Tom Stellard [Thu, 5 Feb 2015 15:32:15 +0000 (15:32 +0000)]
R600/SI: Fix bug from insertion of llvm.SI.end.cf into loop headers
The llvm.SI.end.cf intrinsic is used to mark the end of if-then blocks,
if-then-else blocks, and loops. It is responsible for updating the
exec mask to re-enable threads that had been masked during the preceding
control flow block. For example:
The bug fixed by this patch was one where the llvm.SI.end.cf intrinsic
was being inserted into the header of loops. This would happen when
an if block terminated in a loop header and we would end up with
code like this:
LOOP: ; Start of loop header
s_or_b64 exec, exec, s[0:1] ; llvm.SI.end.cf <-BUG: The exec mask has the
same value at the beginning of each loop
iteration.
do_stuff();
s_cbranch_execnz LOOP
The fix is to create a new basic block before the loop and insert the
llvm.SI.end.cf there. This way the exec mask is restored before the
start of the loop instead of at the beginning of each iteration.
Bill Schmidt [Thu, 5 Feb 2015 15:24:47 +0000 (15:24 +0000)]
[PowerPC] Implement the vclz instructions for PWR8
Patch by Kit Barton.
Add the vector count leading zeros instruction for byte, halfword,
word, and doubleword sizes. This is a fairly straightforward addition
after the changes made for vpopcnt:
1. Add the correct definitions for the various instructions in
PPCInstrAltivec.td
2. Make the CTLZ operation legal on vector types when using P8Altivec
in PPCISelLowering.cpp
Test Plan
Created new test case in test/CodeGen/PowerPC/vec_clz.ll to check the
instructions are being generated when the CTLZ operation is used in
LLVM.
Check the encoding and decoding in test/MC/PowerPC/ppc_encoding_vmx.s
and test/Disassembler/PowerPC/ppc_encoding_vmx.txt respectively.
Implement a BITCAST dag combine to transform i32->mmx conversion patterns
into a X86 specific node (MMX_MOVW2D) and guarantee that moves between
i32 and x86mmx are better handled, i.e., don't use store-load to do the
conversion..
Avoid regression in previously supported MMX code by adding different
combinations of tests which exercise MMX bitcasts. Small improvements
to these patterns should come next.
Teach isDereferenceablePointer() to look through bitcast constant expressions.
This fixes a LICM regression due to the new load+store pair canonicalization.
Craig Topper [Thu, 5 Feb 2015 08:51:02 +0000 (08:51 +0000)]
[X86] Remove two feature flags that covered sets of instructions that have no patterns or intrinsics. Since we don't check feature flags in the assembler parser for any instruction sets, these flags don't provide any value. This frees up 2 of the fully utilized feature flags.
Larisse Voufo [Thu, 5 Feb 2015 04:54:51 +0000 (04:54 +0000)]
Disable enumeral mismatch warning when compiling llvm with gcc.
Tested with gcc 4.9.2.
Compiling with -Werror was producing:
.../llvm/lib/Target/X86/X86ISelLowering.cpp: In function 'llvm::SDValue lowerVectorShuffleAsBitMask(llvm::SDLoc, llvm::MVT, llvm::SDValue, llvm::SDValue, llvm::ArrayRef<int>, llvm::SelectionDAG&)':
.../llvm/lib/Target/X86/X86ISelLowering.cpp:7771:40: error: enumeral mismatch in conditional expression: 'llvm::X86ISD::NodeType' vs 'llvm::ISD::NodeType' [-Werror=enum-compare]
V = DAG.getNode(VT.isFloatingPoint() ? X86ISD::FAND : ISD::AND, DL, VT, V,
^
Matt Arsenault [Thu, 5 Feb 2015 03:35:34 +0000 (03:35 +0000)]
Add addrspacecast node to tablegen
The node is still defined oddly so that the
address spaces are not operands and not accessible
from tablegen, but as-is this can now be used to write
a ComplexPattern with an addrspacecast root node.
Implement new heuristic for complete loop unrolling.
Complete loop unrolling can make some loads constant, thus enabling a
lot of other optimizations. To catch such cases, we look for loads that
might become constants and estimate number of instructions that would be
simplified or become dead after substitution.
Example:
Suppose we have:
int a[] = {0, 1, 0};
v = 0;
for (i = 0; i < 3; i ++)
v += b[i]*a[i];
If we completely unroll the loop, we would get:
v = b[0]*a[0] + b[1]*a[1] + b[2]*a[2]
Which then will be simplified to:
v = b[0]* 0 + b[1]* 1 + b[2]* 0
Value soft float calls as more expensive in the inliner.
Summary: When evaluating floating point instructions in the inliner, ask the TTI whether it is an expensive operation. By default, it's not an expensive operation. This keeps the default behavior the same as before. The ARM TTI has been updated to return back TCC_Expensive for targets which don't have hardware floating point.
Add some API to `APSInt` to make it easier to compare with `int64_t`.
- `APSInt::compareValues(APSInt, APSInt)` returns 1, -1 or 0 for
greater, lesser, or equal, doing the right thing for mismatched
"has-sign" and bitwidths. This is just like `isSameValue()` (and is
now the implementation of it).
- `APSInt::get(int64_t)` gets a signed `APSInt`.
- `operator<(int64_t)`, etc., are implemented trivially via `get()`
and `compareValues()`.
- Also added `APSInt::getUnsigned(uint64_t)` to make it easier to test
`compareValues()`.
Handle LLVM_USE_SANITIZER=Address;Undefined (and the other way around)
Summary:
Handle LLVM_USE_SANITIZER=Address;Undefined to enable ASan and UBSan
If UBSan is compatible with more of the other sanitizers, maybe we should
deal with this in a better way where we allow combining UBSan with any of
the other sanitizers.
AsmParser: Move MDField details to source file, NFC
Move all the types of `MDField` to an anonymous namespace in the source
file. This also eliminates the duplication of `ParseMDField()`
declarations in the header for each new field type.
Tom Stellard [Wed, 4 Feb 2015 20:49:51 +0000 (20:49 +0000)]
R600/SI: Make more store operations legal
v2i32, i32, trunc i32 to i16, and truc i32 to i8 stores are legal for
all address spaces. We had marked them as custom in order to lower
them for the private address space, but this is no longer necessary.
This enables lowering of misaligned stores of these types in the
DAGLegalizer.
Tom Stellard [Wed, 4 Feb 2015 20:49:44 +0000 (20:49 +0000)]
StructurizeCFG: Use a reverse post-order traversal
We were previously doing a post-order traversal and operating on the
list in reverse, however this would occasionaly cause backedges for
loops to be visited before some of the other blocks in the loop.
We know use a reverse post-order traversal, which avoids this issue.
The reverse post-order traversal is not completely ideal, so we need
to manually fixup the list to ensure that inner loop backedges are
visited before outer loop backedges.
Alexey Samsonov [Wed, 4 Feb 2015 17:39:48 +0000 (17:39 +0000)]
SpecialCaseList: Add support for parsing multiple input files.
Summary:
This change allows users to create SpecialCaseList objects from
multiple local files. This is needed to implement a proper support
for -fsanitize-blacklist flag (allow users to specify multiple blacklists,
in addition to default blacklist, see PR22431).
DFSan can also benefit from this change, as DFSan instrumentation pass now
accepts ABI-lists both from -fsanitize-blacklist= and -mllvm -dfsan-abilist flags.
Bradley Smith [Wed, 4 Feb 2015 16:23:24 +0000 (16:23 +0000)]
[ARM] Fix subtarget feature set truncation when using .cpu directive
This is a bug that was caused due to storing the feature bitset in a 32-bit
variable when it is a 64-bit mask, discarding the top half of the feature set.
Renato Golin [Wed, 4 Feb 2015 13:31:29 +0000 (13:31 +0000)]
Adding support to LLVM for targeting Cortex-A72
Currently, Cortex-A72 is modelled as an Cortex-A57 except the fp
load balancing pass isn't enabled for Cortex-A72 as it's not
profitable to have it enabled for this core.
[x86] Give movss and movsd execution domains in the x86 backend.
This associates movss and movsd with the packed single and packed double
execution domains (resp.). While this is largely cosmetic, as we now
don't have weird ping-pong-ing between single and double precision, it
is also useful because it avoids the domain fixing algorithm from seeing
domain breaks that don't actually exist. It will also be much more
important if we have an execution domain default other than packed
single, as that would cause us to mix movss and movsd with integer
vector code on a regular basis, a very bad mixture.
[x86] Mechanically update a bunch of tests' check lines using the latest
version of the script.
Changes include:
- Using the VEX prefix
- Skipping more detail when we have useful shuffle comments to match
- Matching more shuffle comments that have been added to the printer
(yay!)
- Matching the destination registers of some AVX instructions
- Stripping trailing whitespace that crept in
- Fixing indentation issues
Nothing interesting going on here. I'm just trying really hard to ensure
these changes don't show up in the diffs with actual changes to the
backend.
[x86] Teach the test update script to strip trailing whitespace.
This is done in a bit of a strange way to use a multiline RE instead of
looping over the lines. Suggestions welcome here for a more pythonic way
of doing this as long as its reasonably fast.
This reverts patches 223862, 224198, 224203, and 224754, which were all
related to the vector load/store combining and were reverted/reaplied
a few times due to the same alignment problems we're seeing now.
Further tests, mainly self-hosting Clang, will be needed to reapply this
patch in the future.
[x86] Start to introduce bit-masking based blend lowering.
This is the simplest form of bit-math based blending which only fires
when we are blending with zero and is relatively profitable. I've only
enabled this path on very specific lowering strategies. I'm planning to
widen its applicability in subsequent patches, but so far you'll notice
that even though we get fewer shufps instructions, we *still* do the bit
math in the FP execution port. I'm looking into why this is still
happening.
[x86] Add missing patterns for andps, orps, xorps, and andnps.
Specifically, the existing patterns were scalar-only. These cover the
packed vector bitwise operations when specifically requested with pseudo
instructions. This is particularly important in SSE1 where we can't
actually emit a logical operation on a v2i64 as that isn't a legal type.
This will be tested in subsequent patches which form the floating point
and patterns in more places.
projects / llvm / log