This is part of an ongoing transition to use OperandValueInfo which combines OperandValueKind and OperandValueProperties. This change adds some accessor methods and uses them to simplify backend code. The primary motivation of doing so is removing uses of the parameters so that an upcoming api change is less error prone.
Both are reasonable names; this is solely that an upcoming change can use the OpNInfo name, and the compiler can tell me if I forgot to update something (instead of silently passing along properties that might not hold.)
This keeps clang compatible with MSVC defines for the FP environment.
These defines are used by the CRT and other libraries to interrogate
what to expect. Perhaps most importantly, they feed into the definition
of float_t and double_t which may result in ODR violations between MSVC
and clang.
In preparation for adding support for DBG_VALUE_LIST instructions in
InstrRefLDV, this patch updates the logic for joining variables at block
joins to support joining variables that use multiple debug operands.
This is one of the more meaty "logical" changes, although the line count
isn't too high - this changes pickVPHILoc to find a valid joined
location for every operand, with part of the function being split off
into pickValuePHILoc which finds a location for a single operand.
Differential Revision: https://reviews.llvm.org/D128180
This interface allows a target to reject a proposed
SMS schedule. For Hexagon/PowerPC, all schedules
are accepted, leaving behavior unchanged. For ARM,
schedules which exceed register pressure limits are
rejected.
Also, two RegisterPressureTracker methods now need to be public so
that register pressure can be computed by more callers.
Reapplication of D128941/(reversion:D132037) with small fix.
Differential Revision: https://reviews.llvm.org/D132170
Canonicalize ((x + C1) & C2) --> ((x & C2) + C1) for suitable constants
C1 and C2, instead of the other way round. This should allow more
constant ADDs to be matched as part of addressing modes for loads and
stores.
Differential Revision: https://reviews.llvm.org/D130080
Don't demand low order bits from the LHS of an Add if:
- they are not demanded in the result, and
- they are known to be zero in the RHS, so they can't possibly
overflow and affect higher bit positions
This is intended to avoid a regression from a future patch to change
the order of canonicalization of ADD and AND.
Differential Revision: https://reviews.llvm.org/D130075
They were mapped as follows:
- /fp:except to --ftrapping-math
- /fp:except- to --fno-trapping-math
- /fp:strict to --fno-fast-math
- /fp:precise to --fno-fast-math
Let's map them as follows:
- /fp:except to --ffp-exception-behavior=strict
- /fp:except- to --ffp-exception-behavior=ignore
- /fp:strict to --ffp-model=strict
- /fp:precise to --ffp-model=ignore
I believe the changes to /fp:except are technically a no-op but it makes
the mapping a lot clearer. The changes for /fp:strict and /fp:precise are not
no-ops, they now match MSVC's behavior.
While we are here, also add support for /fp:contract by mapping it to
-ffp-contract=on.
This removes the last use of OperandValueKind from the client side API, and (once this is fully plumbed through TTI implementation) allow use of the same properties in store costing as arithmetic costing.
This patch does the following:
- Consumes the PIC flags (fPIC/fPIE/fropi/frwpi etc) in flang-new.
tools::ParsePICArgs() in ToolChains/CommonArgs.cpp is used for this.
- Adds FC1Option to "-mrelocation-model", "-pic-level", and "-pic-is-pie"
command line options.
- Adds the above options to flang/Frontend/CodeGenOptions' data structure.
- Sets the relocation model in the target machine, and
- Sets module flags for the respective PIC/PIE type in LLVM IR.
I have tried my best to replicate how clang does things.
Differential Revision: https://reviews.llvm.org/D131533
Change-Id: I68fe64910be28147dc5617826641cea71b92d94d
This completes the client side transition to the OperandValueInfo version of this routine. Backend TTI implementations still use the prior versions for now.
We currently use CMake's find_library function to detect whether
libpthread exists on the system to determine if pthread should
be added on the link step. However, there are configurations in
which CMake's path checking fails to find the library even though
the toolchain has it.
One such case is with Clang 14.0.0 on PowerPC. Due to a recent
change, the build puts libc++ and related libraries in a
subdirectory that appears to depend on the default target triple.
CMake then uses that subdirectory to determine the architecture
and adds that name to its search paths. However, the triple for
the system GNU toolchain is different so CMake fails to find it.
Namely, Clang 14.0.0's default target triple and the subdirectory
name is powerpc64le-unknown-linux-gnu whereas the system GNU
toolchain has powerpc64le-linux-gnu. Clang's driver has no trouble
finding either the GNU includes/libraries or Clang's own. But
CMake seems to get this wrong.
The net result of this is that we can't do a shared libraries
build of ToT with Clang 14.0.0.
This patch proposes using HAVE_LIBPTHREAD which CMake seems to
determine by compiling a test file with -lpthread (or perhaps
-pthread, I can't really get CMake to tell me how it is figuring
this out). If that variable tells CMake that the build compiler
accepts the pthread option, it seems reasonable to depend on
that variable to determine if we should add it to the link step
when building the llvm_gtest library.
This patch fixes:
llvm/lib/CodeGen/LiveDebugValues/InstrRefBasedImpl.h:330:5: error:
anonymous types declared in an anonymous union are an extension
[-Werror,-Wnested-anon-types]
In preparation for allowing the prefer_type list in the append_args clause,
use the OMPInteropInfo in the attribute for 'declare variant'.
This requires adding a new Argument kind to the attribute code. This change
adds a specific attribute to pass an array of OMPInteropInfo. It implements
new tablegen needed to handle the interop-type part of the structure. When
prefer_type is added, more work will be needed to dump, instantiate, and
serialize the PreferTypes field in OMPInteropInfo.
Differential Revision: https://reviews.llvm.org/D132270
Perform compile-time folding of the F08 parity intrinsic when all parameters are compile-time constants.
Differential Revision: https://reviews.llvm.org/D129785
In preparation for allowing InstrRefBasedLDV to handle DBG_VALUE_LIST,
this patch updates the internal representation that it uses to represent
debug values to store a list of values. This is one of the more
significant changes in terms of line count, but is fairly simple and
should not affect the output of this pass.
Differential Revision: https://reviews.llvm.org/D128177
OperandValueKind and OperandValueProperties both provide facts about the operands of an instruction for purposes of cost modeling. We've discussed merging them several times; before I plumb through more flags, let's go ahead and do so.
This change only adds the client side interface for getArithmeticInstrCost and makes a couple of minor changes in client code to prove that it works. Target TTI implementations still use the split flags. I'm deliberately splitting what could be one big change into a series of smaller ones so that I can lean on the compiler to catch errors along the way.
`RegisterClassInfo` caches information like allocation orders and reuses
it for multiple machine functions where possible. However the `MCPhysReg
*CalleeSavedRegs` field used to test whether the set of callee saved
registers changed did not work: After D28566
`MachineRegisterInfo::getCalleeSavedRegs()` can return dynamically
computed CSR sets that are only valid while the `MachineRegisterInfo`
object of the current function exists.
This changes the code to make a copy of the CSR list instead of keeping
a possibly invalid pointer around.
Differential Revision: https://reviews.llvm.org/D132080
Before this patch, open-source clang would consider
`-target x86_64-apple-darwin -mios-simulator-version-min=11.0` as
targeting the iOS simulator, due to the mios flag informing it
that we want to target iOS, and logic in the driver then realizing
that x86 iOS builds must be the simulator.
However, for `-target arm64-apple-darwin -mios-simulator-version-min=11.0`
that didn't work and clang thought that it's building for actual iOS,
and not for the simulator.
Due to this, building compiler-rt for arm64 iossim would lead to
all .o files in RTSanitizerCommonSymbolizer.iossim.dir being built
for iOS instead of for iOS simulator, and clang would ask ld64 to
link for iOS, but using the iPhoneSimulator sysroot. This would then
lead to many warnings from ld64 looking like:
ld: warning: building for iOS, but linking in .tbd file
(.../iPhoneSimulator.sdk/usr/lib/libc++abi.tbd) built for iOS Simulator
Worse, with ld64.lld, this diagnostic is currently an error instead
of a warning.
This patch makes it so that the presence of -mios-simulator-version-min=
now informs clang that we're building for simulator. That way, all the
.o files are built for simulator, the linker is informed that we're
building for simulator, and everything Just Works.
(Xcode's clang already behaves like this, so this makes open-source clang
match Xcode clang.)
We can now likely remove the hack to treat non-mac darwin x86 as
simulator, but doing that feels slightly risky, so I'm leaving that
for a follow-up patch.
(This patch is made necessary by the existence of arm64 macs.)
Differential Revision: https://reviews.llvm.org/D132258
Currently, InstrRefLDV only handles DBG_VALUE instructions, not
DBG_VALUE_LIST, and as a result of this it handles these instructions
using functions that only work for that type of debug value, i.e. using
getOperand(0) to get the debug operand. This patch changes this to use
the generic debug value functions, such as getDebugOperand and
isDebugOffsetImm, as well as adding an IsVariadic field to the
DbgValueProperties class and a few other minor changes to acknowledge
DBG_VALUE_LISTs. Note that this patch does not add support for
DBG_VALUE_LIST here, but is a precursor to other patches that do add
that support.
Differential Revision: https://reviews.llvm.org/D128174
tensor.pad is lowered to tensor.generate + tensor.insert_slice during bufferization. For best performance with constant padding values, users should vectorize the IR before bufferizing it.
This change also relaxes tje restriction that no new ops that bufferize to a memory write should be added during bufferization. Since bufferization has been split into two steps a while ago (tensor copy insertion + bufferization), it is reasonable to allow this now.
Differential Revision: https://reviews.llvm.org/D132355
TileToForeachThreadOp now accepts mixed SSA value operands / index attributes for tile_sizes and num_threads. (Reusing OperandsOrIntegersSizesList.) In case of an operand, a PDL_Operation must be specified that is mapped to a payload op that returns the tile size or number of threads.
Differential Revision: https://reviews.llvm.org/D131949
In the InstrRefBasedImpl for LiveDebugValues, we attempt to propagate
debug values through basic blocks in part by checking to see whether all
a variable's incoming debug values to a BB "agree", i.e. whether their
properties match and they refer to the same underlying value.
Prior to this patch, the check for agreement between incoming values
relied on exact equality, which meant that a VPHI and a Def DbgValue
that referred to the same underlying value would be seen as disagreeing.
This patch changes this behaviour to treat them as referring to the same
value, allowing the shared value to propagate into the BB.
Differential Revision: https://reviews.llvm.org/D125953
It was originally left off thinking the paper only impacts the C
standard library, but Clang supports diagnostics for incorrect use of
a format specifier, so this paper has some frontend impacts as well.
This check was put in place to prevent static functions
from translation units outside the one that the current
expression is evaluated from taking precedence over functions
in the global namespace. However, this is really a different
bug. LLDB lumps functions from all CUs into a single AST and
ends up picking the file-static even when C++ context rules
wouldn't allow that to happen.
This patch removes the check so we apply the AsmLabel to all
FunctionDecls we create from DWARF if we have a linkage name
available. This makes the code-path easier to reason about and
allows calling static functions in contexts where we previously
would've chosen the wrong function.
We also flip the XFAILs in the API test to reflect what effect
this change has.
**Testing**
* Fixed API tests and added XFAIL
Differential Revision: https://reviews.llvm.org/D132231
This will be useful in preparation for some reshuffling
of assertions in this file since we won't have to
adjust the persitent variable names during the process.
sed commands:
```
s/expect("expr -- /expect_expr("/g
s/startstr="(int) [$0-9]* = /result_type="int", result_value="/g
```
**Testing**
* API tests still pass
Differential Revision: https://reviews.llvm.org/D132271
This patch makes it possible for lambdas, implicit copy/move ctors
and structured bindings to handle non-POD multidimensional arrays.
Differential Revision: https://reviews.llvm.org/D131840
A fixed length SK_Splice shuffle vector is lowered to a Ext under
AArch64, which should have a cost of 1.
Differential Revision: https://reviews.llvm.org/D132299
Chris Bieneman