[MLIR][DLTI][Transform] Introduce transform.dlti.query #101561
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
6164a5d to
2454ee2
Compare
|
Hi @rengolin, @ftynse, @banach-space, @joker-eph, @adam-smnk & @Dinistro, as I think you are interested in making progress on making use of DLTI's target descriptors, I thought to ping you on this. If any of you could help with review, that would be appreciated - thanks! |
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-mlir-dlti Author: Rolf Morel (rolfmorel) ChangesThis transform op makes it possible to query attributes associated to IR by means of the DLTI dialect. The op takes both a This op uses new free-standing functions in the Patch is 27.65 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/101561.diff 11 Files Affected:
diff --git a/mlir/include/mlir/Dialect/DLTI/CMakeLists.txt b/mlir/include/mlir/Dialect/DLTI/CMakeLists.txt
index 5b0cb6c37bfab..4f8382e8e6e6b 100644
--- a/mlir/include/mlir/Dialect/DLTI/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/DLTI/CMakeLists.txt
@@ -1,3 +1,5 @@
+add_subdirectory(TransformOps)
+
add_mlir_dialect(DLTI dlti)
add_mlir_doc(DLTIAttrs DLTIDialect Dialects/ -gen-dialect-doc)
diff --git a/mlir/include/mlir/Dialect/DLTI/DLTI.h b/mlir/include/mlir/Dialect/DLTI/DLTI.h
index f50a654f3885d..4061c98bf9eca 100644
--- a/mlir/include/mlir/Dialect/DLTI/DLTI.h
+++ b/mlir/include/mlir/Dialect/DLTI/DLTI.h
@@ -22,6 +22,12 @@ namespace detail {
class DataLayoutEntryAttrStorage;
} // namespace detail
} // namespace mlir
+namespace mlir {
+namespace dlti {
+DataLayoutSpecInterface getDataLayoutSpec(Operation *op);
+TargetSystemSpecInterface getTargetSystemSpec(Operation *op);
+} // namespace dlti
+} // namespace mlir
#define GET_ATTRDEF_CLASSES
#include "mlir/Dialect/DLTI/DLTIAttrs.h.inc"
diff --git a/mlir/include/mlir/Dialect/DLTI/TransformOps/CMakeLists.txt b/mlir/include/mlir/Dialect/DLTI/TransformOps/CMakeLists.txt
new file mode 100644
index 0000000000000..1188d1a9b24d6
--- /dev/null
+++ b/mlir/include/mlir/Dialect/DLTI/TransformOps/CMakeLists.txt
@@ -0,0 +1,6 @@
+set(LLVM_TARGET_DEFINITIONS DLTITransformOps.td)
+mlir_tablegen(DLTITransformOps.h.inc -gen-op-decls)
+mlir_tablegen(DLTITransformOps.cpp.inc -gen-op-defs)
+add_public_tablegen_target(MLIRDLTITransformOpsIncGen)
+
+add_mlir_doc(DLTITransformOps DLTITransformOps Dialects/ -gen-op-doc)
diff --git a/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h b/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h
new file mode 100644
index 0000000000000..970f2597ef7ad
--- /dev/null
+++ b/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h
@@ -0,0 +1,38 @@
+//===- DLTITransformOps.h - DLTI transform ops ------------------*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_DIALECT_DLTI_TRANSFORMOPS_DLTITRANSFORMOPS_H
+#define MLIR_DIALECT_DLTI_TRANSFORMOPS_DLTITRANSFORMOPS_H
+
+#include "mlir/Dialect/Transform/IR/TransformAttrs.h"
+#include "mlir/Dialect/Transform/IR/TransformDialect.h"
+#include "mlir/Dialect/Transform/IR/TransformTypes.h"
+#include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.h"
+
+namespace mlir {
+namespace transform {
+class QueryOp;
+} // namespace transform
+} // namespace mlir
+
+namespace mlir {
+class DialectRegistry;
+
+namespace dlti {
+void registerTransformDialectExtension(DialectRegistry ®istry);
+} // namespace dlti
+} // namespace mlir
+
+////===----------------------------------------------------------------------===//
+//// DLTI Transform Operations
+////===----------------------------------------------------------------------===//
+
+#define GET_OP_CLASSES
+#include "mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h.inc"
+
+#endif // MLIR_DIALECT_DLTI_TRANSFORMOPS_DLTITRANSFORMOPS_H
diff --git a/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.td b/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.td
new file mode 100644
index 0000000000000..438912a6b2623
--- /dev/null
+++ b/mlir/include/mlir/Dialect/DLTI/TransformOps/DLTITransformOps.td
@@ -0,0 +1,61 @@
+//===- DLTITransformOps.td - DLTI transform ops ------------*- tablegen -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef DLTI_TRANSFORM_OPS
+#define DLTI_TRANSFORM_OPS
+
+include "mlir/Dialect/Transform/IR/TransformDialect.td"
+include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.td"
+include "mlir/Dialect/Transform/IR/TransformTypes.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
+include "mlir/IR/OpBase.td"
+
+def QueryOp : Op<Transform_Dialect, "dlti.query", [
+ TransformOpInterface, TransformEachOpTrait,
+ DeclareOpInterfaceMethods<MemoryEffectsOpInterface>
+]> {
+ let summary = "Return attribute (as param) associated to key via DTLI";
+ let description = [{
+ This op queries data layout and target information associated to payload
+ IR by way of the DLTI dialect. A lookup is performed for the given `key`
+ at the `target` op, with the DLTI dialect determining which interfaces and
+ attributes are consulted.
+
+ When only `key` is provided, the lookup occurs with respect to the data
+ layout specification of DLTI. When `device` is provided, the lookup occurs
+ with respect to DLTI's target device specifications associated to a DLTI
+ system device specification.
+
+ #### Return modes
+
+ When succesfull, the result, `associated_attr`, associates one attribute
+ as a param for each op in `target`'s payload.
+
+ If the lookup fails - as DLTI specifications or entries with the right
+ names are missing (i.e. the values of `device` and `key`) - a definite
+ failure is returned.
+ }];
+
+ let arguments = (ins TransformHandleTypeInterface:$target,
+ OptionalAttr<StrAttr>:$device,
+ StrAttr:$key);
+ let results = (outs TransformParamTypeInterface:$associated_attr);
+ let assemblyFormat =
+ "(`:``:` $device^ `:``:`)? $key `at` $target attr-dict `:`"
+ "functional-type(operands, results)";
+
+ let extraClassDeclaration = [{
+ ::mlir::DiagnosedSilenceableFailure applyToOne(
+ ::mlir::transform::TransformRewriter &rewriter,
+ ::mlir::Operation *target,
+ ::mlir::transform::ApplyToEachResultList &results,
+ TransformState &state);
+ }];
+}
+
+#endif // DLTI_TRANSFORM_OPS
diff --git a/mlir/include/mlir/InitAllExtensions.h b/mlir/include/mlir/InitAllExtensions.h
index 20a4ab6f18a28..0adc5e52f2a0e 100644
--- a/mlir/include/mlir/InitAllExtensions.h
+++ b/mlir/include/mlir/InitAllExtensions.h
@@ -25,6 +25,7 @@
#include "mlir/Conversion/UBToLLVM/UBToLLVM.h"
#include "mlir/Dialect/Affine/TransformOps/AffineTransformOps.h"
#include "mlir/Dialect/Bufferization/TransformOps/BufferizationTransformOps.h"
+#include "mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h"
#include "mlir/Dialect/Func/Extensions/AllExtensions.h"
#include "mlir/Dialect/Func/TransformOps/FuncTransformOps.h"
#include "mlir/Dialect/GPU/TransformOps/GPUTransformOps.h"
@@ -69,6 +70,7 @@ inline void registerAllExtensions(DialectRegistry ®istry) {
// Register all transform dialect extensions.
affine::registerTransformDialectExtension(registry);
bufferization::registerTransformDialectExtension(registry);
+ dlti::registerTransformDialectExtension(registry);
func::registerTransformDialectExtension(registry);
gpu::registerTransformDialectExtension(registry);
linalg::registerTransformDialectExtension(registry);
diff --git a/mlir/lib/Dialect/DLTI/CMakeLists.txt b/mlir/lib/Dialect/DLTI/CMakeLists.txt
index f74f5f8013090..7691a8a10a37a 100644
--- a/mlir/lib/Dialect/DLTI/CMakeLists.txt
+++ b/mlir/lib/Dialect/DLTI/CMakeLists.txt
@@ -1,3 +1,4 @@
+add_subdirectory(TransformOps)
add_mlir_dialect_library(MLIRDLTIDialect
DLTI.cpp
Traits.cpp
diff --git a/mlir/lib/Dialect/DLTI/DLTI.cpp b/mlir/lib/Dialect/DLTI/DLTI.cpp
index 420c605d1a19b..488c4c97e819d 100644
--- a/mlir/lib/Dialect/DLTI/DLTI.cpp
+++ b/mlir/lib/Dialect/DLTI/DLTI.cpp
@@ -389,6 +389,41 @@ TargetSystemSpecAttr::verify(function_ref<InFlightDiagnostic()> emitError,
return success();
}
+DataLayoutSpecInterface dlti::getDataLayoutSpec(Operation *op) {
+ DataLayoutSpecInterface dlSpec = nullptr;
+
+ for (Operation *cur = op; cur && !dlSpec; cur = cur->getParentOp()) {
+ if (auto dataLayoutOp = dyn_cast<DataLayoutOpInterface>(cur))
+ dlSpec = dataLayoutOp.getDataLayoutSpec();
+ else if (auto moduleOp = dyn_cast<ModuleOp>(cur))
+ dlSpec = moduleOp.getDataLayoutSpec();
+ else
+ for (NamedAttribute attr : cur->getAttrs())
+ if ((dlSpec = llvm::dyn_cast<DataLayoutSpecInterface>(attr.getValue())))
+ break;
+ }
+
+ return dlSpec;
+}
+
+TargetSystemSpecInterface dlti::getTargetSystemSpec(Operation *op) {
+ TargetSystemSpecInterface sysSpec = nullptr;
+
+ for (Operation *cur = op; cur && !sysSpec; cur = cur->getParentOp()) {
+ if (auto dataLayoutOp = dyn_cast<DataLayoutOpInterface>(cur))
+ sysSpec = dataLayoutOp.getTargetSystemSpec();
+ else if (auto moduleOp = dyn_cast<ModuleOp>(cur))
+ sysSpec = moduleOp.getTargetSystemSpec();
+ else
+ for (NamedAttribute attr : cur->getAttrs())
+ if ((sysSpec =
+ llvm::dyn_cast<TargetSystemSpecInterface>(attr.getValue())))
+ break;
+ }
+
+ return sysSpec;
+}
+
//===----------------------------------------------------------------------===//
// DLTIDialect
//===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/DLTI/TransformOps/CMakeLists.txt b/mlir/lib/Dialect/DLTI/TransformOps/CMakeLists.txt
new file mode 100644
index 0000000000000..05734353f9d11
--- /dev/null
+++ b/mlir/lib/Dialect/DLTI/TransformOps/CMakeLists.txt
@@ -0,0 +1,15 @@
+add_mlir_dialect_library(MLIRDLTITransformOps
+ DLTITransformOps.cpp
+
+ ADDITIONAL_HEADER_DIRS
+ ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/DLTI/TransformOps
+
+ DEPENDS
+ MLIRDLTITransformOpsIncGen
+ MLIRDLTIDialect
+
+ LINK_LIBS PUBLIC
+ MLIRDLTIDialect
+ MLIRSideEffectInterfaces
+ MLIRTransformDialect
+ )
diff --git a/mlir/lib/Dialect/DLTI/TransformOps/DLTITransformOps.cpp b/mlir/lib/Dialect/DLTI/TransformOps/DLTITransformOps.cpp
new file mode 100644
index 0000000000000..dc601046b2a04
--- /dev/null
+++ b/mlir/lib/Dialect/DLTI/TransformOps/DLTITransformOps.cpp
@@ -0,0 +1,98 @@
+
+//===- LinalgTransformOps.cpp - Implementation of Linalg transform ops ----===//
+//
+// 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/Dialect/DLTI/TransformOps/DLTITransformOps.h"
+
+#include "mlir/Analysis/DataLayoutAnalysis.h"
+#include "mlir/Dialect/DLTI/DLTI.h"
+#include "mlir/Dialect/DLTI/TransformOps/DLTITransformOps.h"
+#include "mlir/Dialect/Transform/IR/TransformOps.h"
+#include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.h"
+#include "mlir/Dialect/Transform/Utils/DiagnosedSilenceableFailure.h"
+#include "mlir/Dialect/Transform/Utils/Utils.h"
+#include "mlir/Interfaces/DataLayoutInterfaces.h"
+
+using namespace mlir;
+using namespace mlir::transform;
+
+#define DEBUG_TYPE "dlti-transforms"
+
+//===----------------------------------------------------------------------===//
+// FuseOp
+//===----------------------------------------------------------------------===//
+
+void transform::QueryOp::getEffects(
+ SmallVectorImpl<MemoryEffects::EffectInstance> &effects) {
+ onlyReadsHandle(getTargetMutable(), effects);
+ producesHandle(getOperation()->getOpResults(), effects);
+ onlyReadsPayload(effects);
+}
+
+DiagnosedSilenceableFailure transform::QueryOp::applyToOne(
+ transform::TransformRewriter &rewriter, Operation *target,
+ transform::ApplyToEachResultList &results, TransformState &state) {
+ StringAttr deviceId = getDeviceAttr();
+ StringAttr key = getKeyAttr();
+
+ DataLayoutEntryInterface entry;
+ if (deviceId) {
+ TargetSystemSpecInterface sysSpec = dlti::getTargetSystemSpec(target);
+ if (!sysSpec)
+ return mlir::emitDefiniteFailure(target->getLoc())
+ << "no target system spec associated to: " << target;
+
+ if (auto targetSpec = sysSpec.getDeviceSpecForDeviceID(deviceId))
+ entry = targetSpec->getSpecForIdentifier(key);
+ else
+ return mlir::emitDefiniteFailure(target->getLoc())
+ << "no " << deviceId << " target device spec found";
+ } else {
+ DataLayoutSpecInterface dlSpec = dlti::getDataLayoutSpec(target);
+ if (!dlSpec)
+ return mlir::emitDefiniteFailure(target->getLoc())
+ << "no data layout spec associated to: " << target;
+
+ entry = dlSpec.getSpecForIdentifier(key);
+ }
+
+ if (!entry)
+ return mlir::emitDefiniteFailure(target->getLoc())
+ << "no DLTI entry for key: " << key;
+
+ results.push_back(entry.getValue());
+
+ return DiagnosedSilenceableFailure::success();
+}
+
+//===----------------------------------------------------------------------===//
+// Transform op registration
+//===----------------------------------------------------------------------===//
+
+namespace {
+class DLTITransformDialectExtension
+ : public transform::TransformDialectExtension<
+ DLTITransformDialectExtension> {
+public:
+ using Base::Base;
+
+ void init() {
+ registerTransformOps<
+#define GET_OP_LIST
+#include "mlir/Dialect/DLTI/TransformOps/DLTITransformOps.cpp.inc"
+ >();
+ }
+};
+} // namespace
+
+#define GET_OP_CLASSES
+#include "mlir/Dialect/DLTI/TransformOps/DLTITransformOps.cpp.inc"
+
+void mlir::dlti::registerTransformDialectExtension(DialectRegistry ®istry) {
+ registry.addExtensions<DLTITransformDialectExtension>();
+}
diff --git a/mlir/test/Dialect/DLTI/query.mlir b/mlir/test/Dialect/DLTI/query.mlir
new file mode 100644
index 0000000000000..2504958de0928
--- /dev/null
+++ b/mlir/test/Dialect/DLTI/query.mlir
@@ -0,0 +1,296 @@
+// RUN: mlir-opt -transform-interpreter -canonicalize -split-input-file -verify-diagnostics %s | FileCheck %s
+
+// expected-remark @below {{associated attr 42 : i32}}
+module attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 42 : i32>>} {
+ func.func private @f()
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %funcs = transform.structured.match ops{["func.func"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %module = transform.get_parent_op %funcs : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query "test.id" at %module : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %module : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+module attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 42 : i32>>} {
+ // expected-remark @below {{associated attr 24 : i32}}
+ func.func private @f() attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 24 : i32>>}
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %funcs = transform.structured.match ops{["func.func"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query "test.id" at %funcs : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %funcs : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+// expected-remark @below {{associated attr 42 : i32}}
+module attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 42 : i32>>} {
+ func.func private @f() attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 24 : i32>>}
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %funcs = transform.structured.match ops{["func.func"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %module = transform.get_parent_op %funcs : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query "test.id" at %module : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %module : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+module attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 42 : i32>>} {
+ func.func @matmul_tensors(
+ %arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32>, %arg2: tensor<?x?xf32>)
+ -> tensor<?x?xf32> {
+ // expected-remark @below {{associated attr 42 : i32}}
+ %0 = linalg.matmul ins(%arg0, %arg1: tensor<?x?xf32>, tensor<?x?xf32>)
+ outs(%arg2: tensor<?x?xf32>)
+ -> tensor<?x?xf32>
+ return %0 : tensor<?x?xf32>
+ }
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %matmul = transform.structured.match ops{["linalg.matmul"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query "test.id" at %matmul : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %matmul : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+module attributes { test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 42 : i32>>} {
+ func.func @matmul_tensors(
+ %arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32>, %arg2: tensor<?x?xf32>)
+ -> tensor<?x?xf32> attributes {test.dlti = #dlti.dl_spec<#dlti.dl_entry<"test.id", 24 : i32>>} {
+ // expected-remark @below {{associated attr 24 : i32}}
+ %0 = linalg.matmul ins(%arg0, %arg1: tensor<?x?xf32>, tensor<?x?xf32>)
+ outs(%arg2: tensor<?x?xf32>)
+ -> tensor<?x?xf32>
+ return %0 : tensor<?x?xf32>
+ }
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %matmul = transform.structured.match ops{["linalg.matmul"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query "test.id" at %matmul : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %matmul : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+// expected-remark @below {{associated attr 42 : i32}}
+module attributes { test.dlti = #dlti.target_system_spec<"CPU": #dlti.target_device_spec<#dlti.dl_entry<"test.id", 42 : i32>>>} {
+ func.func private @f()
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %func = transform.structured.match ops{["func.func"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %module = transform.get_parent_op %func : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query ::"CPU"::"test.id" at %module : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %module : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+module attributes { test.dlti = #dlti.target_system_spec<"CPU": #dlti.target_device_spec<#dlti.dl_entry<"test.id", 42 : i32>>,
+ "GPU": #dlti.target_device_spec<#dlti.dl_entry<"test.id", 43 : i32>>>} {
+ // expected-remark @below {{associated attr 43 : i32}}
+ func.func private @f()
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg: !transform.any_op) {
+ %func = transform.structured.match ops{["func.func"]} in %arg : (!transform.any_op) -> !transform.any_op
+ %param = transform.dlti.query ::"GPU"::"test.id" at %func : (!transform.any_op) -> !transform.any_param
+ transform.debug.emit_param_as_remark %param, "associated attr" at %func : !transform.any_param, !transform.any_op
+ transform.yield
+ }
+}
+
+// -----
+
+module attributes { test.dlti = #dlti.target_system_spec<"CPU": #dlti.target_device_spec<#dlti.dl_entry<"test.id", 42 : i32>>,
+ "GPU": #dlti.target_device_spec<#dlti.dl_entry<"test.id", 43 : i32>>>} {
+ // expected-remark @below {{associated attr 24 : i32}}
+ func.func private @f() attributes { test.dlti = #dlti.target_system_spec<"CPU": #dlti.target_device_spec<#dlti.dl_entry<"tes...
[truncated]
|
f6e8061 to
7448ef1
Compare
rengolin
reviewed
Aug 6, 2024
| at the `target` op, with the DLTI dialect determining which interfaces and | ||
| attributes are consulted. | ||
|
|
||
| When only `key` is provided, the lookup occurs with respect to the data |
There was a problem hiding this comment.
This is a currently arbitrary semantics, but I can't see a better way to encode this.
@ftynse we could probably add a similar root entry to DL like we have for TI. Then there would be no ambiguity.
Not that I like the root entry itself, but it does simplify things a lot.
There was a problem hiding this comment.
With the current state of DLTI, Data Layout info is always keyed by a single name. For querying Target Information, you must always specify a device ID in addition to a key to look up w.r.t. that device. So currently, whether a second key, deviceId, is provided is enough to indicate which type of DLTI info the user wants to query. (If this were no longer to be the case going forward, I would advocate for having both types of DLTI info implement a DLTIQueryInterface, or the like, probably supporting any number of keys so that nested dictionary lookups become possible.)
As for the ::$deviceId::$key syntax: I too would prefer for the leading :: to go. It's currently there to help the parser. With a custom parser, I am sure $deviceId::$key syntax is possible as well. As more changes for DLTI are planned, I was hoping to address this in later PRs (e.g. when would allow nested dictionary lookups).
There was a problem hiding this comment.
I don't mind there being a root entry.
banach-space
approved these changes
Aug 6, 2024
|
Thanks for checking it over, @banach-space and @rengolin. I think I'll wait one more day and then ask that the PR is merged. |
c4e08cd to
7ab0a58
Compare
This transform op makes it possible to query attributes associated to IR by means of the DLTI dialect. The op takes both a `key` and a target `op` to perform the query at. Facility functions automatically find the closest ancestor op which defines the appropriate DLTI interface or has an attribute implementing a DLTI interface. By default the lookup uses the data layout interfaces of DLTI. If the optional `device` parameter is provided, the lookup happens with respect to the interfaces for TargetSystemSpec and TargetDeviceSpec. This op uses new free-standing functions in the `dlti` namespace to not only look up specifications via the DataLayoutSpecOpInterface and on ModuleOps but also on any ancestor op that has an appropriate DLTI attribute.
7ab0a58 to
7491f66
Compare
|
I am happy for the PR to be merged. As such, I have squashed the documentation fix-up commit. If somebody, e.g. @banach-space or @rengolin, could help with merging, that would be much appreciated! |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/177/builds/2914 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/129/builds/3431 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/138/builds/2276 Here is the relevant piece of the build log for the reference: |
|
FYI, we're aware of the issue and will revert soon if there is no quick fix. |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/169/builds/1957 Here is the relevant piece of the build log for the reference: |
rengolin
added a commit
that referenced
this pull request
Aug 9, 2024
…)" This reverts commit 8f21ff9. Crashed CI builds
|
With weekend coming, I just reverted. Let's pick this up Monday. |
|
The issue seems to be related to this change: #101119. "Existing DialectExtension subclasses will need a type ID to be registered for each subclass. " |
|
Thanks, I just hit the same issue when I'm doing an integrate. I'll advance it ahead of the revert. |
|
Apologies all. I should have rebased locally before declaring the PR was ready to go. Lesson learned.
Indeed, that commit introduced a change to how dialect extensions should be registered (the PR's title captures it well). I have it working locally now. Will submit another PR for a second go (EDIT: #102652). The change that I needed was: --- a/mlir/lib/Dialect/DLTI/TransformOps/DLTITransformOps.cpp
+++ b/mlir/lib/Dialect/DLTI/TransformOps/DLTITransformOps.cpp
@@ -75,6 +75,9 @@ class DLTITransformDialectExtension
: public transform::TransformDialectExtension<
DLTITransformDialectExtension> {
public:
+ MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(DLTITransformDialectExtension)
+
using Base::Base;
void init() |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/35/builds/1748 Here is the relevant piece of the build log for the reference: |
rolfmorel
added a commit
to rolfmorel/llvm-project
that referenced
this pull request
Aug 9, 2024
This transform op makes it possible to query attributes associated to IR by means of the DLTI dialect. The op takes both a `key` and a target `op` to perform the query at. Facility functions automatically find the closest ancestor op which defines the appropriate DLTI interface or has an attribute implementing a DLTI interface. By default the lookup uses the data layout interfaces of DLTI. If the optional `device` parameter is provided, the lookup happens with respect to the interfaces for TargetSystemSpec and TargetDeviceSpec. This op uses new free-standing functions in the `dlti` namespace to not only look up specifications via the `DataLayoutSpecOpInterface` and on `ModuleOp`s but also on any ancestor op that has an appropriate DLTI attribute.
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/153/builds/5509 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/164/builds/1879 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/25/builds/1519 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/17/builds/1852 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/4/builds/1236 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/143/builds/1330 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/41/builds/1166 Here is the relevant piece of the build log for the reference: |
kutemeikito
added a commit
to kutemeikito/llvm-project
that referenced
this pull request
Aug 10, 2024
* 'main' of https://github.com/llvm/llvm-project: (700 commits) [SandboxIR][NFC] SingleLLVMInstructionImpl class (llvm#102687) [ThinLTO]Clean up 'import-assume-unique-local' flag. (llvm#102424) [nsan] Make #include more conventional [SandboxIR][NFC] Use Tracker.emplaceIfTracking() [libc] Moved range_reduction_double ifdef statement (llvm#102659) [libc] Fix CFP long double and add tests (llvm#102660) [TargetLowering] Handle vector types in expandFixedPointMul (llvm#102635) [compiler-rt][NFC] Replace environment variable with %t (llvm#102197) [UnitTests] Convert a test to use opaque pointers (llvm#102668) [CodeGen][NFCI] Don't re-implement parts of ASTContext::getIntWidth (llvm#101765) [SandboxIR] Clean up tracking code with the help of emplaceIfTracking() (llvm#102406) [mlir][bazel] remove extra blanks in mlir-tblgen test [NVPTX][NFC] Update tests to use bfloat type (llvm#101493) [mlir] Add support for parsing nested PassPipelineOptions (llvm#101118) [mlir][bazel] add missing td dependency in mlir-tblgen test [flang][cuda] Fix lib dependency [libc] Clean up remaining use of *_WIDTH macros in printf (llvm#102679) [flang][cuda] Convert cuf.alloc for box to fir.alloca in device context (llvm#102662) [SandboxIR] Implement the InsertElementInst class (llvm#102404) [libc] Fix use of cpp::numeric_limits<...>::digits (llvm#102674) [mlir][ODS] Verify type constraints in Types and Attributes (llvm#102326) [LTO] enable `ObjCARCContractPass` only on optimized build (llvm#101114) [mlir][ODS] Consistent `cppType` / `cppClassName` usage (llvm#102657) [lldb] Move definition of SBSaveCoreOptions dtor out of header (llvm#102539) [libc] Use cpp::numeric_limits in preference to C23 <limits.h> macros (llvm#102665) [clang] Implement -fptrauth-auth-traps. (llvm#102417) [LLVM][rtsan] rtsan transform to preserve CFGAnalyses (llvm#102651) Revert "[AMDGPU] Move `AMDGPUAttributorPass` to full LTO post link stage (llvm#102086)" [RISCV][GISel] Add missing tests for G_CTLZ/CTTZ instruction selection. NFC Return available function types for BindingDecls. (llvm#102196) [clang] Wire -fptrauth-returns to "ptrauth-returns" fn attribute. (llvm#102416) [RISCV] Remove riscv-experimental-rv64-legal-i32. (llvm#102509) [RISCV] Move PseudoVSET(I)VLI expansion to use PseudoInstExpansion. (llvm#102496) [NVPTX] support switch statement with brx.idx (reland) (llvm#102550) [libc][newhdrgen]sorted function names in yaml (llvm#102544) [GlobalIsel] Combine G_ADD and G_SUB with constants (llvm#97771) Suppress spurious warnings due to R_RISCV_SET_ULEB128 [scudo] Separated committed and decommitted entries. (llvm#101409) [MIPS] Fix missing ANDI optimization (llvm#97689) [Clang] Add env var for nvptx-arch/amdgpu-arch timeout (llvm#102521) [asan] Switch allocator to dynamic base address (llvm#98511) [AMDGPU] Move `AMDGPUAttributorPass` to full LTO post link stage (llvm#102086) [libc][math][c23] Add fadd{l,f128} C23 math functions (llvm#102531) [mlir][bazel] revert bazel rule change for DLTITransformOps [msan] Support vst{2,3,4}_lane instructions (llvm#101215) Revert "[MLIR][DLTI][Transform] Introduce transform.dlti.query (llvm#101561)" [X86] pr57673.ll - generate MIR test checks [mlir][vector][test] Split tests from vector-transfer-flatten.mlir (llvm#102584) [mlir][bazel] add bazel rule for DLTITransformOps OpenMPOpt: Remove dead include [IR] Add method to GlobalVariable to change type of initializer. (llvm#102553) [flang][cuda] Force default allocator in device code (llvm#102238) [llvm] Construct SmallVector<SDValue> with ArrayRef (NFC) (llvm#102578) [MLIR][DLTI][Transform] Introduce transform.dlti.query (llvm#101561) [AMDGPU][AsmParser][NFC] Remove a misleading comment. (llvm#102604) [Arm][AArch64][Clang] Respect function's branch protection attributes. (llvm#101978) [mlir] Verifier: steal bit to track seen instead of set. (llvm#102626) [Clang] Fix Handling of Init Capture with Parameter Packs in LambdaScopeForCallOperatorInstantiationRAII (llvm#100766) [X86] Convert truncsat clamping patterns to use SDPatternMatch. NFC. [gn] Give two scripts argparse.RawDescriptionHelpFormatter [bazel] Add missing dep for the SPIRVToLLVM target [Clang] Simplify specifying passes via -Xoffload-linker (llvm#102483) [bazel] Port for d45de80 [SelectionDAG] Use unaligned store/load to move AVX registers onto stack for `insertelement` (llvm#82130) [Clang][OMPX] Add the code generation for multi-dim `num_teams` (llvm#101407) [ARM] Regenerate big-endian-vmov.ll. NFC [AMDGPU][AsmParser][NFCI] All NamedIntOperands to be of the i32 type. (llvm#102616) [libc][math][c23] Add totalorderl function. (llvm#102564) [mlir][spirv] Support `memref` in `convert-to-spirv` pass (llvm#102534) [MLIR][GPU-LLVM] Convert `gpu.func` to `llvm.func` (llvm#101664) Fix a unit test input file (llvm#102567) [llvm-readobj][COFF] Dump hybrid objects for ARM64X files. (llvm#102245) AMDGPU/NewPM: Port SIFixSGPRCopies to new pass manager (llvm#102614) [MemoryBuiltins] Simplify getCalledFunction() helper (NFC) [AArch64] Add invalid 1 x vscale costs for reductions and reduction-operations. (llvm#102105) [MemoryBuiltins] Handle allocator attributes on call-site LSV/test/AArch64: add missing lit.local.cfg; fix build (llvm#102607) Revert "Enable logf128 constant folding for hosts with 128bit floats (llvm#96287)" [RISCV] Add Syntacore SCR5 RV32/64 processors definition (llvm#102285) [InstCombine] Remove unnecessary RUN line from test (NFC) [flang][OpenMP] Handle multiple ranges in `num_teams` clause (llvm#102535) [mlir][vector] Add tests for scalable vectors in one-shot-bufferize.mlir (llvm#102361) [mlir][vector] Disable `vector.matrix_multiply` for scalable vectors (llvm#102573) [clang] Implement CWG2627 Bit-fields and narrowing conversions (llvm#78112) [NFC] Use references to avoid copying (llvm#99863) Revert "[mlir][ArmSME] Pattern to swap shape_cast(tranpose) with transpose(shape_cast) (llvm#100731)" (llvm#102457) [IRBuilder] Generate nuw GEPs for struct member accesses (llvm#99538) [bazel] Port for 9b06e25 [CodeGen][NewPM] Improve start/stop pass error message CodeGenPassBuilder (llvm#102591) [AArch64] Implement TRBMPAM_EL1 system register (llvm#102485) [InstCombine] Fixing wrong select folding in vectors with undef elements (llvm#102244) [AArch64] Sink operands to fmuladd. (llvm#102297) LSV: document hang reported in llvm#37865 (llvm#102479) Enable logf128 constant folding for hosts with 128bit floats (llvm#96287) [RISCV][clang] Remove bfloat base type in non-zvfbfmin vcreate (llvm#102146) [RISCV][clang] Add missing `zvfbfmin` to `vget_v` intrinsic (llvm#102149) [mlir][vector] Add mask elimination transform (llvm#99314) [Clang][Interp] Fix display of syntactically-invalid note for member function calls (llvm#102170) [bazel] Port for 3fffa6d [DebugInfo][RemoveDIs] Use iterator-inserters in clang (llvm#102006) ... Signed-off-by: Edwiin Kusuma Jaya <kutemeikito0905@gmail.com>
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/52/builds/1489 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/55/builds/1190 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/85/builds/1045 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/84/builds/612 Here is the relevant piece of the build log for the reference: |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/94/builds/833 Here is the relevant piece of the build log for the reference: |
ftynse
reviewed
Aug 13, 2024
| at the `target` op, with the DLTI dialect determining which interfaces and | ||
| attributes are consulted. | ||
|
|
||
| When only `key` is provided, the lookup occurs with respect to the data |
There was a problem hiding this comment.
I don't mind there being a root entry.
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/24/builds/905 Here is the relevant piece of the build log for the reference: |
This was referenced Aug 16, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This transform op makes it possible to query attributes associated to IR by means of the DLTI dialect.
The op takes both a
keyand a targetopto perform the query at. Facility functions automatically find the closest ancestor op which defines the appropriate DLTI interface or has an attribute implementing a DLTI interface. By default the lookup uses the data layout interfaces of DLTI. If the optionaldeviceparameter is provided, the lookup happens with respect to the interfaces for TargetSystemSpec and TargetDeviceSpec.This op uses new free-standing functions in the
dltinamespace to not only look up specifications via theDataLayoutSpecOpInterfaceand onModuleOps but also on any ancestor op that has an appropriate DLTI attribute.