This change adds the new unroll metadata "llvm.loop.unroll.enable" which directs
the optimizer to unroll a loop fully if the trip count is known at compile time, and
unroll partially if the trip count is not known at compile time. This differs from
"llvm.loop.unroll.full" which explicitly does not unroll a loop if the trip count is not
known at compile time
With this change "#pragma unroll" generates "llvm.loop.unroll.enable" rather than
"llvm.loop.unroll.full" metadata. This changes the semantics of "#pragma unroll" slightly
to mean "unroll aggressively (fully or partially)" rather than "unroll fully or not at all".
The motivating example for this change was some internal code with a loop marked
with "#pragma unroll" which only sometimes had a compile-time trip count depending
on template magic. When the trip count was a compile-time constant, everything works
as expected and the loop is fully unrolled. However, when the trip count was not a
compile-time constant the "#pragma unroll" explicitly disabled unrolling of the loop(!).
Removing "#pragma unroll" caused the loop to be unrolled partially which was desirable
from a performance perspective.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@244467 91177308-0d34-0410-b5e6-96231b3b80d8
Specifying #pragma clang loop vectorize(assume_safety) on a loop adds the
mem.parallel_loop_access metadata to each load/store operation in the loop. This
metadata tells loop access analysis (LAA) to skip memory dependency checking.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@239572 91177308-0d34-0410-b5e6-96231b3b80d8
Previously loop hints such as #pragma loop vectorize_width(#) required a constant. This patch allows a constant expression to be used as well. Such as a non-type template parameter or an expression (2 * c + 1).
Reviewed by Richard Smith
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@219589 91177308-0d34-0410-b5e6-96231b3b80d8
This patch is necessary to support constant expressions which replaces the integer value in the loop hint attribute with an expression. The integer value was also storing the pragma’s state for options like vectorize(enable/disable) and the pragma unroll and nounroll directive. The state variable is introduced to hold the state of those options/pragmas. This moves the validation of the state (keywords) from SemaStmtAttr handler to the loop hint annotation token handler.
Resubmit with changes to try to fix the build-bot issue.
Reviewed by Aaron Ballman
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@214432 91177308-0d34-0410-b5e6-96231b3b80d8
This patch is necessary to support constant expressions which replaces the integer value in the loop hint attribute with an expression. The integer value was also storing the pragma’s state for options like vectorize(enable/disable) and the pragma unroll and nounroll directive. The state variable is introduced to hold the state of those options/pragmas. This moves the validation of the state (keywords) from SemaStmtAttr handler to the loop hint annotation token handler.
Reviewed by Aaron Ballman
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@214333 91177308-0d34-0410-b5e6-96231b3b80d8
This patch fixes a crash when handling malformed arguments to loop pragmas such
as: "#pragma clang loop vectorize(()". Essentially any argument which is not an
identifier or constant resulted in a crash. This patch also changes a couple of
the error messages which weren't quite correct. New behavior with this patch vs
old behavior:
#pragma clang loop vectorize(1)
OLD: error: missing keyword; expected 'enable' or 'disable'
NEW: error: invalid argument; expected 'enable' or 'disable'
#pragma clang loop vectorize()
OLD: error: expected ')'
NEW: error: missing argument to loop pragma 'vectorize'
#pragma clang loop vectorize_width(bad)
OLD: error: missing value; expected a positive integer value
NEW: error: invalid argument; expected a positive integer value
#pragma clang loop vectorize(bad)
OLD: invalid keyword 'bad'; expected 'enable' or 'disable'
NEW: error: invalid argument; expected 'enable' or 'disable'
http://reviews.llvm.org/D4197
Patch by Mark Heffernan
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@211292 91177308-0d34-0410-b5e6-96231b3b80d8
These cases in particular were incurring an extra strlen() when we already knew
the length. They appear to be leftovers from when the interfaces worked with C
strings that have continued to compile due to the implicit StringRef ctor.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@210403 91177308-0d34-0410-b5e6-96231b3b80d8
Additionally, remove the optional nature of the spelling list index when creating attributes. This is supported by table generating a Spelling enumeration when the spellings for an attribute are distinct enough to warrant it.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@199378 91177308-0d34-0410-b5e6-96231b3b80d8
attributes yet, so just issue the appropriate diagnostics. Also generalize the
fixit for attributes-in-the-wrong-place code and reuse it here, if attributes
are placed after the access-specifier or 'virtual' in a base specifier.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@175575 91177308-0d34-0410-b5e6-96231b3b80d8
- General C++11 attributes were previously parsed and ignored. Now they are parsed and stored in AST.
- Add support to parse arguments of attributes that in 'gnu' namespace.
- Differentiate unknown attributes and known attributes that can't be applied to statements when emitting diagnostic.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@165082 91177308-0d34-0410-b5e6-96231b3b80d8
caused a crash in an obscure case. On the plus side, it caused me to catch
another bug by inspection.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@158767 91177308-0d34-0410-b5e6-96231b3b80d8
The original r158700 caused crashes in the gcc test suite,
g++.abi/vtable3a.C among others. It also caused failures in the libc++
test suite.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@158749 91177308-0d34-0410-b5e6-96231b3b80d8
Note that this is mostly a structural patch that handles the change from the old
spelling style to the new one. One consequence of this is that all AT_foo_bar
enum values have changed to not be based off of the first spelling, but rather
off of the class name, so they are now AT_FooBar and the like (a straw poll on
IRC showed support for this). Apologies for code churn.
Most attributes have GNU spellings as a temporary solution until everything else
is sorted out (such as a Keyword spelling, which I intend to add if someone else
doesn't beat me to it). This is definitely a WIP.
I've also killed BaseCheckAttr since it was unused, and I had to go through
every attribute anyway.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@158700 91177308-0d34-0410-b5e6-96231b3b80d8
Now, as long as the 'Namespaces' variable is correct inside Attr.td, the
generated code will correctly admit a C++11 attribute only when it has the
appropriate namespace(s).
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@158661 91177308-0d34-0410-b5e6-96231b3b80d8
cases in switch statements. Also add a [[clang::fallthrough]] attribute, which
can be used to suppress the warning in the case of intentional fallthrough.
Patch by Alexander Kornienko!
The handling of C++11 attribute namespaces in this patch is temporary, and will
be replaced with a cleaner mechanism in a subsequent patch.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@156086 91177308-0d34-0410-b5e6-96231b3b80d8
attached. Since we do not support any attributes which appertain to a statement
(yet), testing of this is necessarily quite minimal.
Patch by Alexander Kornienko!
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@154723 91177308-0d34-0410-b5e6-96231b3b80d8
Anastasia Stulova