relocation across sections. Do this for DW_AT_stmt list in the
skeleton CU and check the relocations in the debug_info section.
Add a FIXME for multiple CUs.
llvm-svn: 178969
Integer return values are sign or zero extended by the callee, and
structs up to 32 bytes in size can be returned in registers.
The CC_Sparc64 CallingConv definition is shared between
LowerFormalArguments_64 and LowerReturn_64. Function arguments and
return values are passed in the same registers.
The inreg flag is also used for return values. This is required to handle
C functions returning structs containing floats and ints:
struct ifp {
int i;
float f;
};
struct ifp f(void);
LLVM IR:
define inreg { i32, float } @f() {
...
ret { i32, float } %retval
}
The ABI requires that %retval.i is returned in the high bits of %i0
while %retval.f goes in %f1.
Without the inreg return value attribute, %retval.i would go in %i0 and
%retval.f would go in %f3 which is a more efficient way of returning
%multiple values, but it is not ABI compliant for returning C structs.
llvm-svn: 178966
64-bit SPARC v9 processes use biased stack and frame pointers, so the
current function's stack frame is located at %sp+BIAS .. %fp+BIAS where
BIAS = 2047.
This makes more local variables directly accessible via [%fp+simm13]
addressing.
llvm-svn: 178965
There are certain PPC instructions into which we can fold a zero immediate
operand. We can detect such cases by looking at the register class required
by the using operand (so long as it is not otherwise constrained).
llvm-svn: 178961
All arguments are formally assigned to stack positions and then promoted
to floating point and integer registers. Since there are more floating
point registers than integer registers, this can cause situations where
floating point arguments are assigned to registers after integer
arguments that where assigned to the stack.
Use the inreg flag to indicate 32-bit fragments of structs containing
both float and int members.
The three-way shadowing between stack, integer, and floating point
registers requires custom argument lowering. The good news is that
return values are passed in the exact same way, and we can share the
code.
Still missing:
- Update LowerReturn to handle structs returned in registers.
- LowerCall.
- Variadic functions.
llvm-svn: 178958