This optimization is unstable at this moment; it
1) block us on a very important application
2) PR15200
3) test6 and test7 in test/Transforms/ScalarRepl/dynamic-vector-gep.ll
(the CHECK command compare the output against wrong result)
I personally believe this optimization should not have any impact on the
autovectorized code, as auto-vectorizer is supposed to put gather/scatter
in a "right" way. Although in theory downstream optimizaters might reveal
some gather/scatter optimization opportunities, the chance is quite slim.
For the hand-crafted vectorizing code, in term of redundancy elimination,
load-CSE, copy-propagation and DSE can collectively achieve the same result,
but in much simpler way. On the other hand, these optimizers are able to
improve the code in a incremental way; in contrast, SROA is sort of all-or-none
approach. However, SROA might slighly win in stack size, as it tries to figure
out a stretch of memory tightenly cover the area accessed by the dynamic index.
rdar://13174884
PR15200
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178912 91177308-0d34-0410-b5e6-96231b3b80d8
llvm-mips-linux green.
llvm-mips-linux runs on a big endian machine. This test passes if I change 'e'
to 'E' in the target data layout string.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178910 91177308-0d34-0410-b5e6-96231b3b80d8
memory operands.
Essentially, this layers an infix calculator on top of the parsing state
machine. The scale on the index register is still expected to be an immediate
__asm mov eax, [eax + ebx*4]
and will not work with more complex expressions. For example,
__asm mov eax, [eax + ebx*(2*2)]
The plus and minus binary operators assume the numeric value of a register is
zero so as to not change the displacement. Register operands should never
be an operand for a multiply or divide operation; the scale*indexreg
expression is always replaced with a zero on the operand stack to prevent
such a case.
rdar://13521380
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178881 91177308-0d34-0410-b5e6-96231b3b80d8
InMemoryStruct is extremely dangerous as it returns data from an internal
buffer when the endiannes doesn't match. This should fix the tests on big
endian hosts.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178875 91177308-0d34-0410-b5e6-96231b3b80d8
When the RuntimeDyldELF::processRelocationRef routine finds the target
symbol of a relocation in the local or global symbol table, it performs
a section-relative relocation:
Value.SectionID = lsi->second.first;
Value.Addend = lsi->second.second;
At this point, however, any Addend that might have been specified in
the original relocation record is lost. This is somewhat difficult to
trigger for relocations within the code section since they usually
do not contain non-zero Addends (when built with the default JIT code
model, in any case). However, the problem can be reliably triggered
by a relocation within the data section caused by code like:
int test[2] = { -1, 0 };
int *p = &test[1];
The initializer of "p" will need a relocation to "test + 4". On
platforms using RelA relocations this means an Addend of 4 is required.
Current code ignores this addend when processing the relocation,
resulting in incorrect execution.
Fixed by taking the Addend into account when processing relocations
to symbols found in the local or global symbol table.
Tested on x86_64-linux and powerpc64-linux.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178869 91177308-0d34-0410-b5e6-96231b3b80d8
Pass down the fact that an operand is going to be a vector of constants.
This should bring the performance of MultiSource/Benchmarks/PAQ8p/paq8p on x86
back. It had degraded to scalar performance due to my pervious shift cost change
that made all shifts expensive on x86.
radar://13576547
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178809 91177308-0d34-0410-b5e6-96231b3b80d8
SSE2 has efficient support for shifts by a scalar. My previous change of making
shifts expensive did not take this into account marking all shifts as expensive.
This would prevent vectorization from happening where it is actually beneficial.
With this change we differentiate between shifts of constants and other shifts.
radar://13576547
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178808 91177308-0d34-0410-b5e6-96231b3b80d8
The DAGCombine logic that recognized a/sqrt(b) and transformed it into
a multiplication by the reciprocal sqrt did not handle cases where the
sqrt and the division were separated by an fpext or fptrunc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178801 91177308-0d34-0410-b5e6-96231b3b80d8
It had been dropped during the switch to yaml::IO. Also add a test going
from yaml2obj to llvm-readobj. It can be extended as we add more
fields/formats to yaml2obj.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178786 91177308-0d34-0410-b5e6-96231b3b80d8
At the time when the XCore backend was added there were some issues with
with overlapping register classes but these all seem to be fixed now.
Describing the register classes correctly allow us to get rid of a
codegen only instruction (LDAWSP_lru6_RRegs) and it means we can
disassemble ru6 instructions that use registers above r11.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178782 91177308-0d34-0410-b5e6-96231b3b80d8
The Thumb2SizeReduction pass avoids false CPSR dependencies, except it
still aggressively creates tMOVi8 instructions because they are so
common.
Avoid creating false CPSR dependencies even for tMOVi8 instructions when
the the CPSR flags are known to have high latency. This allows integer
computation to overlap floating point computations.
Also process blocks in a reverse post-order and propagate high-latency
flags to successors.
<rdar://problem/13468102>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178773 91177308-0d34-0410-b5e6-96231b3b80d8
This requires v9 cmov instructions using the %xcc flags instead of the
%icc flags.
Still missing:
- Select floats on %xcc flags.
- Select i64 on %fcc flags.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178737 91177308-0d34-0410-b5e6-96231b3b80d8
The default logic does not correctly identify costs of casts because they are
marked as custom on x86.
For some cases, where the shift amount is a scalar we would be able to generate
better code. Unfortunately, when this is the case the value (the splat) will get
hoisted out of the loop, thereby making it invisible to ISel.
radar://13130673
radar://13537826
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178703 91177308-0d34-0410-b5e6-96231b3b80d8
Normally r_info is just a 32 of 64 bit number matching the endian of the rest
of the file. Unfortunately, mips 64 bit little endian is special: The top 32
bits are a little endian number and the following 32 are a big endian one.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178694 91177308-0d34-0410-b5e6-96231b3b80d8
ELF with support for:
- File headers
- Section headers + data
- Relocations
- Symbols
- Unwind data (only COFF/Win64)
The output format follows a few rules:
- Values are almost always output one per line (as elf-dump/coff-dump already do). - Many values are translated to something readable (like enum names), with the raw value in parentheses.
- Hex numbers are output in uppercase, prefixed with "0x".
- Flags are sorted alphabetically.
- Lists and groups are always delimited.
Example output:
---------- snip ----------
Sections [
Section {
Index: 1
Name: .text (5)
Type: SHT_PROGBITS (0x1)
Flags [ (0x6)
SHF_ALLOC (0x2)
SHF_EXECINSTR (0x4)
]
Address: 0x0
Offset: 0x40
Size: 33
Link: 0
Info: 0
AddressAlignment: 16
EntrySize: 0
Relocations [
0x6 R_386_32 .rodata.str1.1 0x0
0xB R_386_PC32 puts 0x0
0x12 R_386_32 .rodata.str1.1 0x0
0x17 R_386_PC32 puts 0x0
]
SectionData (
0000: 83EC04C7 04240000 0000E8FC FFFFFFC7 |.....$..........|
0010: 04240600 0000E8FC FFFFFF31 C083C404 |.$.........1....|
0020: C3 |.|
)
}
]
---------- snip ----------
Relocations and symbols can be output standalone or together with the section header as displayed in the example.
This feature set supports all tests in test/MC/COFF and test/MC/ELF (and I suspect all additional tests using elf-dump), making elf-dump and coff-dump deprecated.
Patch by Nico Rieck!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178679 91177308-0d34-0410-b5e6-96231b3b80d8
For this we need to use a libcall. Previously LLVM didn't implement
libcall support for frem, so I've added it in the usual
straightforward manner. A test case from the bug report is included.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178639 91177308-0d34-0410-b5e6-96231b3b80d8
The same compare instruction is used for 32-bit and 64-bit compares. It
sets two different sets of flags: icc and xcc.
This patch adds a conditional branch instruction using the xcc flags for
64-bit compares.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178621 91177308-0d34-0410-b5e6-96231b3b80d8
When unsafe FP math operations are enabled, we can use the fre[s] and
frsqrte[s] instructions, which generate reciprocal (sqrt) estimates, together
with some Newton iteration, in order to quickly generate floating-point
division and sqrt results. All of these instructions are separately optional,
and so each has its own feature flag (except for the Altivec instructions,
which are covered under the existing Altivec flag). Doing this is not only
faster than using the IEEE-compliant fdiv/fsqrt instructions, but allows these
computations to be pipelined with other computations in order to hide their
overall latency.
I've also added a couple of missing fnmsub patterns which turned out to be
missing (but are necessary for good code generation of the Newton iterations).
Altivec needs a similar fix, but that will probably be more complicated because
fneg is expanded for Altivec's v4f32.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178617 91177308-0d34-0410-b5e6-96231b3b80d8
This finally fixes the encoding. The patch also
* Removes eh-frame.ll. It was an unnecessary .ll to .o test that was checking
the wrong value.
* Merge fde-reloc.s and eh-frame.s into a single test, since the only difference
was the run lines.
* Don't blindly test the content of the entire .eh_frame section. It makes it
hard to anyone actually fixing a bug and hitting a difference in a binary
blob. Instead, use a CHECK for each field and document what is being checked.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178615 91177308-0d34-0410-b5e6-96231b3b80d8
The semantics of ARC implies that a pointer passed into an objc_autorelease
must live until some point (potentially down the stack) where an
autorelease pool is popped. On the other hand, an
objc_autoreleaseReturnValue just signifies that the object must live
until the end of the given function at least.
Thus objc_autorelease is stronger than objc_autoreleaseReturnValue in
terms of the semantics of ARC* implying that performing the given
strength reduction without any knowledge of how this relates to
the autorelease pool pop that is further up the stack violates the
semantics of ARC.
*Even though objc_autoreleaseReturnValue if you know that no RV
optimization will occur is more computationally expensive.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178612 91177308-0d34-0410-b5e6-96231b3b80d8
This patch initializes t9 to the handler address, but only if the relocation
model is pic. This handles the case where handler to which eh.return jumps
points to the start of the function.
Patch by Sasa Stankovic.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178588 91177308-0d34-0410-b5e6-96231b3b80d8
When doing a partword atomic operation, a lwarx was being paired with
a stdcx. instead of a stwcx. when compiling for a 64-bit target. The
target has nothing to do with it in this case; we always need a stwcx.
Thanks to Kai Nacke for reporting the problem.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178559 91177308-0d34-0410-b5e6-96231b3b80d8
This is helps on architectures where i8,i16 are not legal but we have byte, and
short loads/stores. Allowing us to merge copies like the one below on ARM.
copy(char *a, char *b, int n) {
do {
int t0 = a[0];
int t1 = a[1];
b[0] = t0;
b[1] = t1;
radar://13536387
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178546 91177308-0d34-0410-b5e6-96231b3b80d8
The iterator could be invalidated when it's recursively deleting a whole bunch
of constant expressions in a constant initializer.
Note: This was only reproducible if `opt' was run on a `.bc' file. If `opt' was
run on a `.ll' file, it wouldn't crash. This is why the test first pushes the
`.ll' file through `llvm-as' before feeding it to `opt'.
PR15440
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178531 91177308-0d34-0410-b5e6-96231b3b80d8
SPARC v9 extends all ALU instructions to 64 bits, so we simply need to
add patterns to use them for both i32 and i64 values.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178527 91177308-0d34-0410-b5e6-96231b3b80d8
The last resort pattern produces 6 instructions, and there are still
opportunities for materializing some immediates in fewer instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178526 91177308-0d34-0410-b5e6-96231b3b80d8
SPARC v9 defines new 64-bit shift instructions. The 32-bit shift right
instructions are still usable as zero and sign extensions.
This adds new F3_Sr and F3_Si instruction formats that probably should
be used for the 32-bit shifts as well. They don't really encode an
simm13 field.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178525 91177308-0d34-0410-b5e6-96231b3b80d8
This is far from complete, but it is enough to make it possible to write
test cases using i64 arguments.
Missing features:
- Floating point arguments.
- Receiving arguments on the stack.
- Calls.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178523 91177308-0d34-0410-b5e6-96231b3b80d8
Revision 177141 caused a regression in all but
mips64 little endian. That is because none of the
other Mips targets had test cases checking the
contents of the .eh_frame section. This patch fixes
both the llvm code and adds an assembler test case
to include the current 4 flavors.
The test cases unfortunately rely on llvm-objdump. A
preferable method would be to use a pretty printer output
such as what readelf -wf <elf_file> would give.
I also changed the name of the test case to correct a typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178506 91177308-0d34-0410-b5e6-96231b3b80d8
We would also like to merge sequences that involve a variable index like in the
example below.
int index = *idx++
int i0 = c[index+0];
int i1 = c[index+1];
b[0] = i0;
b[1] = i1;
By extending the parsing of the base pointer to handle dags that contain a
base, index, and offset we can handle examples like the one above.
The dag for the code above will look something like:
(load (i64 add (i64 copyfromreg %c)
(i64 signextend (i8 load %index))))
(load (i64 add (i64 copyfromreg %c)
(i64 signextend (i32 add (i32 signextend (i8 load %index))
(i32 1)))))
The code that parses the tree ignores the intermediate sign extensions. However,
if there is a sign extension it needs to be on all indexes.
(load (i64 add (i64 copyfromreg %c)
(i64 signextend (add (i8 load %index)
(i8 1))))
vs
(load (i64 add (i64 copyfromreg %c)
(i64 signextend (i32 add (i32 signextend (i8 load %index))
(i32 1)))))
radar://13536387
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178483 91177308-0d34-0410-b5e6-96231b3b80d8
The P7 and A2 have additional floating-point conversion instructions which
allow a direct two-instruction sequence (plus load/store) to convert from all
combinations (signed/unsigned i32/i64) <--> (float/double) (on previous cores,
only some combinations were directly available).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178480 91177308-0d34-0410-b5e6-96231b3b80d8
The popcntw instruction is available whenever the popcntd instruction is
available, and performs a separate popcnt on the lower and upper 32-bits.
Ignoring the high-order count, this can be used for the 32-bit input case
(saving on the explicit zero extension otherwise required to use popcntd).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178470 91177308-0d34-0410-b5e6-96231b3b80d8
This instruction is available on modern PPC64 CPUs, and is now used
to improve the SINT_TO_FP lowering (by eliminating the need for the
separate sign extension instruction and decreasing the amount of
needed stack space).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178446 91177308-0d34-0410-b5e6-96231b3b80d8
The existing SINT_TO_FP code for i32 -> float/double conversion was disabled
because it relied on broken EXTSW_32/STD_32 instruction definitions. The
original intent had been to enable these 64-bit instructions to be used on CPUs
that support them even in 32-bit mode. Unfortunately, this form of lying to
the infrastructure was buggy (as explained in the FIXME comment) and had
therefore been disabled.
This re-enables this functionality, using regular DAG nodes, but only when
compiling in 64-bit mode. The old STD_32/EXTSW_32 definitions (which were dead)
are removed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178438 91177308-0d34-0410-b5e6-96231b3b80d8
'@SECREL' is what is used by the Microsoft assembler, but GNU as expects '@SECREL32'.
With the patch, the MC-generated code works fine in combination with a recent GNU as (2.23.51.20120920 here).
Patch by David Nadlinger!
Differential Revision: http://llvm-reviews.chandlerc.com/D429
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178427 91177308-0d34-0410-b5e6-96231b3b80d8
specific code paths.
This allows us to write code like:
if (__nvvm_reflect("FOO"))
// Do something
else
// Do something else
and compile into a library, then give "FOO" a value at kernel
compile-time so the check becomes a no-op.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178416 91177308-0d34-0410-b5e6-96231b3b80d8
Check that instruction selection can select multiply-add/sub DSP instructions
from a pattern that doesn't have intrinsics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178406 91177308-0d34-0410-b5e6-96231b3b80d8
derived class MipsSETargetLowering.
We shouldn't be generating madd/msub nodes if target is Mips16, since Mips16
doesn't have support for multipy-add/sub instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178404 91177308-0d34-0410-b5e6-96231b3b80d8
Specifically, objc-arc-expand will make sure that the
objc_retainAutoreleasedReturnValue, objc_autoreleaseReturnValue, and ret
will all have %call as an argument.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178382 91177308-0d34-0410-b5e6-96231b3b80d8
clang.arc.used is an interesting call for ARC since ObjCARCContract
needs to run to remove said intrinsic to avoid a linker error (since the
call does not exist).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178369 91177308-0d34-0410-b5e6-96231b3b80d8
Like nearbyint, rint can be implemented on PPC using the frin instruction. The
complication comes from the fact that rint needs to set the FE_INEXACT flag
when the result does not equal the input value (and frin does not do that). As
a result, we use a custom inserter which, after the rounding, compares the
rounded value with the original, and if they differ, explicitly sets the XX bit
in the FPSCR register (which corresponds to FE_INEXACT).
Once LLVM has better modeling of the floating-point environment we should be
able to (often) eliminate this extra complexity.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178362 91177308-0d34-0410-b5e6-96231b3b80d8
These instructions are available on the P5x (and later) and on the A2. They
implement the standard floating-point rounding operations (floor, trunc, etc.).
One caveat: frin (round to nearest) does not implement "ties to even", and so
is only enabled in fast-math mode.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178337 91177308-0d34-0410-b5e6-96231b3b80d8
Mips assembler supports macros that allows the OR instruction
to have an immediate parameter. This patch adds an instruction
alias that converts this macro into a Mips ORI instruction.
Contributer: Vladimir Medic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178316 91177308-0d34-0410-b5e6-96231b3b80d8
- RDRAND always clears the destination value when a random value is not
available (i.e. CF == 0). This value is truncated or zero-extended as
the false boolean value to be returned. Boolean simplification needs
to skip this 'zext' or 'trunc' node.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178312 91177308-0d34-0410-b5e6-96231b3b80d8
Mips assembler allows following to be used as aliased instructions:
jal $rs for jalr $rs
jal $rd,$rd for jalr $rd,$rs
This patch provides alias definitions in td files and test cases to show the usage.
Contributer: Vladimir Medic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178304 91177308-0d34-0410-b5e6-96231b3b80d8
Compiling in 32-bit mode on a P7 would assert after 64-bit DAG combines were
added for bswap with load/store. This is because these combines are really only
valid in 64-bit mode, regardless of the CPU (and this was not being checked).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178286 91177308-0d34-0410-b5e6-96231b3b80d8
Since we handle optimizable objc_retainBlocks through strength reduction
in OptimizableIndividualCalls, we know that all code after that point
will only see non-optimizable objc_retainBlock calls. IsForwarding is
only called by functions after that point, so it is ok to just classify
objc_retainBlock as non-forwarding.
<rdar://problem/13249661>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178285 91177308-0d34-0410-b5e6-96231b3b80d8
If an objc_retainBlock has the copy_on_escape metadata attached to it
AND if the block pointer argument only escapes down the stack, we are
allowed to strength reduce the objc_retainBlock to to an objc_retain and
thus optimize it.
Current there is logic in the ARC data flow analysis to handle
this case which is complicated and involved making distinctions in
between objc_retainBlock and objc_retain in certain places and
considering them the same in others.
This patch simplifies said code by:
1. Performing the strength reduction in the initial ARC peephole
analysis (ObjCARCOpts::OptimizeIndividualCalls).
2. Changes the ARC dataflow analysis (which runs after the peephole
analysis) to consider all objc_retainBlock calls to not be optimizable
(since if the call was optimizable, we would have strength reduced it
already).
This patch leaves in the infrastructure in the ARC dataflow analysis to
handle this case, which due to 2 will just be dead code. I am doing this
on purpose to separate the removal of the old code from the testing of
the new code.
<rdar://problem/13249661>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178284 91177308-0d34-0410-b5e6-96231b3b80d8
