Move non-common files from FortranCommon to FortranSupport (analogous to
LLVMSupport) such that
* declarations and definitions that are only used by the Flang compiler,
but not by the runtime, are moved to FortranSupport
* declarations and definitions that are used by both ("common"), the
compiler and the runtime, remain in FortranCommon
* generic STL-like/ADT/utility classes and algorithms remain in
FortranCommon
This allows a for cleaner separation between compiler and runtime
components, which are compiled differently. For instance, runtime
sources must not use STL's `<optional>` which causes problems with CUDA
support. Instead, the surrogate header `flang/Common/optional.h` must be
used. This PR fixes this for `fast-int-sel.h`.
Declarations in include/Runtime are also used by both, but are
header-only. `ISO_Fortran_binding_wrapper.h`, a header used by compiler
and runtime, is also moved into FortranCommon.
When a global variable is used in the OpenMP target region, it is passed
as an argument to the function that implements target region. But the
`DeclareOp` for this incarnation still have the original name of the
variable. As some of our checks to decide if a variable is global or nor
are based on the name, this can result in a local variable being treated
as global. This PR hardens the check a bit. We now also check that
memory ref is actually an `AddrOfOp` before looking at the name.
This patch shares core interface methods dealing with argument and
result attributes from CallableOpInterface with the CallOpInterface and
makes them mandatory to gives more consistent guarantees about concrete
operations using these interfaces.
This allows adding argument attributes on call like operations, which is
sometimes required to get proper ABI, like with llvm.call (and llvm.invoke).
The patch adds optional `arg_attrs` and `res_attrs` attributes to operations using
these interfaces that did not have that already.
They can then re-use the common "rich function signature"
printing/parsing helpers if they want (for the LLVM dialect, this is
done in the next patch).
Part of RFC: https://discourse.llvm.org/t/mlir-rfc-adding-argument-and-result-attributes-to-llvm-call/84107
As there is now certain areas where we now have the possibility of
having either a ModuleOp or GPUModuleOp and both of these modules can
have DataLayout's and we may require utilising the DataLayout utilities
in these areas I've taken the liberty of trying to extend them for use
with both.
Those with more knowledge of how they wish the GPUModuleOp's to interact
with their parent ModuleOp's DataLayout may have further alterations
they wish to make in the future, but for the moment, it'll simply
utilise the basic data layout construction which I believe combines
parent and child datalayouts from the ModuleOp and GPUModuleOp. If there
is no GPUModuleOp DataLayout it should default to the parent ModuleOp.
It's worth noting there is some weirdness if you have two module
operations defining builtin dialect DataLayout Entries, it appears the
combinatorial functionality for DataLayouts doesn't support the merging
of these.
This behaviour is useful for areas like:
https://github.com/llvm/llvm-project/pull/119585/files#diff-19fc4bcb38829d085e25d601d344bbd85bf7ef749ca359e348f4a7c750eae89dR1412
where we have a crossroads between the two different module operations.
This PR adds debug support for common block in flang. As variable which
are part of a common block don't have a special marker to recognize
them, we use the following check to find them.
%0 = fir.address_of(@a)
%1 = fir.convert %0
%2 = fir.coordinate_of %1, %c0
%3 = fir.convert %2
%4 = fircg.ext_declare %3
If the memref of a fircg.ext_declare points to a fir.coordinate_of and
that in turn points to an fir.address_of (ignoring immediate
fir.convert) then we assume that it is a common block variable. The
fir.address_of gives us the global symbol which is the storage for
common block and fir.coordinate_of provides the offset in this storage.
The debug hierarchy looks like as
subroutine f3
integer :: x, y
common /a/ x, y
end subroutine
@a_ = global { ... } { ... }, !dbg !26, !dbg !28!23 = !DISubprogram(name: "f3"...)
!24 = !DICommonBlock(scope: !23, name: "a", ...)
!25 = !DIGlobalVariable(name: "x", scope: !24 ...)
!26 = !DIGlobalVariableExpression(var: !25, expr: !DIExpression())
!27 = !DIGlobalVariable(name: "y", scope: !24 ...)
!28 = !DIGlobalVariableExpression(var: !27, expr:
!DIExpression(DW_OP_plus_uconst, 4))
This required following changes:
1. Instead of using DIGlobalVariableAttr in the FusedLoc of GlobalOp, we
use DIGlobalVariableExpressionAttr. This allows us the generate the
DIExpression where we have the information.
2. Previously, only one DIGlobalVariableExpressionAttr could be linked
to one global op. I recently removed this restriction in mlir. To make
use of it, we add an ArrayAttr to the FusedLoc of a GlobalOp. This
allows us to pass multiple DIGlobalVariableExpressionAttr.
3. I was depending on the name of global for the name of the common
block. The name gets a '_' appended. I could not find a utility function
in flang to remove it so I have to brute force it.
Since https://github.com/llvm/llvm-project/pull/122770, we have seen
that compile time have become extremely slow for cyclic derived types.
In #122770, we made the criteria to cache a derived type very strict. As
a result, some types which are safe to cache were also being
re-generated every type they were required. This increased the compile
time and also the size of the debug info.
Please see the description of PR# 122770. We decided that when
processing `t1`, the type generated for `t2` and `t3` were not safe to
cached. But our algorithm also denied caching to `t1` which as top level
type was safe.
```
type t1
type(t2), pointer :: p1
end type
type t2
type(t3), pointer :: p2
end type
type t3
type(t1), pointer :: p3
end type
```
I have tinkered the check a bit so that top level type is always cached.
To detect a top level type, we use a depth counter that get incremented
before call to `convertRecordType` and decremented after it returns.
After this change, the following
[file](https://github.com/fujitsu/compiler-test-suite/blob/main/Fortran/0394/0394_0031.f90)
from Fujitsu get compiled around 40s which is same as it was before
#122770.
The smaller testcase present in issue #124049 takes less than half a
second. I also added check to make sure that duplicate entries of the
`DICompositeType` are not present in the IR.
Fixes#124049 and #123960.
When `RecordType` is converted to corresponding `DIType`, we cache the
information to avoid doing the conversion again.
Our conversion of `RecordType` looks like this:
`ConvertRecordType(RecordType Ty)`
1. If type `Ty` is already in the cache, then return the corresponding
item.
2. Create a place holder `DICompositeTypeAttr` (called `ty_self` below)
for `Ty`
3. Put `Ty->ty_self` in the cache
4. Convert members of `Ty`. This may cause `ConvertRecordType` to be
called again with other types.
5. Create final `DICompositeTypeAttr`
6. Replace the `ty_self` in the cache with one created in step 5 end
The purpose of creating `ty_self` is to handle cases where a member may
have reference to parent type.
Now consider the code below:
```
type t1
type(t2), pointer :: p1
end type
type t2
type(t1), pointer :: p2
end type
```
While processing t1, we could have a structure like below. `t1 -> t2 ->
t1_self`
The `t2` created during handling of `t1` cant be cached on its own as it
contains a place holder reference. It will fail an assert in MLIR if it
is processed standalone. To avoid this problem, we have a check in the
step 6 above to not cache such types. But this check was not tight
enough. It just checked if a type should not have a place holder
reference to another type. It missed the following case where the place
holder reference can be in a type further down the line.
```
type t1
type(t2), pointer :: p1
end type
type t2
type(t3), pointer :: p2
end type
type t3
type(t1), pointer :: p3
end type
```
So while processing `t1`, we have to stop caching of not only `t3` but
also of `t2`. This PR improves the check and moves the logic inside
`convertRecordType`.
Please note that this limitation of why a type cant have a placeholder
reference is because of how such references are resolved in the mlir.
Please see the discussion at the end of this
[PR](https://github.com/llvm/llvm-project/pull/106571).
I have to change `getDerivedType` so that it will also get the derived
type for things like `type(t2), pointer :: p1` which are wrapped in
`BoxType`. Happy to move it to a new function or a local helper in case
this change is problematic.
Fixes#122024.
Introduce a new `MLIR_LIBS` argument to `add_flang_library`, that uses
`mlir_target_link_libraries` to link the MLIR dylib alterantively to the
component libraries. Use it, along with a few inline
`mlir_target_link_libraries` in tools, to support linking Flang to MLIR
dylib rather than the static libraries.
With these changes, the vast majority of Flang can be linked
dynamically. The only parts still using static libraries are these
requiring MLIR test libraries, that are not included in the dylib.
Introduce a new op to get the device address from a host symbol. This
simplify the current conversion and this is also in preparation for some
legalization work that need to be done in cuf kernel and cuf kernel
launch similar to
https://github.com/llvm/llvm-project/pull/122802
`findAllocaLoopInsertionPoint()` hit assertion not being able
to find the `fir.freemem` because of the `fir.convert`.
I think it is better to look for `fir.freemem` same way
with the look-through walk.
This does add a little computational complexity because now every
freemem operation has to be tested for every allocation. This could be
improved with some more memoisation but I think it is easier to read
this way. Let me know if you would prefer me to change this to
pre-compute the normalised addresses each freemem operation is using.
Weirdly, this change resulted in a verifier failure for the fir.declare
in the previous test case. Maybe it was previously removed as dead code
and now it isn't. Anyway I fixed that too.
Currently fir::isDummyArgument is being used to check if a DeclareOp
represents a dummy argument. The argument passed to the function is
declOp.getMemref(). This bypasses the code in isDummyArgument that
checks for dummy_scope because the `Value` returned by the getMemref()
may not have DeclareOp as its defining op.
This bypassing mean that sometime a variable will be marked as argument
when it should not. This happened in this case where same arg was being
used for 2 different result variables with use of `entry` in the
function.
The solution is to check directly if the declOp has a dummy_scope. If
yes, we know this is dummy argument. We can now check if the memref
points to the BlockArgument and use its number. This will still miss
arguments where memref does not directly point to a BlockArgument but
that is missed currently too. Note that we can still evaluate those
variable in debugger. It is just that they are not marked as arguments.
Fixes#116525.
Introduce cuf.sync_descriptor to be used to sync device global
descriptor after pointer association.
Also move CUFCommon so it can be used in FIRBuilder lib as well.
We were passing size in bytes for the sizeInBits field in
DIDerivedTypeAttr with DW_TAG_pointer_type. Although this field is
un-used in this case but better to be accurate.
Loops resulting from array expressions like array(:,i)
may be versioned for the unit stride of the innermost dimension,
when the initial array is an assumed-shape array (which are contiguous
in many Fortran programs).
This speeds up facerec for about 12% due to further vectorization
of the innermost loop produced for the total SUM reduction.
Note that PointerUnion::{is,get} have been soft deprecated in
PointerUnion.h:
// FIXME: Replace the uses of is(), get() and dyn_cast() with
// isa<T>, cast<T> and the llvm::dyn_cast<T>
I'm not touching PointerUnion::dyn_cast for now because it's a bit
complicated; we could blindly migrate it to dyn_cast_if_present, but
we should probably use dyn_cast when the operand is known to be
non-null.
The greedy rewriter is used in many different flows and it has a lot of
convenience (work list management, debugging actions, tracing, etc). But
it combines two kinds of greedy behavior 1) how ops are matched, 2)
folding wherever it can.
These are independent forms of greedy and leads to inefficiency. E.g.,
cases where one need to create different phases in lowering and is
required to applying patterns in specific order split across different
passes. Using the driver one ends up needlessly retrying folding/having
multiple rounds of folding attempts, where one final run would have
sufficed.
Of course folks can locally avoid this behavior by just building their
own, but this is also a common requested feature that folks keep on
working around locally in suboptimal ways.
For downstream users, there should be no behavioral change. Updating
from the deprecated should just be a find and replace (e.g., `find ./
-type f -exec sed -i
's|applyPatternsAndFoldGreedily|applyPatternsGreedily|g' {} \;` variety)
as the API arguments hasn't changed between the two.
In order to get the pointer to a structure member, `getelementptr`
typically requires two indices: one to indicate the structure itself,
and another to specify the member's position. We are missing the former
in `GPULaunchKernelConversion`, so generated code may cause stack
corruption. This PR corrects the indices of a structure used as a kernel
launch temp.
For the call to _FortranACUFLaunchKernel, we store the pointer to a
member of a temporary structure in a parameter array. However, when we
obtain an element pointer from the parameter array, its address is
calculated based on the type of the structure. This PR properly treats
the parameter array as an array of pointers.
Example:
```mlir
%30 = llvm.load %29 : !llvm.ptr -> i32
%31 = llvm.mlir.constant(1 : i32) : i32
%32 = llvm.alloca %31 x !llvm.struct<(i64, i64, i32, ptr)> : (i32) -> !llvm.ptr
%33 = llvm.mlir.constant(4 : i32) : i32
%34 = llvm.alloca %33 x !llvm.ptr : (i32) -> !llvm.ptr
%35 = llvm.mlir.constant(0 : i32) : i32
%36 = llvm.getelementptr %32[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>
llvm.store %8, %36 : i64, !llvm.ptr
%37 = llvm.getelementptr %34[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>
llvm.store %36, %37 : !llvm.ptr, !llvm.ptr
...
llvm.call @_FortranACUFLaunchKernel(%47, %8, %8, %8, %2, %8, %8, %7, %34, %48) : (!llvm.ptr, i64, i64, i64, i64, i64, i64, i32, !llvm.ptr, !llvm.ptr) -> ()
```
In this example, `%37 = llvm.getelementptr %34[%35] : (!llvm.ptr, i32)
-> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>` will be `%37 =
llvm.getelementptr %34[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.ptr`.
Use `pm.nest` to schedule the pass on nested `func.func` and `gpu.func`
in the `gpu.module`.
AbstractResult pass is not meant to run on the whole gpu.module at once.
1. In `CufOpConversion` `isDeviceGlobal` was renamed
`isRegisteredGlobal` and moved to the common file. `isRegisteredGlobal`
excludes constant `fir.global` operation from registration. This is to
avoid calls to `_FortranACUFGetDeviceAddress` on globals which do not
have any symbols in the runtime. This was done for
`_FortranACUFRegisterVariable` in #118582, but also needs to be done
here after #118591
2. `CufDeviceGlobal` no longer adds the `#cuf.cuda<constant>` attribute
to the constant global. As discussed in #118582 a module variable with
the #cuf.cuda<constant> attribute is not a compile time constant. Yet,
the compile time constant also needs to be copied into the GPU module.
The candidates for copy to the GPU modules are
- the globals needing regsitrations regardless of their uses in device
code (they can be referred to in host code as well)
- the compile time constant when used in device code
3. The registration of "constant" module device variables (
#cuf.cuda<constant>) can be restored in `CufAddConstructor`
Add pattern that update fir.declare memref when it comes from a device
global and is not a descriptor. In that case, we recover the device
address that needs to be used in ops like `fir.array_coor` and so on.
Global constants have no symbols in library files. They are replaced
with literal constants during lowering before kernels are moved into a
GPU module. Do not register them because they will result in unresolved
symbols.
in CUDA Fortran, device function are converted to `gpu.func` inside the
`gpu.module` operation. Update the AbstractResult pass to be able to run
on `func.func` and `gpu.func` operations inside the `gpu.module`.
