llvm-project/mlir/lib/IR/SymbolTable.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1128 lines
45 KiB
C++
Raw Normal View History

//===- SymbolTable.cpp - MLIR Symbol Table Class --------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "mlir/IR/SymbolTable.h"
#include "mlir/IR/Builders.h"
#include "mlir/IR/OpImplementation.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallString.h"
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
#include "llvm/ADT/StringSwitch.h"
#include <optional>
using namespace mlir;
/// Return true if the given operation is unknown and may potentially define a
/// symbol table.
static bool isPotentiallyUnknownSymbolTable(Operation *op) {
return op->getNumRegions() == 1 && !op->getDialect();
}
/// Returns the string name of the given symbol, or null if this is not a
/// symbol.
static StringAttr getNameIfSymbol(Operation *op) {
return op->getAttrOfType<StringAttr>(SymbolTable::getSymbolAttrName());
}
static StringAttr getNameIfSymbol(Operation *op, StringAttr symbolAttrNameId) {
return op->getAttrOfType<StringAttr>(symbolAttrNameId);
}
/// Computes the nested symbol reference attribute for the symbol 'symbolName'
/// that are usable within the symbol table operations from 'symbol' as far up
/// to the given operation 'within', where 'within' is an ancestor of 'symbol'.
/// Returns success if all references up to 'within' could be computed.
static LogicalResult
collectValidReferencesFor(Operation *symbol, StringAttr symbolName,
Operation *within,
SmallVectorImpl<SymbolRefAttr> &results) {
assert(within->isAncestor(symbol) && "expected 'within' to be an ancestor");
MLIRContext *ctx = symbol->getContext();
auto leafRef = FlatSymbolRefAttr::get(symbolName);
results.push_back(leafRef);
// Early exit for when 'within' is the parent of 'symbol'.
Operation *symbolTableOp = symbol->getParentOp();
if (within == symbolTableOp)
return success();
// Collect references until 'symbolTableOp' reaches 'within'.
SmallVector<FlatSymbolRefAttr, 1> nestedRefs(1, leafRef);
StringAttr symbolNameId =
StringAttr::get(ctx, SymbolTable::getSymbolAttrName());
do {
// Each parent of 'symbol' should define a symbol table.
if (!symbolTableOp->hasTrait<OpTrait::SymbolTable>())
return failure();
// Each parent of 'symbol' should also be a symbol.
StringAttr symbolTableName = getNameIfSymbol(symbolTableOp, symbolNameId);
if (!symbolTableName)
return failure();
results.push_back(SymbolRefAttr::get(symbolTableName, nestedRefs));
symbolTableOp = symbolTableOp->getParentOp();
if (symbolTableOp == within)
break;
nestedRefs.insert(nestedRefs.begin(),
FlatSymbolRefAttr::get(symbolTableName));
} while (true);
return success();
}
/// Walk all of the operations within the given set of regions, without
/// traversing into any nested symbol tables. Stops walking if the result of the
/// callback is anything other than `WalkResult::advance`.
static std::optional<WalkResult>
walkSymbolTable(MutableArrayRef<Region> regions,
function_ref<std::optional<WalkResult>(Operation *)> callback) {
SmallVector<Region *, 1> worklist(llvm::make_pointer_range(regions));
while (!worklist.empty()) {
for (Operation &op : worklist.pop_back_val()->getOps()) {
std::optional<WalkResult> result = callback(&op);
if (result != WalkResult::advance())
return result;
// If this op defines a new symbol table scope, we can't traverse. Any
// symbol references nested within 'op' are different semantically.
if (!op.hasTrait<OpTrait::SymbolTable>()) {
for (Region &region : op.getRegions())
worklist.push_back(&region);
}
}
}
return WalkResult::advance();
}
/// Walk all of the operations nested under, and including, the given operation,
/// without traversing into any nested symbol tables. Stops walking if the
/// result of the callback is anything other than `WalkResult::advance`.
static std::optional<WalkResult>
walkSymbolTable(Operation *op,
function_ref<std::optional<WalkResult>(Operation *)> callback) {
std::optional<WalkResult> result = callback(op);
if (result != WalkResult::advance() || op->hasTrait<OpTrait::SymbolTable>())
return result;
return walkSymbolTable(op->getRegions(), callback);
}
//===----------------------------------------------------------------------===//
// SymbolTable
//===----------------------------------------------------------------------===//
/// Build a symbol table with the symbols within the given operation.
SymbolTable::SymbolTable(Operation *symbolTableOp)
: symbolTableOp(symbolTableOp) {
assert(symbolTableOp->hasTrait<OpTrait::SymbolTable>() &&
"expected operation to have SymbolTable trait");
assert(symbolTableOp->getNumRegions() == 1 &&
"expected operation to have a single region");
assert(llvm::hasSingleElement(symbolTableOp->getRegion(0)) &&
"expected operation to have a single block");
StringAttr symbolNameId = StringAttr::get(symbolTableOp->getContext(),
SymbolTable::getSymbolAttrName());
for (auto &op : symbolTableOp->getRegion(0).front()) {
StringAttr name = getNameIfSymbol(&op, symbolNameId);
if (!name)
continue;
auto inserted = symbolTable.insert({name, &op});
(void)inserted;
assert(inserted.second &&
"expected region to contain uniquely named symbol operations");
}
}
/// Look up a symbol with the specified name, returning null if no such name
/// exists. Names never include the @ on them.
Operation *SymbolTable::lookup(StringRef name) const {
return lookup(StringAttr::get(symbolTableOp->getContext(), name));
}
Operation *SymbolTable::lookup(StringAttr name) const {
return symbolTable.lookup(name);
}
void SymbolTable::remove(Operation *op) {
StringAttr name = getNameIfSymbol(op);
assert(name && "expected valid 'name' attribute");
assert(op->getParentOp() == symbolTableOp &&
"expected this operation to be inside of the operation with this "
"SymbolTable");
auto it = symbolTable.find(name);
if (it != symbolTable.end() && it->second == op)
symbolTable.erase(it);
}
void SymbolTable::erase(Operation *symbol) {
remove(symbol);
symbol->erase();
}
// TODO: Consider if this should be renamed to something like insertOrUpdate
/// Insert a new symbol into the table and associated operation if not already
/// there and rename it as necessary to avoid collisions. Return the name of
/// the symbol after insertion as attribute.
StringAttr SymbolTable::insert(Operation *symbol, Block::iterator insertPt) {
// The symbol cannot be the child of another op and must be the child of the
// symbolTableOp after this.
//
// TODO: consider if SymbolTable's constructor should behave the same.
if (!symbol->getParentOp()) {
auto &body = symbolTableOp->getRegion(0).front();
if (insertPt == Block::iterator()) {
insertPt = Block::iterator(body.end());
} else {
assert((insertPt == body.end() ||
insertPt->getParentOp() == symbolTableOp) &&
"expected insertPt to be in the associated module operation");
}
// Insert before the terminator, if any.
if (insertPt == Block::iterator(body.end()) && !body.empty() &&
std::prev(body.end())->hasTrait<OpTrait::IsTerminator>())
insertPt = std::prev(body.end());
body.getOperations().insert(insertPt, symbol);
}
assert(symbol->getParentOp() == symbolTableOp &&
"symbol is already inserted in another op");
// Add this symbol to the symbol table, uniquing the name if a conflict is
// detected.
StringAttr name = getSymbolName(symbol);
if (symbolTable.insert({name, symbol}).second)
return name;
// If the symbol was already in the table, also return.
if (symbolTable.lookup(name) == symbol)
return name;
MLIRContext *context = symbol->getContext();
[mlir][gpu] Introduce `gpu.dynamic_shared_memory` Op (#71546) While the `gpu.launch` Op allows setting the size via the `dynamic_shared_memory_size` argument, accessing the dynamic shared memory is very convoluted. This PR implements the proposed Op, `gpu.dynamic_shared_memory` that aims to simplify the utilization of dynamic shared memory. RFC: https://discourse.llvm.org/t/rfc-simplifying-dynamic-shared-memory-access-in-gpu/ **Proposal from RFC** This PR `gpu.dynamic.shared.memory` Op to use dynamic shared memory feature efficiently. It is is a powerful feature that enables the allocation of shared memory at runtime with the kernel launch on the host. Afterwards, the memory can be accessed directly from the device. I believe similar story exists for AMDGPU. **Current way Using Dynamic Shared Memory with MLIR** Let me illustrate the challenges of using dynamic shared memory in MLIR with an example below. The process involves several steps: - memref.global 0-sized array LLVM's NVPTX backend expects - dynamic_shared_memory_size Set the size of dynamic shared memory - memref.get_global Access the global symbol - reinterpret_cast and subview Many OPs for pointer arithmetic ``` // Step 1. Create 0-sized global symbol. Manually set the alignment memref.global "private" @dynamicShmem : memref<0xf16, 3> { alignment = 16 } func.func @main() { // Step 2. Allocate shared memory gpu.launch blocks(...) threads(...) dynamic_shared_memory_size %c10000 { // Step 3. Access the global object %shmem = memref.get_global @dynamicShmem : memref<0xf16, 3> // Step 4. A sequence of `memref.reinterpret_cast` and `memref.subview` operations. %4 = memref.reinterpret_cast %shmem to offset: [0], sizes: [14, 64, 128], strides: [8192,128,1] : memref<0xf16, 3> to memref<14x64x128xf16,3> %5 = memref.subview %4[7, 0, 0][7, 64, 128][1,1,1] : memref<14x64x128xf16,3> to memref<7x64x128xf16, strided<[8192, 128, 1], offset: 57344>, 3> %6 = memref.subview %5[2, 0, 0][1, 64, 128][1,1,1] : memref<7x64x128xf16, strided<[8192, 128, 1], offset: 57344>, 3> to memref<64x128xf16, strided<[128, 1], offset: 73728>, 3> %7 = memref.subview %6[0, 0][64, 64][1,1] : memref<64x128xf16, strided<[128, 1], offset: 73728>, 3> to memref<64x64xf16, strided<[128, 1], offset: 73728>, 3> %8 = memref.subview %6[32, 0][64, 64][1,1] : memref<64x128xf16, strided<[128, 1], offset: 73728>, 3> to memref<64x64xf16, strided<[128, 1], offset: 77824>, 3> // Step.5 Use "test.use.shared.memory"(%7) : (memref<64x64xf16, strided<[128, 1], offset: 73728>, 3>) -> (index) "test.use.shared.memory"(%8) : (memref<64x64xf16, strided<[128, 1], offset: 77824>, 3>) -> (index) gpu.terminator } ``` Let’s write the program above with that: ``` func.func @main() { gpu.launch blocks(...) threads(...) dynamic_shared_memory_size %c10000 { %i = arith.constant 18 : index // Step 1: Obtain shared memory directly %shmem = gpu.dynamic_shared_memory : memref<?xi8, 3> %c147456 = arith.constant 147456 : index %c155648 = arith.constant 155648 : index %7 = memref.view %shmem[%c147456][] : memref<?xi8, 3> to memref<64x64xf16, 3> %8 = memref.view %shmem[%c155648][] : memref<?xi8, 3> to memref<64x64xf16, 3> // Step 2: Utilize the shared memory "test.use.shared.memory"(%7) : (memref<64x64xf16, 3>) -> (index) "test.use.shared.memory"(%8) : (memref<64x64xf16, 3>) -> (index) } } ``` This PR resolves #72513
2023-11-16 14:42:17 +01:00
SmallString<128> nameBuffer = generateSymbolName<128>(
name.getValue(),
[&](StringRef candidate) {
return !symbolTable
.insert({StringAttr::get(context, candidate), symbol})
.second;
},
uniquingCounter);
setSymbolName(symbol, nameBuffer);
return getSymbolName(symbol);
}
LogicalResult SymbolTable::rename(StringAttr from, StringAttr to) {
Operation *op = lookup(from);
return rename(op, to);
}
LogicalResult SymbolTable::rename(Operation *op, StringAttr to) {
StringAttr from = getNameIfSymbol(op);
(void)from;
assert(from && "expected valid 'name' attribute");
assert(op->getParentOp() == symbolTableOp &&
"expected this operation to be inside of the operation with this "
"SymbolTable");
assert(lookup(from) == op && "current name does not resolve to op");
assert(lookup(to) == nullptr && "new name already exists");
if (failed(SymbolTable::replaceAllSymbolUses(op, to, getOp())))
return failure();
// Remove op with old name, change name, add with new name. The order is
// important here due to how `remove` and `insert` rely on the op name.
remove(op);
setSymbolName(op, to);
insert(op);
assert(lookup(to) == op && "new name does not resolve to renamed op");
assert(lookup(from) == nullptr && "old name still exists");
return success();
}
LogicalResult SymbolTable::rename(StringAttr from, StringRef to) {
auto toAttr = StringAttr::get(getOp()->getContext(), to);
return rename(from, toAttr);
}
LogicalResult SymbolTable::rename(Operation *op, StringRef to) {
auto toAttr = StringAttr::get(getOp()->getContext(), to);
return rename(op, toAttr);
}
FailureOr<StringAttr>
SymbolTable::renameToUnique(StringAttr oldName,
ArrayRef<SymbolTable *> others) {
// Determine new name that is unique in all symbol tables.
StringAttr newName;
{
MLIRContext *context = oldName.getContext();
SmallString<64> prefix = oldName.getValue();
int uniqueId = 0;
prefix.push_back('_');
while (true) {
newName = StringAttr::get(context, prefix + Twine(uniqueId++));
auto lookupNewName = [&](SymbolTable *st) { return st->lookup(newName); };
if (!lookupNewName(this) && llvm::none_of(others, lookupNewName)) {
break;
}
}
}
// Apply renaming.
if (failed(rename(oldName, newName)))
return failure();
return newName;
}
FailureOr<StringAttr>
SymbolTable::renameToUnique(Operation *op, ArrayRef<SymbolTable *> others) {
StringAttr from = getNameIfSymbol(op);
assert(from && "expected valid 'name' attribute");
return renameToUnique(from, others);
}
/// Returns the name of the given symbol operation.
StringAttr SymbolTable::getSymbolName(Operation *symbol) {
StringAttr name = getNameIfSymbol(symbol);
assert(name && "expected valid symbol name");
return name;
}
/// Sets the name of the given symbol operation.
void SymbolTable::setSymbolName(Operation *symbol, StringAttr name) {
symbol->setAttr(getSymbolAttrName(), name);
}
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
/// Returns the visibility of the given symbol operation.
SymbolTable::Visibility SymbolTable::getSymbolVisibility(Operation *symbol) {
// If the attribute doesn't exist, assume public.
StringAttr vis = symbol->getAttrOfType<StringAttr>(getVisibilityAttrName());
if (!vis)
return Visibility::Public;
// Otherwise, switch on the string value.
return StringSwitch<Visibility>(vis.getValue())
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
.Case("private", Visibility::Private)
.Case("nested", Visibility::Nested)
.Case("public", Visibility::Public);
}
/// Sets the visibility of the given symbol operation.
void SymbolTable::setSymbolVisibility(Operation *symbol, Visibility vis) {
MLIRContext *ctx = symbol->getContext();
// If the visibility is public, just drop the attribute as this is the
// default.
if (vis == Visibility::Public) {
symbol->removeAttr(StringAttr::get(ctx, getVisibilityAttrName()));
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
return;
}
// Otherwise, update the attribute.
assert((vis == Visibility::Private || vis == Visibility::Nested) &&
"unknown symbol visibility kind");
StringRef visName = vis == Visibility::Private ? "private" : "nested";
symbol->setAttr(getVisibilityAttrName(), StringAttr::get(ctx, visName));
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
}
/// Returns the nearest symbol table from a given operation `from`. Returns
/// nullptr if no valid parent symbol table could be found.
Operation *SymbolTable::getNearestSymbolTable(Operation *from) {
assert(from && "expected valid operation");
if (isPotentiallyUnknownSymbolTable(from))
return nullptr;
while (!from->hasTrait<OpTrait::SymbolTable>()) {
from = from->getParentOp();
// Check that this is a valid op and isn't an unknown symbol table.
if (!from || isPotentiallyUnknownSymbolTable(from))
return nullptr;
}
return from;
}
/// Walks all symbol table operations nested within, and including, `op`. For
/// each symbol table operation, the provided callback is invoked with the op
/// and a boolean signifying if the symbols within that symbol table can be
/// treated as if all uses are visible. `allSymUsesVisible` identifies whether
/// all of the symbol uses of symbols within `op` are visible.
void SymbolTable::walkSymbolTables(
Operation *op, bool allSymUsesVisible,
function_ref<void(Operation *, bool)> callback) {
bool isSymbolTable = op->hasTrait<OpTrait::SymbolTable>();
if (isSymbolTable) {
SymbolOpInterface symbol = dyn_cast<SymbolOpInterface>(op);
allSymUsesVisible |= !symbol || symbol.isPrivate();
} else {
// Otherwise if 'op' is not a symbol table, any nested symbols are
// guaranteed to be hidden.
allSymUsesVisible = true;
}
for (Region &region : op->getRegions())
for (Block &block : region)
for (Operation &nestedOp : block)
walkSymbolTables(&nestedOp, allSymUsesVisible, callback);
// If 'op' had the symbol table trait, visit it after any nested symbol
// tables.
if (isSymbolTable)
callback(op, allSymUsesVisible);
}
/// Returns the operation registered with the given symbol name with the
/// regions of 'symbolTableOp'. 'symbolTableOp' is required to be an operation
/// with the 'OpTrait::SymbolTable' trait. Returns nullptr if no valid symbol
/// was found.
Operation *SymbolTable::lookupSymbolIn(Operation *symbolTableOp,
StringAttr symbol) {
assert(symbolTableOp->hasTrait<OpTrait::SymbolTable>());
Region &region = symbolTableOp->getRegion(0);
if (region.empty())
return nullptr;
// Look for a symbol with the given name.
StringAttr symbolNameId = StringAttr::get(symbolTableOp->getContext(),
SymbolTable::getSymbolAttrName());
for (auto &op : region.front())
if (getNameIfSymbol(&op, symbolNameId) == symbol)
return &op;
return nullptr;
}
Operation *SymbolTable::lookupSymbolIn(Operation *symbolTableOp,
SymbolRefAttr symbol) {
SmallVector<Operation *, 4> resolvedSymbols;
if (failed(lookupSymbolIn(symbolTableOp, symbol, resolvedSymbols)))
return nullptr;
return resolvedSymbols.back();
}
/// Internal implementation of `lookupSymbolIn` that allows for specialized
/// implementations of the lookup function.
static LogicalResult lookupSymbolInImpl(
Operation *symbolTableOp, SymbolRefAttr symbol,
SmallVectorImpl<Operation *> &symbols,
function_ref<Operation *(Operation *, StringAttr)> lookupSymbolFn) {
assert(symbolTableOp->hasTrait<OpTrait::SymbolTable>());
// Lookup the root reference for this symbol.
symbolTableOp = lookupSymbolFn(symbolTableOp, symbol.getRootReference());
if (!symbolTableOp)
return failure();
symbols.push_back(symbolTableOp);
// If there are no nested references, just return the root symbol directly.
ArrayRef<FlatSymbolRefAttr> nestedRefs = symbol.getNestedReferences();
if (nestedRefs.empty())
return success();
// Verify that the root is also a symbol table.
if (!symbolTableOp->hasTrait<OpTrait::SymbolTable>())
return failure();
// Otherwise, lookup each of the nested non-leaf references and ensure that
// each corresponds to a valid symbol table.
for (FlatSymbolRefAttr ref : nestedRefs.drop_back()) {
symbolTableOp = lookupSymbolFn(symbolTableOp, ref.getAttr());
if (!symbolTableOp || !symbolTableOp->hasTrait<OpTrait::SymbolTable>())
return failure();
symbols.push_back(symbolTableOp);
}
symbols.push_back(lookupSymbolFn(symbolTableOp, symbol.getLeafReference()));
return success(symbols.back());
}
LogicalResult
SymbolTable::lookupSymbolIn(Operation *symbolTableOp, SymbolRefAttr symbol,
SmallVectorImpl<Operation *> &symbols) {
auto lookupFn = [](Operation *symbolTableOp, StringAttr symbol) {
return lookupSymbolIn(symbolTableOp, symbol);
};
return lookupSymbolInImpl(symbolTableOp, symbol, symbols, lookupFn);
}
/// Returns the operation registered with the given symbol name within the
/// closes parent operation with the 'OpTrait::SymbolTable' trait. Returns
/// nullptr if no valid symbol was found.
Operation *SymbolTable::lookupNearestSymbolFrom(Operation *from,
StringAttr symbol) {
Operation *symbolTableOp = getNearestSymbolTable(from);
return symbolTableOp ? lookupSymbolIn(symbolTableOp, symbol) : nullptr;
}
Operation *SymbolTable::lookupNearestSymbolFrom(Operation *from,
SymbolRefAttr symbol) {
Operation *symbolTableOp = getNearestSymbolTable(from);
return symbolTableOp ? lookupSymbolIn(symbolTableOp, symbol) : nullptr;
}
raw_ostream &mlir::operator<<(raw_ostream &os,
SymbolTable::Visibility visibility) {
switch (visibility) {
case SymbolTable::Visibility::Public:
return os << "public";
case SymbolTable::Visibility::Private:
return os << "private";
case SymbolTable::Visibility::Nested:
return os << "nested";
}
llvm_unreachable("Unexpected visibility");
}
//===----------------------------------------------------------------------===//
// SymbolTable Trait Types
//===----------------------------------------------------------------------===//
LogicalResult detail::verifySymbolTable(Operation *op) {
if (op->getNumRegions() != 1)
return op->emitOpError()
<< "Operations with a 'SymbolTable' must have exactly one region";
if (!llvm::hasSingleElement(op->getRegion(0)))
return op->emitOpError()
<< "Operations with a 'SymbolTable' must have exactly one block";
// Check that all symbols are uniquely named within child regions.
DenseMap<Attribute, Location> nameToOrigLoc;
for (auto &block : op->getRegion(0)) {
for (auto &op : block) {
// Check for a symbol name attribute.
auto nameAttr =
op.getAttrOfType<StringAttr>(mlir::SymbolTable::getSymbolAttrName());
if (!nameAttr)
continue;
// Try to insert this symbol into the table.
auto it = nameToOrigLoc.try_emplace(nameAttr, op.getLoc());
if (!it.second)
return op.emitError()
.append("redefinition of symbol named '", nameAttr.getValue(), "'")
.attachNote(it.first->second)
.append("see existing symbol definition here");
}
}
// Verify any nested symbol user operations.
SymbolTableCollection symbolTable;
auto verifySymbolUserFn = [&](Operation *op) -> std::optional<WalkResult> {
if (SymbolUserOpInterface user = dyn_cast<SymbolUserOpInterface>(op))
return WalkResult(user.verifySymbolUses(symbolTable));
return WalkResult::advance();
};
std::optional<WalkResult> result =
walkSymbolTable(op->getRegions(), verifySymbolUserFn);
return success(result && !result->wasInterrupted());
}
LogicalResult detail::verifySymbol(Operation *op) {
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
// Verify the name attribute.
if (!op->getAttrOfType<StringAttr>(mlir::SymbolTable::getSymbolAttrName()))
return op->emitOpError() << "requires string attribute '"
<< mlir::SymbolTable::getSymbolAttrName() << "'";
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
// Verify the visibility attribute.
if (Attribute vis = op->getAttr(mlir::SymbolTable::getVisibilityAttrName())) {
[mlir] Update method cast calls to function calls The MLIR classes Type/Attribute/Operation/Op/Value support cast/dyn_cast/isa/dyn_cast_or_null functionality through llvm's doCast functionality in addition to defining methods with the same name. This change begins the migration of uses of the method to the corresponding function call as has been decided as more consistent. Note that there still exist classes that only define methods directly, such as AffineExpr, and this does not include work currently to support a functional cast/isa call. Context: * https://mlir.llvm.org/deprecation/ at "Use the free function variants for dyn_cast/cast/isa/…" * Original discussion at https://discourse.llvm.org/t/preferred-casting-style-going-forward/68443 Implementation: This follows a previous patch that updated calls `op.cast<T>()-> cast<T>(op)`. However some cases could not handle an unprefixed `cast` call due to occurrences of variables named cast, or occurring inside of class definitions which would resolve to the method. All C++ files that did not work automatically with `cast<T>()` are updated here to `llvm::cast` and similar with the intention that they can be easily updated after the methods are removed through a find-replace. See https://github.com/llvm/llvm-project/compare/main...tpopp:llvm-project:tidy-cast-check for the clang-tidy check that is used and then update printed occurrences of the function to include `llvm::` before. One can then run the following: ``` ninja -C $BUILD_DIR clang-tidy run-clang-tidy -clang-tidy-binary=$BUILD_DIR/bin/clang-tidy -checks='-*,misc-cast-functions'\ -export-fixes /tmp/cast/casts.yaml mlir/*\ -header-filter=mlir/ -fix rm -rf $BUILD_DIR/tools/mlir/**/*.inc ``` Differential Revision: https://reviews.llvm.org/D150348
2023-05-11 11:10:46 +02:00
StringAttr visStrAttr = llvm::dyn_cast<StringAttr>(vis);
[mlir] Add support for attaching a visibility to symbols. Summary: The visibility defines the structural reachability of the symbol within the IR. Symbols can define one of three visibilities: * Public The symbol \may be accessed from outside of the visible IR. We cannot assume that we can observe all of the uses of this symbol. * Private The symbol may only be referenced from within the operations in the current symbol table, via SymbolRefAttr. * Nested The symbol may be referenced by operations in symbol tables above the current symbol table, as long as each symbol table parent also defines a non-private symbol. This allows or referencing the symbol from outside of the defining symbol table, while retaining the ability for the compiler to see all uses. These properties help to reason about the properties of a symbol, and will be used in a follow up to implement a dce pass on dead symbols. A few examples of what this would look like in the IR are shown below: module @public_module { // This function can be accessed by 'live.user' func @nested_function() attributes { sym_visibility = "nested" } // This function cannot be accessed outside of 'public_module' func @private_function() attributes { sym_visibility = "private" } } // This function can only be accessed from within this module. func @private_function() attributes { sym_visibility = "private" } // This function may be referenced externally. func @public_function() "live.user"() {uses = [@public_module::@nested_function, @private_function, @public_function]} : () -> () Depends On D72043 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D72044
2020-01-13 15:54:08 -08:00
if (!visStrAttr)
return op->emitOpError() << "requires visibility attribute '"
<< mlir::SymbolTable::getVisibilityAttrName()
<< "' to be a string attribute, but got " << vis;
if (!llvm::is_contained(ArrayRef<StringRef>{"public", "private", "nested"},
visStrAttr.getValue()))
return op->emitOpError()
<< "visibility expected to be one of [\"public\", \"private\", "
"\"nested\"], but got "
<< visStrAttr;
}
return success();
}
//===----------------------------------------------------------------------===//
// Symbol Use Lists
//===----------------------------------------------------------------------===//
/// Walk all of the symbol references within the given operation, invoking the
/// provided callback for each found use. The callbacks takes the use of the
/// symbol.
static WalkResult
walkSymbolRefs(Operation *op,
function_ref<WalkResult(SymbolTable::SymbolUse)> callback) {
return op->getAttrDictionary().walk<WalkOrder::PreOrder>(
[&](SymbolRefAttr symbolRef) {
if (callback({op, symbolRef}).wasInterrupted())
return WalkResult::interrupt();
// Don't walk nested references.
return WalkResult::skip();
});
}
/// Walk all of the uses, for any symbol, that are nested within the given
/// regions, invoking the provided callback for each. This does not traverse
/// into any nested symbol tables.
static std::optional<WalkResult>
walkSymbolUses(MutableArrayRef<Region> regions,
function_ref<WalkResult(SymbolTable::SymbolUse)> callback) {
return walkSymbolTable(regions,
[&](Operation *op) -> std::optional<WalkResult> {
// Check that this isn't a potentially unknown symbol
// table.
if (isPotentiallyUnknownSymbolTable(op))
return std::nullopt;
return walkSymbolRefs(op, callback);
});
}
/// Walk all of the uses, for any symbol, that are nested within the given
/// operation 'from', invoking the provided callback for each. This does not
/// traverse into any nested symbol tables.
static std::optional<WalkResult>
walkSymbolUses(Operation *from,
function_ref<WalkResult(SymbolTable::SymbolUse)> callback) {
// If this operation has regions, and it, as well as its dialect, isn't
// registered then conservatively fail. The operation may define a
// symbol table, so we can't opaquely know if we should traverse to find
// nested uses.
if (isPotentiallyUnknownSymbolTable(from))
return std::nullopt;
// Walk the uses on this operation.
if (walkSymbolRefs(from, callback).wasInterrupted())
return WalkResult::interrupt();
// Only recurse if this operation is not a symbol table. A symbol table
// defines a new scope, so we can't walk the attributes from within the symbol
// table op.
if (!from->hasTrait<OpTrait::SymbolTable>())
return walkSymbolUses(from->getRegions(), callback);
return WalkResult::advance();
}
namespace {
/// This class represents a single symbol scope. A symbol scope represents the
/// set of operations nested within a symbol table that may reference symbols
/// within that table. A symbol scope does not contain the symbol table
/// operation itself, just its contained operations. A scope ends at leaf
/// operations or another symbol table operation.
struct SymbolScope {
/// Walk the symbol uses within this scope, invoking the given callback.
/// This variant is used when the callback type matches that expected by
/// 'walkSymbolUses'.
template <typename CallbackT,
std::enable_if_t<!std::is_same<
typename llvm::function_traits<CallbackT>::result_t,
void>::value> * = nullptr>
std::optional<WalkResult> walk(CallbackT cback) {
[mlir] Move casting calls from methods to function calls The MLIR classes Type/Attribute/Operation/Op/Value support cast/dyn_cast/isa/dyn_cast_or_null functionality through llvm's doCast functionality in addition to defining methods with the same name. This change begins the migration of uses of the method to the corresponding function call as has been decided as more consistent. Note that there still exist classes that only define methods directly, such as AffineExpr, and this does not include work currently to support a functional cast/isa call. Context: - https://mlir.llvm.org/deprecation/ at "Use the free function variants for dyn_cast/cast/isa/…" - Original discussion at https://discourse.llvm.org/t/preferred-casting-style-going-forward/68443 Implementation: This patch updates all remaining uses of the deprecated functionality in mlir/. This was done with clang-tidy as described below and further modifications to GPUBase.td and OpenMPOpsInterfaces.td. Steps are described per line, as comments are removed by git: 0. Retrieve the change from the following to build clang-tidy with an additional check: main...tpopp:llvm-project:tidy-cast-check 1. Build clang-tidy 2. Run clang-tidy over your entire codebase while disabling all checks and enabling the one relevant one. Run on all header files also. 3. Delete .inc files that were also modified, so the next build rebuilds them to a pure state. ``` ninja -C $BUILD_DIR clang-tidy run-clang-tidy -clang-tidy-binary=$BUILD_DIR/bin/clang-tidy -checks='-*,misc-cast-functions'\ -header-filter=mlir/ mlir/* -fix rm -rf $BUILD_DIR/tools/mlir/**/*.inc ``` Differential Revision: https://reviews.llvm.org/D151542
2023-05-26 10:17:47 +02:00
if (Region *region = llvm::dyn_cast_if_present<Region *>(limit))
return walkSymbolUses(*region, cback);
return walkSymbolUses(limit.get<Operation *>(), cback);
}
/// This variant is used when the callback type matches a stripped down type:
/// void(SymbolTable::SymbolUse use)
template <typename CallbackT,
std::enable_if_t<std::is_same<
typename llvm::function_traits<CallbackT>::result_t,
void>::value> * = nullptr>
std::optional<WalkResult> walk(CallbackT cback) {
return walk([=](SymbolTable::SymbolUse use) {
return cback(use), WalkResult::advance();
});
}
/// Walk all of the operations nested under the current scope without
/// traversing into any nested symbol tables.
template <typename CallbackT>
std::optional<WalkResult> walkSymbolTable(CallbackT &&cback) {
[mlir] Move casting calls from methods to function calls The MLIR classes Type/Attribute/Operation/Op/Value support cast/dyn_cast/isa/dyn_cast_or_null functionality through llvm's doCast functionality in addition to defining methods with the same name. This change begins the migration of uses of the method to the corresponding function call as has been decided as more consistent. Note that there still exist classes that only define methods directly, such as AffineExpr, and this does not include work currently to support a functional cast/isa call. Context: - https://mlir.llvm.org/deprecation/ at "Use the free function variants for dyn_cast/cast/isa/…" - Original discussion at https://discourse.llvm.org/t/preferred-casting-style-going-forward/68443 Implementation: This patch updates all remaining uses of the deprecated functionality in mlir/. This was done with clang-tidy as described below and further modifications to GPUBase.td and OpenMPOpsInterfaces.td. Steps are described per line, as comments are removed by git: 0. Retrieve the change from the following to build clang-tidy with an additional check: main...tpopp:llvm-project:tidy-cast-check 1. Build clang-tidy 2. Run clang-tidy over your entire codebase while disabling all checks and enabling the one relevant one. Run on all header files also. 3. Delete .inc files that were also modified, so the next build rebuilds them to a pure state. ``` ninja -C $BUILD_DIR clang-tidy run-clang-tidy -clang-tidy-binary=$BUILD_DIR/bin/clang-tidy -checks='-*,misc-cast-functions'\ -header-filter=mlir/ mlir/* -fix rm -rf $BUILD_DIR/tools/mlir/**/*.inc ``` Differential Revision: https://reviews.llvm.org/D151542
2023-05-26 10:17:47 +02:00
if (Region *region = llvm::dyn_cast_if_present<Region *>(limit))
return ::walkSymbolTable(*region, cback);
return ::walkSymbolTable(limit.get<Operation *>(), cback);
}
/// The representation of the symbol within this scope.
SymbolRefAttr symbol;
/// The IR unit representing this scope.
llvm::PointerUnion<Operation *, Region *> limit;
};
} // namespace
/// Collect all of the symbol scopes from 'symbol' to (inclusive) 'limit'.
static SmallVector<SymbolScope, 2> collectSymbolScopes(Operation *symbol,
Operation *limit) {
StringAttr symName = SymbolTable::getSymbolName(symbol);
assert(!symbol->hasTrait<OpTrait::SymbolTable>() || symbol != limit);
// Compute the ancestors of 'limit'.
SetVector<Operation *, SmallVector<Operation *, 4>,
SmallPtrSet<Operation *, 4>>
limitAncestors;
Operation *limitAncestor = limit;
do {
// Check to see if 'symbol' is an ancestor of 'limit'.
if (limitAncestor == symbol) {
// Check that the nearest symbol table is 'symbol's parent. SymbolRefAttr
// doesn't support parent references.
if (SymbolTable::getNearestSymbolTable(limit->getParentOp()) ==
symbol->getParentOp())
return {{SymbolRefAttr::get(symName), limit}};
return {};
}
limitAncestors.insert(limitAncestor);
} while ((limitAncestor = limitAncestor->getParentOp()));
// Try to find the first ancestor of 'symbol' that is an ancestor of 'limit'.
Operation *commonAncestor = symbol->getParentOp();
do {
if (limitAncestors.count(commonAncestor))
break;
} while ((commonAncestor = commonAncestor->getParentOp()));
assert(commonAncestor && "'limit' and 'symbol' have no common ancestor");
// Compute the set of valid nested references for 'symbol' as far up to the
// common ancestor as possible.
SmallVector<SymbolRefAttr, 2> references;
bool collectedAllReferences = succeeded(
collectValidReferencesFor(symbol, symName, commonAncestor, references));
// Handle the case where the common ancestor is 'limit'.
if (commonAncestor == limit) {
SmallVector<SymbolScope, 2> scopes;
// Walk each of the ancestors of 'symbol', calling the compute function for
// each one.
Operation *limitIt = symbol->getParentOp();
for (size_t i = 0, e = references.size(); i != e;
++i, limitIt = limitIt->getParentOp()) {
assert(limitIt->hasTrait<OpTrait::SymbolTable>());
scopes.push_back({references[i], &limitIt->getRegion(0)});
}
return scopes;
}
// Otherwise, we just need the symbol reference for 'symbol' that will be
// used within 'limit'. This is the last reference in the list we computed
// above if we were able to collect all references.
if (!collectedAllReferences)
return {};
return {{references.back(), limit}};
}
static SmallVector<SymbolScope, 2> collectSymbolScopes(Operation *symbol,
Region *limit) {
auto scopes = collectSymbolScopes(symbol, limit->getParentOp());
// If we collected some scopes to walk, make sure to constrain the one for
// limit to the specific region requested.
if (!scopes.empty())
scopes.back().limit = limit;
return scopes;
}
static SmallVector<SymbolScope, 1> collectSymbolScopes(StringAttr symbol,
[mlir] Make overloads of SymbolTable::replaceAllSymbolUses consistent. (#68320) This function has several overloads that allow to specify the symbol that should be renamed and the scope for that renaming in different ways. The overloads were inconsistent in the following way (quoted strings are `StringAttr`s, other variables are `Operation *`): * `replaceAllSymbolUses(symbolOp, "new_symbol", scopeOp)` would traverse into the nested regions of `scopeOp` and hence rename the symbol inside of `scopeOp`. * `replaceAllSymbolUses("symbol", "new_symbol", scopeOp)` would *not* traverse into the nested regions of `scopeOp` and hence *not* rename the symbol. The underlying behavior was spread over different places and is somewhat hard to understand. The two overloads above mainly differed by what `collectSymbolScopes` computed, which is itself overloaded. If `scopeOp` is a top-level module, then the overload on `(Operation *, Operation *)`, which is used in the first of the above cases, computes a scope where the body region of the module is the `limit`; however, the overload on `(StringAttr, Operation *)` computed the module op itself as the `limit`. Later, `walkSymbolTable` would walk the body of the module if it was given as a region but it would *not* enter the regions of the module op because that op has a symbol table (which was assumed to be a *different* scope). The fix in this commit is change the behavior of `collectSymbolScopes` such that the `(StringAttr, Operation *)` overload returns a scope for each region in the `limit` argument.
2023-10-10 07:47:08 +02:00
Region *limit) {
return {{SymbolRefAttr::get(symbol), limit}};
}
[mlir] Make overloads of SymbolTable::replaceAllSymbolUses consistent. (#68320) This function has several overloads that allow to specify the symbol that should be renamed and the scope for that renaming in different ways. The overloads were inconsistent in the following way (quoted strings are `StringAttr`s, other variables are `Operation *`): * `replaceAllSymbolUses(symbolOp, "new_symbol", scopeOp)` would traverse into the nested regions of `scopeOp` and hence rename the symbol inside of `scopeOp`. * `replaceAllSymbolUses("symbol", "new_symbol", scopeOp)` would *not* traverse into the nested regions of `scopeOp` and hence *not* rename the symbol. The underlying behavior was spread over different places and is somewhat hard to understand. The two overloads above mainly differed by what `collectSymbolScopes` computed, which is itself overloaded. If `scopeOp` is a top-level module, then the overload on `(Operation *, Operation *)`, which is used in the first of the above cases, computes a scope where the body region of the module is the `limit`; however, the overload on `(StringAttr, Operation *)` computed the module op itself as the `limit`. Later, `walkSymbolTable` would walk the body of the module if it was given as a region but it would *not* enter the regions of the module op because that op has a symbol table (which was assumed to be a *different* scope). The fix in this commit is change the behavior of `collectSymbolScopes` such that the `(StringAttr, Operation *)` overload returns a scope for each region in the `limit` argument.
2023-10-10 07:47:08 +02:00
static SmallVector<SymbolScope, 1> collectSymbolScopes(StringAttr symbol,
Operation *limit) {
SmallVector<SymbolScope, 1> scopes;
auto symbolRef = SymbolRefAttr::get(symbol);
for (auto &region : limit->getRegions())
scopes.push_back({symbolRef, &region});
return scopes;
}
/// Returns true if the given reference 'SubRef' is a sub reference of the
/// reference 'ref', i.e. 'ref' is a further qualified reference.
static bool isReferencePrefixOf(SymbolRefAttr subRef, SymbolRefAttr ref) {
if (ref == subRef)
return true;
// If the references are not pointer equal, check to see if `subRef` is a
// prefix of `ref`.
[mlir] Update method cast calls to function calls The MLIR classes Type/Attribute/Operation/Op/Value support cast/dyn_cast/isa/dyn_cast_or_null functionality through llvm's doCast functionality in addition to defining methods with the same name. This change begins the migration of uses of the method to the corresponding function call as has been decided as more consistent. Note that there still exist classes that only define methods directly, such as AffineExpr, and this does not include work currently to support a functional cast/isa call. Context: * https://mlir.llvm.org/deprecation/ at "Use the free function variants for dyn_cast/cast/isa/…" * Original discussion at https://discourse.llvm.org/t/preferred-casting-style-going-forward/68443 Implementation: This follows a previous patch that updated calls `op.cast<T>()-> cast<T>(op)`. However some cases could not handle an unprefixed `cast` call due to occurrences of variables named cast, or occurring inside of class definitions which would resolve to the method. All C++ files that did not work automatically with `cast<T>()` are updated here to `llvm::cast` and similar with the intention that they can be easily updated after the methods are removed through a find-replace. See https://github.com/llvm/llvm-project/compare/main...tpopp:llvm-project:tidy-cast-check for the clang-tidy check that is used and then update printed occurrences of the function to include `llvm::` before. One can then run the following: ``` ninja -C $BUILD_DIR clang-tidy run-clang-tidy -clang-tidy-binary=$BUILD_DIR/bin/clang-tidy -checks='-*,misc-cast-functions'\ -export-fixes /tmp/cast/casts.yaml mlir/*\ -header-filter=mlir/ -fix rm -rf $BUILD_DIR/tools/mlir/**/*.inc ``` Differential Revision: https://reviews.llvm.org/D150348
2023-05-11 11:10:46 +02:00
if (llvm::isa<FlatSymbolRefAttr>(ref) ||
ref.getRootReference() != subRef.getRootReference())
return false;
auto refLeafs = ref.getNestedReferences();
auto subRefLeafs = subRef.getNestedReferences();
return subRefLeafs.size() < refLeafs.size() &&
subRefLeafs == refLeafs.take_front(subRefLeafs.size());
}
//===----------------------------------------------------------------------===//
// SymbolTable::getSymbolUses
/// The implementation of SymbolTable::getSymbolUses below.
template <typename FromT>
static std::optional<SymbolTable::UseRange> getSymbolUsesImpl(FromT from) {
std::vector<SymbolTable::SymbolUse> uses;
auto walkFn = [&](SymbolTable::SymbolUse symbolUse) {
uses.push_back(symbolUse);
return WalkResult::advance();
};
auto result = walkSymbolUses(from, walkFn);
return result ? std::optional<SymbolTable::UseRange>(std::move(uses))
: std::nullopt;
}
/// Get an iterator range for all of the uses, for any symbol, that are nested
/// within the given operation 'from'. This does not traverse into any nested
/// symbol tables, and will also only return uses on 'from' if it does not
/// also define a symbol table. This is because we treat the region as the
/// boundary of the symbol table, and not the op itself. This function returns
/// std::nullopt if there are any unknown operations that may potentially be
/// symbol tables.
auto SymbolTable::getSymbolUses(Operation *from) -> std::optional<UseRange> {
return getSymbolUsesImpl(from);
}
auto SymbolTable::getSymbolUses(Region *from) -> std::optional<UseRange> {
return getSymbolUsesImpl(MutableArrayRef<Region>(*from));
}
//===----------------------------------------------------------------------===//
// SymbolTable::getSymbolUses
/// The implementation of SymbolTable::getSymbolUses below.
template <typename SymbolT, typename IRUnitT>
static std::optional<SymbolTable::UseRange> getSymbolUsesImpl(SymbolT symbol,
IRUnitT *limit) {
std::vector<SymbolTable::SymbolUse> uses;
for (SymbolScope &scope : collectSymbolScopes(symbol, limit)) {
if (!scope.walk([&](SymbolTable::SymbolUse symbolUse) {
if (isReferencePrefixOf(scope.symbol, symbolUse.getSymbolRef()))
uses.push_back(symbolUse);
}))
return std::nullopt;
}
return SymbolTable::UseRange(std::move(uses));
}
/// Get all of the uses of the given symbol that are nested within the given
/// operation 'from', invoking the provided callback for each. This does not
/// traverse into any nested symbol tables. This function returns std::nullopt
/// if there are any unknown operations that may potentially be symbol tables.
auto SymbolTable::getSymbolUses(StringAttr symbol, Operation *from)
-> std::optional<UseRange> {
return getSymbolUsesImpl(symbol, from);
}
auto SymbolTable::getSymbolUses(Operation *symbol, Operation *from)
-> std::optional<UseRange> {
return getSymbolUsesImpl(symbol, from);
}
auto SymbolTable::getSymbolUses(StringAttr symbol, Region *from)
-> std::optional<UseRange> {
return getSymbolUsesImpl(symbol, from);
}
auto SymbolTable::getSymbolUses(Operation *symbol, Region *from)
-> std::optional<UseRange> {
return getSymbolUsesImpl(symbol, from);
}
//===----------------------------------------------------------------------===//
// SymbolTable::symbolKnownUseEmpty
/// The implementation of SymbolTable::symbolKnownUseEmpty below.
template <typename SymbolT, typename IRUnitT>
static bool symbolKnownUseEmptyImpl(SymbolT symbol, IRUnitT *limit) {
for (SymbolScope &scope : collectSymbolScopes(symbol, limit)) {
// Walk all of the symbol uses looking for a reference to 'symbol'.
if (scope.walk([&](SymbolTable::SymbolUse symbolUse) {
return isReferencePrefixOf(scope.symbol, symbolUse.getSymbolRef())
? WalkResult::interrupt()
: WalkResult::advance();
}) != WalkResult::advance())
return false;
}
return true;
}
/// Return if the given symbol is known to have no uses that are nested within
/// the given operation 'from'. This does not traverse into any nested symbol
/// tables. This function will also return false if there are any unknown
/// operations that may potentially be symbol tables.
bool SymbolTable::symbolKnownUseEmpty(StringAttr symbol, Operation *from) {
return symbolKnownUseEmptyImpl(symbol, from);
}
bool SymbolTable::symbolKnownUseEmpty(Operation *symbol, Operation *from) {
return symbolKnownUseEmptyImpl(symbol, from);
}
bool SymbolTable::symbolKnownUseEmpty(StringAttr symbol, Region *from) {
return symbolKnownUseEmptyImpl(symbol, from);
}
bool SymbolTable::symbolKnownUseEmpty(Operation *symbol, Region *from) {
return symbolKnownUseEmptyImpl(symbol, from);
}
//===----------------------------------------------------------------------===//
// SymbolTable::replaceAllSymbolUses
/// Generates a new symbol reference attribute with a new leaf reference.
static SymbolRefAttr generateNewRefAttr(SymbolRefAttr oldAttr,
FlatSymbolRefAttr newLeafAttr) {
[mlir] Update method cast calls to function calls The MLIR classes Type/Attribute/Operation/Op/Value support cast/dyn_cast/isa/dyn_cast_or_null functionality through llvm's doCast functionality in addition to defining methods with the same name. This change begins the migration of uses of the method to the corresponding function call as has been decided as more consistent. Note that there still exist classes that only define methods directly, such as AffineExpr, and this does not include work currently to support a functional cast/isa call. Context: * https://mlir.llvm.org/deprecation/ at "Use the free function variants for dyn_cast/cast/isa/…" * Original discussion at https://discourse.llvm.org/t/preferred-casting-style-going-forward/68443 Implementation: This follows a previous patch that updated calls `op.cast<T>()-> cast<T>(op)`. However some cases could not handle an unprefixed `cast` call due to occurrences of variables named cast, or occurring inside of class definitions which would resolve to the method. All C++ files that did not work automatically with `cast<T>()` are updated here to `llvm::cast` and similar with the intention that they can be easily updated after the methods are removed through a find-replace. See https://github.com/llvm/llvm-project/compare/main...tpopp:llvm-project:tidy-cast-check for the clang-tidy check that is used and then update printed occurrences of the function to include `llvm::` before. One can then run the following: ``` ninja -C $BUILD_DIR clang-tidy run-clang-tidy -clang-tidy-binary=$BUILD_DIR/bin/clang-tidy -checks='-*,misc-cast-functions'\ -export-fixes /tmp/cast/casts.yaml mlir/*\ -header-filter=mlir/ -fix rm -rf $BUILD_DIR/tools/mlir/**/*.inc ``` Differential Revision: https://reviews.llvm.org/D150348
2023-05-11 11:10:46 +02:00
if (llvm::isa<FlatSymbolRefAttr>(oldAttr))
return newLeafAttr;
auto nestedRefs = llvm::to_vector<2>(oldAttr.getNestedReferences());
nestedRefs.back() = newLeafAttr;
return SymbolRefAttr::get(oldAttr.getRootReference(), nestedRefs);
}
/// The implementation of SymbolTable::replaceAllSymbolUses below.
template <typename SymbolT, typename IRUnitT>
static LogicalResult
replaceAllSymbolUsesImpl(SymbolT symbol, StringAttr newSymbol, IRUnitT *limit) {
// Generate a new attribute to replace the given attribute.
FlatSymbolRefAttr newLeafAttr = FlatSymbolRefAttr::get(newSymbol);
for (SymbolScope &scope : collectSymbolScopes(symbol, limit)) {
SymbolRefAttr oldAttr = scope.symbol;
SymbolRefAttr newAttr = generateNewRefAttr(scope.symbol, newLeafAttr);
AttrTypeReplacer replacer;
replacer.addReplacement(
[&](SymbolRefAttr attr) -> std::pair<Attribute, WalkResult> {
// Regardless of the match, don't walk nested SymbolRefAttrs, we don't
// want to accidentally replace an inner reference.
if (attr == oldAttr)
return {newAttr, WalkResult::skip()};
// Handle prefix matches.
if (isReferencePrefixOf(oldAttr, attr)) {
auto oldNestedRefs = oldAttr.getNestedReferences();
auto nestedRefs = attr.getNestedReferences();
if (oldNestedRefs.empty())
return {SymbolRefAttr::get(newSymbol, nestedRefs),
WalkResult::skip()};
auto newNestedRefs = llvm::to_vector<4>(nestedRefs);
newNestedRefs[oldNestedRefs.size() - 1] = newLeafAttr;
return {SymbolRefAttr::get(attr.getRootReference(), newNestedRefs),
WalkResult::skip()};
}
return {attr, WalkResult::skip()};
});
auto walkFn = [&](Operation *op) -> std::optional<WalkResult> {
replacer.replaceElementsIn(op);
return WalkResult::advance();
};
if (!scope.walkSymbolTable(walkFn))
return failure();
}
return success();
}
/// Attempt to replace all uses of the given symbol 'oldSymbol' with the
/// provided symbol 'newSymbol' that are nested within the given operation
/// 'from'. This does not traverse into any nested symbol tables. If there are
/// any unknown operations that may potentially be symbol tables, no uses are
/// replaced and failure is returned.
LogicalResult SymbolTable::replaceAllSymbolUses(StringAttr oldSymbol,
StringAttr newSymbol,
Operation *from) {
return replaceAllSymbolUsesImpl(oldSymbol, newSymbol, from);
}
LogicalResult SymbolTable::replaceAllSymbolUses(Operation *oldSymbol,
StringAttr newSymbol,
Operation *from) {
return replaceAllSymbolUsesImpl(oldSymbol, newSymbol, from);
}
LogicalResult SymbolTable::replaceAllSymbolUses(StringAttr oldSymbol,
StringAttr newSymbol,
Region *from) {
return replaceAllSymbolUsesImpl(oldSymbol, newSymbol, from);
}
LogicalResult SymbolTable::replaceAllSymbolUses(Operation *oldSymbol,
StringAttr newSymbol,
Region *from) {
return replaceAllSymbolUsesImpl(oldSymbol, newSymbol, from);
}
//===----------------------------------------------------------------------===//
// SymbolTableCollection
//===----------------------------------------------------------------------===//
Operation *SymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
StringAttr symbol) {
return getSymbolTable(symbolTableOp).lookup(symbol);
}
Operation *SymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
SymbolRefAttr name) {
SmallVector<Operation *, 4> symbols;
if (failed(lookupSymbolIn(symbolTableOp, name, symbols)))
return nullptr;
return symbols.back();
}
/// A variant of 'lookupSymbolIn' that returns all of the symbols referenced by
/// a given SymbolRefAttr. Returns failure if any of the nested references could
/// not be resolved.
LogicalResult
SymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
SymbolRefAttr name,
SmallVectorImpl<Operation *> &symbols) {
auto lookupFn = [this](Operation *symbolTableOp, StringAttr symbol) {
return lookupSymbolIn(symbolTableOp, symbol);
};
return lookupSymbolInImpl(symbolTableOp, name, symbols, lookupFn);
}
/// Returns the operation registered with the given symbol name within the
/// closest parent operation of, or including, 'from' with the
/// 'OpTrait::SymbolTable' trait. Returns nullptr if no valid symbol was
/// found.
Operation *SymbolTableCollection::lookupNearestSymbolFrom(Operation *from,
StringAttr symbol) {
Operation *symbolTableOp = SymbolTable::getNearestSymbolTable(from);
return symbolTableOp ? lookupSymbolIn(symbolTableOp, symbol) : nullptr;
}
Operation *
SymbolTableCollection::lookupNearestSymbolFrom(Operation *from,
SymbolRefAttr symbol) {
Operation *symbolTableOp = SymbolTable::getNearestSymbolTable(from);
return symbolTableOp ? lookupSymbolIn(symbolTableOp, symbol) : nullptr;
}
/// Lookup, or create, a symbol table for an operation.
SymbolTable &SymbolTableCollection::getSymbolTable(Operation *op) {
auto it = symbolTables.try_emplace(op, nullptr);
if (it.second)
it.first->second = std::make_unique<SymbolTable>(op);
return *it.first->second;
}
//===----------------------------------------------------------------------===//
// LockedSymbolTableCollection
//===----------------------------------------------------------------------===//
Operation *LockedSymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
StringAttr symbol) {
return getSymbolTable(symbolTableOp).lookup(symbol);
}
Operation *
LockedSymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
FlatSymbolRefAttr symbol) {
return lookupSymbolIn(symbolTableOp, symbol.getAttr());
}
Operation *LockedSymbolTableCollection::lookupSymbolIn(Operation *symbolTableOp,
SymbolRefAttr name) {
SmallVector<Operation *> symbols;
if (failed(lookupSymbolIn(symbolTableOp, name, symbols)))
return nullptr;
return symbols.back();
}
LogicalResult LockedSymbolTableCollection::lookupSymbolIn(
Operation *symbolTableOp, SymbolRefAttr name,
SmallVectorImpl<Operation *> &symbols) {
auto lookupFn = [this](Operation *symbolTableOp, StringAttr symbol) {
return lookupSymbolIn(symbolTableOp, symbol);
};
return lookupSymbolInImpl(symbolTableOp, name, symbols, lookupFn);
}
SymbolTable &
2023-02-22 14:05:50 -08:00
LockedSymbolTableCollection::getSymbolTable(Operation *symbolTableOp) {
assert(symbolTableOp->hasTrait<OpTrait::SymbolTable>());
// Try to find an existing symbol table.
{
llvm::sys::SmartScopedReader<true> lock(mutex);
auto it = collection.symbolTables.find(symbolTableOp);
if (it != collection.symbolTables.end())
return *it->second;
}
// Create a symbol table for the operation. Perform construction outside of
// the critical section.
auto symbolTable = std::make_unique<SymbolTable>(symbolTableOp);
// Insert the constructed symbol table.
llvm::sys::SmartScopedWriter<true> lock(mutex);
return *collection.symbolTables
.insert({symbolTableOp, std::move(symbolTable)})
.first->second;
}
//===----------------------------------------------------------------------===//
// SymbolUserMap
//===----------------------------------------------------------------------===//
SymbolUserMap::SymbolUserMap(SymbolTableCollection &symbolTable,
Operation *symbolTableOp)
: symbolTable(symbolTable) {
// Walk each of the symbol tables looking for discardable callgraph nodes.
SmallVector<Operation *> symbols;
auto walkFn = [&](Operation *symbolTableOp, bool allUsesVisible) {
for (Operation &nestedOp : symbolTableOp->getRegion(0).getOps()) {
auto symbolUses = SymbolTable::getSymbolUses(&nestedOp);
assert(symbolUses && "expected uses to be valid");
for (const SymbolTable::SymbolUse &use : *symbolUses) {
symbols.clear();
(void)symbolTable.lookupSymbolIn(symbolTableOp, use.getSymbolRef(),
symbols);
for (Operation *symbolOp : symbols)
symbolToUsers[symbolOp].insert(use.getUser());
}
}
};
// We just set `allSymUsesVisible` to false here because it isn't necessary
// for building the user map.
SymbolTable::walkSymbolTables(symbolTableOp, /*allSymUsesVisible=*/false,
walkFn);
}
void SymbolUserMap::replaceAllUsesWith(Operation *symbol,
StringAttr newSymbolName) {
auto it = symbolToUsers.find(symbol);
if (it == symbolToUsers.end())
return;
// Replace the uses within the users of `symbol`.
for (Operation *user : it->second)
(void)SymbolTable::replaceAllSymbolUses(symbol, newSymbolName, user);
// Move the current users of `symbol` to the new symbol if it is in the
// symbol table.
Operation *newSymbol =
symbolTable.lookupSymbolIn(symbol->getParentOp(), newSymbolName);
if (newSymbol != symbol) {
// Transfer over the users to the new symbol. The reference to the old one
// is fetched again as the iterator is invalidated during the insertion.
auto newIt = symbolToUsers.try_emplace(newSymbol, SetVector<Operation *>{});
auto oldIt = symbolToUsers.find(symbol);
assert(oldIt != symbolToUsers.end() && "missing old users list");
if (newIt.second)
newIt.first->second = std::move(oldIt->second);
else
newIt.first->second.set_union(oldIt->second);
symbolToUsers.erase(oldIt);
}
}
//===----------------------------------------------------------------------===//
// Visibility parsing implementation.
//===----------------------------------------------------------------------===//
ParseResult impl::parseOptionalVisibilityKeyword(OpAsmParser &parser,
NamedAttrList &attrs) {
StringRef visibility;
if (parser.parseOptionalKeyword(&visibility, {"public", "private", "nested"}))
return failure();
StringAttr visibilityAttr = parser.getBuilder().getStringAttr(visibility);
attrs.push_back(parser.getBuilder().getNamedAttr(
SymbolTable::getVisibilityAttrName(), visibilityAttr));
return success();
}
//===----------------------------------------------------------------------===//
// Symbol Interfaces
//===----------------------------------------------------------------------===//
/// Include the generated symbol interfaces.
#include "mlir/IR/SymbolInterfaces.cpp.inc"