[mlir][bytecode] Add bytecode writer config API to skip serialization of resources #71991
Conversation
This file contains hidden or 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
|
@llvm/pr-subscribers-mlir-core Author: Matteo Franciolini (mfrancio) ChangesWhen serializing to bytecode, users can select the option to elide resources from the bytecode file. This will instruct the bytecode writer to serialize only the key and resource kind, while skipping serialization of the data buffer. At parsing, the IR built in memory with valid (but empty) resource handlers. Full diff: https://github.com/llvm/llvm-project/pull/71991.diff 5 Files Affected:
diff --git a/mlir/include/mlir/Bytecode/BytecodeWriter.h b/mlir/include/mlir/Bytecode/BytecodeWriter.h
index b82d8ddad38ed1c..ea4b36832e0bac3 100644
--- a/mlir/include/mlir/Bytecode/BytecodeWriter.h
+++ b/mlir/include/mlir/Bytecode/BytecodeWriter.h
@@ -152,6 +152,9 @@ class BytecodeWriterConfig {
// Resources
//===--------------------------------------------------------------------===//
+ /// Set a boolean flag to skip emission of resources into the bytecode file.
+ void setElideResourceDataFlag(bool shouldElideResourceData = true);
+
/// Attach the given resource printer to the writer configuration.
void attachResourcePrinter(std::unique_ptr<AsmResourcePrinter> printer);
diff --git a/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h b/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
index a1530936f55caee..e255d9fa70b6594 100644
--- a/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
+++ b/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
@@ -82,6 +82,9 @@ class MlirOptMainConfig {
return *this;
}
bool shouldEmitBytecode() const { return emitBytecodeFlag; }
+ bool shouldElideResourceDataFromBytecode() const {
+ return elideResourceDataFromBytecodeFlag;
+ }
/// Set the IRDL file to load before processing the input.
MlirOptMainConfig &setIrdlFile(StringRef file) {
@@ -185,6 +188,9 @@ class MlirOptMainConfig {
/// Emit bytecode instead of textual assembly when generating output.
bool emitBytecodeFlag = false;
+ /// Elide resources when generating bytecode.
+ bool elideResourceDataFromBytecodeFlag = false;
+
/// Enable the Debugger action hook: Debugger can intercept MLIR Actions.
bool enableDebuggerActionHookFlag = false;
diff --git a/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp b/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
index 01dcea1ca3848eb..6097f0eda469cd2 100644
--- a/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
+++ b/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
@@ -39,6 +39,10 @@ struct BytecodeWriterConfig::Impl {
/// Note: This only differs from kVersion if a specific version is set.
int64_t bytecodeVersion = bytecode::kVersion;
+ /// A flag specifying whether to elide emission of resources into the bytecode
+ /// file.
+ bool shouldElideResourceData = false;
+
/// A map containing dialect version information for each dialect to emit.
llvm::StringMap<std::unique_ptr<DialectVersion>> dialectVersionMap;
@@ -89,6 +93,11 @@ void BytecodeWriterConfig::attachResourcePrinter(
impl->externalResourcePrinters.emplace_back(std::move(printer));
}
+void BytecodeWriterConfig::setElideResourceDataFlag(
+ bool shouldElideResourceData) {
+ impl->shouldElideResourceData = shouldElideResourceData;
+}
+
void BytecodeWriterConfig::setDesiredBytecodeVersion(int64_t bytecodeVersion) {
impl->bytecodeVersion = bytecodeVersion;
}
@@ -1170,22 +1179,25 @@ class ResourceBuilder : public AsmResourceBuilder {
using PostProcessFn = function_ref<void(StringRef, AsmResourceEntryKind)>;
ResourceBuilder(EncodingEmitter &emitter, StringSectionBuilder &stringSection,
- PostProcessFn postProcessFn)
+ PostProcessFn postProcessFn, bool shouldElideData)
: emitter(emitter), stringSection(stringSection),
- postProcessFn(postProcessFn) {}
+ postProcessFn(postProcessFn), shouldElideData(shouldElideData) {}
~ResourceBuilder() override = default;
void buildBlob(StringRef key, ArrayRef<char> data,
uint32_t dataAlignment) final {
- emitter.emitOwnedBlobAndAlignment(data, dataAlignment);
+ if (!shouldElideData)
+ emitter.emitOwnedBlobAndAlignment(data, dataAlignment);
postProcessFn(key, AsmResourceEntryKind::Blob);
}
void buildBool(StringRef key, bool data) final {
- emitter.emitByte(data);
+ if (!shouldElideData)
+ emitter.emitByte(data);
postProcessFn(key, AsmResourceEntryKind::Bool);
}
void buildString(StringRef key, StringRef data) final {
- emitter.emitVarInt(stringSection.insert(data));
+ if (!shouldElideData)
+ emitter.emitVarInt(stringSection.insert(data));
postProcessFn(key, AsmResourceEntryKind::String);
}
@@ -1193,6 +1205,7 @@ class ResourceBuilder : public AsmResourceBuilder {
EncodingEmitter &emitter;
StringSectionBuilder &stringSection;
PostProcessFn postProcessFn;
+ bool shouldElideData = false;
};
} // namespace
@@ -1225,7 +1238,8 @@ void BytecodeWriter::writeResourceSection(Operation *op,
// Builder used to emit resources.
ResourceBuilder entryBuilder(resourceEmitter, stringSection,
- appendResourceOffset);
+ appendResourceOffset,
+ config.shouldElideResourceData);
// Emit the external resource entries.
resourceOffsetEmitter.emitVarInt(config.externalResourcePrinters.size());
diff --git a/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp b/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
index c36afae716b12c5..d7d47619ef4ac98 100644
--- a/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
+++ b/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
@@ -90,6 +90,11 @@ struct MlirOptMainConfigCLOptions : public MlirOptMainConfig {
"emit-bytecode", cl::desc("Emit bytecode when generating output"),
cl::location(emitBytecodeFlag), cl::init(false));
+ static cl::opt<bool, /*ExternalStorage=*/true> elideResourcesFromBytecode(
+ "elide-resource-data-from-bytecode",
+ cl::desc("Elide resources when generating bytecode"),
+ cl::location(elideResourceDataFromBytecodeFlag), cl::init(false));
+
static cl::opt<std::optional<int64_t>, /*ExternalStorage=*/true,
BytecodeVersionParser>
bytecodeVersion(
@@ -385,6 +390,8 @@ performActions(raw_ostream &os,
BytecodeWriterConfig writerConfig(fallbackResourceMap);
if (auto v = config.bytecodeVersionToEmit())
writerConfig.setDesiredBytecodeVersion(*v);
+ if (config.shouldElideResourceDataFromBytecode())
+ writerConfig.setElideResourceDataFlag();
return writeBytecodeToFile(op.get(), os, writerConfig);
}
diff --git a/mlir/test/Bytecode/resources_elision.mlir b/mlir/test/Bytecode/resources_elision.mlir
new file mode 100644
index 000000000000000..5238ae8f1ebc583
--- /dev/null
+++ b/mlir/test/Bytecode/resources_elision.mlir
@@ -0,0 +1,18 @@
+// RUN: mlir-opt -emit-bytecode -elide-resource-data-from-bytecode %s | mlir-opt | FileCheck %s
+
+// CHECK-LABEL: @TestDialectResources
+module @TestDialectResources attributes {
+ bytecode.test = dense_resource<decl_resource> : tensor<2xui32>,
+ bytecode.test2 = dense_resource<resource> : tensor<4xf64>,
+ bytecode.test3 = dense_resource<resource_2> : tensor<4xf64>
+} {}
+
+// CHECK-NOT: dialect_resources
+{-#
+ dialect_resources: {
+ builtin: {
+ resource: "0x08000000010000000000000002000000000000000300000000000000",
+ resource_2: "0x08000000010000000000000002000000000000000300000000000000"
+ }
+ }
+#-}
|
|
@llvm/pr-subscribers-mlir Author: Matteo Franciolini (mfrancio) ChangesWhen serializing to bytecode, users can select the option to elide resources from the bytecode file. This will instruct the bytecode writer to serialize only the key and resource kind, while skipping serialization of the data buffer. At parsing, the IR built in memory with valid (but empty) resource handlers. Full diff: https://github.com/llvm/llvm-project/pull/71991.diff 5 Files Affected:
diff --git a/mlir/include/mlir/Bytecode/BytecodeWriter.h b/mlir/include/mlir/Bytecode/BytecodeWriter.h
index b82d8ddad38ed1c..ea4b36832e0bac3 100644
--- a/mlir/include/mlir/Bytecode/BytecodeWriter.h
+++ b/mlir/include/mlir/Bytecode/BytecodeWriter.h
@@ -152,6 +152,9 @@ class BytecodeWriterConfig {
// Resources
//===--------------------------------------------------------------------===//
+ /// Set a boolean flag to skip emission of resources into the bytecode file.
+ void setElideResourceDataFlag(bool shouldElideResourceData = true);
+
/// Attach the given resource printer to the writer configuration.
void attachResourcePrinter(std::unique_ptr<AsmResourcePrinter> printer);
diff --git a/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h b/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
index a1530936f55caee..e255d9fa70b6594 100644
--- a/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
+++ b/mlir/include/mlir/Tools/mlir-opt/MlirOptMain.h
@@ -82,6 +82,9 @@ class MlirOptMainConfig {
return *this;
}
bool shouldEmitBytecode() const { return emitBytecodeFlag; }
+ bool shouldElideResourceDataFromBytecode() const {
+ return elideResourceDataFromBytecodeFlag;
+ }
/// Set the IRDL file to load before processing the input.
MlirOptMainConfig &setIrdlFile(StringRef file) {
@@ -185,6 +188,9 @@ class MlirOptMainConfig {
/// Emit bytecode instead of textual assembly when generating output.
bool emitBytecodeFlag = false;
+ /// Elide resources when generating bytecode.
+ bool elideResourceDataFromBytecodeFlag = false;
+
/// Enable the Debugger action hook: Debugger can intercept MLIR Actions.
bool enableDebuggerActionHookFlag = false;
diff --git a/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp b/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
index 01dcea1ca3848eb..6097f0eda469cd2 100644
--- a/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
+++ b/mlir/lib/Bytecode/Writer/BytecodeWriter.cpp
@@ -39,6 +39,10 @@ struct BytecodeWriterConfig::Impl {
/// Note: This only differs from kVersion if a specific version is set.
int64_t bytecodeVersion = bytecode::kVersion;
+ /// A flag specifying whether to elide emission of resources into the bytecode
+ /// file.
+ bool shouldElideResourceData = false;
+
/// A map containing dialect version information for each dialect to emit.
llvm::StringMap<std::unique_ptr<DialectVersion>> dialectVersionMap;
@@ -89,6 +93,11 @@ void BytecodeWriterConfig::attachResourcePrinter(
impl->externalResourcePrinters.emplace_back(std::move(printer));
}
+void BytecodeWriterConfig::setElideResourceDataFlag(
+ bool shouldElideResourceData) {
+ impl->shouldElideResourceData = shouldElideResourceData;
+}
+
void BytecodeWriterConfig::setDesiredBytecodeVersion(int64_t bytecodeVersion) {
impl->bytecodeVersion = bytecodeVersion;
}
@@ -1170,22 +1179,25 @@ class ResourceBuilder : public AsmResourceBuilder {
using PostProcessFn = function_ref<void(StringRef, AsmResourceEntryKind)>;
ResourceBuilder(EncodingEmitter &emitter, StringSectionBuilder &stringSection,
- PostProcessFn postProcessFn)
+ PostProcessFn postProcessFn, bool shouldElideData)
: emitter(emitter), stringSection(stringSection),
- postProcessFn(postProcessFn) {}
+ postProcessFn(postProcessFn), shouldElideData(shouldElideData) {}
~ResourceBuilder() override = default;
void buildBlob(StringRef key, ArrayRef<char> data,
uint32_t dataAlignment) final {
- emitter.emitOwnedBlobAndAlignment(data, dataAlignment);
+ if (!shouldElideData)
+ emitter.emitOwnedBlobAndAlignment(data, dataAlignment);
postProcessFn(key, AsmResourceEntryKind::Blob);
}
void buildBool(StringRef key, bool data) final {
- emitter.emitByte(data);
+ if (!shouldElideData)
+ emitter.emitByte(data);
postProcessFn(key, AsmResourceEntryKind::Bool);
}
void buildString(StringRef key, StringRef data) final {
- emitter.emitVarInt(stringSection.insert(data));
+ if (!shouldElideData)
+ emitter.emitVarInt(stringSection.insert(data));
postProcessFn(key, AsmResourceEntryKind::String);
}
@@ -1193,6 +1205,7 @@ class ResourceBuilder : public AsmResourceBuilder {
EncodingEmitter &emitter;
StringSectionBuilder &stringSection;
PostProcessFn postProcessFn;
+ bool shouldElideData = false;
};
} // namespace
@@ -1225,7 +1238,8 @@ void BytecodeWriter::writeResourceSection(Operation *op,
// Builder used to emit resources.
ResourceBuilder entryBuilder(resourceEmitter, stringSection,
- appendResourceOffset);
+ appendResourceOffset,
+ config.shouldElideResourceData);
// Emit the external resource entries.
resourceOffsetEmitter.emitVarInt(config.externalResourcePrinters.size());
diff --git a/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp b/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
index c36afae716b12c5..d7d47619ef4ac98 100644
--- a/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
+++ b/mlir/lib/Tools/mlir-opt/MlirOptMain.cpp
@@ -90,6 +90,11 @@ struct MlirOptMainConfigCLOptions : public MlirOptMainConfig {
"emit-bytecode", cl::desc("Emit bytecode when generating output"),
cl::location(emitBytecodeFlag), cl::init(false));
+ static cl::opt<bool, /*ExternalStorage=*/true> elideResourcesFromBytecode(
+ "elide-resource-data-from-bytecode",
+ cl::desc("Elide resources when generating bytecode"),
+ cl::location(elideResourceDataFromBytecodeFlag), cl::init(false));
+
static cl::opt<std::optional<int64_t>, /*ExternalStorage=*/true,
BytecodeVersionParser>
bytecodeVersion(
@@ -385,6 +390,8 @@ performActions(raw_ostream &os,
BytecodeWriterConfig writerConfig(fallbackResourceMap);
if (auto v = config.bytecodeVersionToEmit())
writerConfig.setDesiredBytecodeVersion(*v);
+ if (config.shouldElideResourceDataFromBytecode())
+ writerConfig.setElideResourceDataFlag();
return writeBytecodeToFile(op.get(), os, writerConfig);
}
diff --git a/mlir/test/Bytecode/resources_elision.mlir b/mlir/test/Bytecode/resources_elision.mlir
new file mode 100644
index 000000000000000..5238ae8f1ebc583
--- /dev/null
+++ b/mlir/test/Bytecode/resources_elision.mlir
@@ -0,0 +1,18 @@
+// RUN: mlir-opt -emit-bytecode -elide-resource-data-from-bytecode %s | mlir-opt | FileCheck %s
+
+// CHECK-LABEL: @TestDialectResources
+module @TestDialectResources attributes {
+ bytecode.test = dense_resource<decl_resource> : tensor<2xui32>,
+ bytecode.test2 = dense_resource<resource> : tensor<4xf64>,
+ bytecode.test3 = dense_resource<resource_2> : tensor<4xf64>
+} {}
+
+// CHECK-NOT: dialect_resources
+{-#
+ dialect_resources: {
+ builtin: {
+ resource: "0x08000000010000000000000002000000000000000300000000000000",
+ resource_2: "0x08000000010000000000000002000000000000000300000000000000"
+ }
+ }
+#-}
|
joker-eph
approved these changes
Nov 11, 2023
… of resources When serializing to bytecode, users can select the option to elide resources from the bytecode file. This will instruct the bytecode writer to serialize only the key and resource kind, while skipping serialization of the data buffer. At parsing, the IR is built in memory with valid (but empty) resource handlers.
mfrancio
force-pushed
the
dev/mfrancio/skipSerializationOfResourcesIntoBytecode
branch
from
3b02bec to
be01399
Compare
mfrancio
deleted the
dev/mfrancio/skipSerializationOfResourcesIntoBytecode
branch
zahiraam
pushed a commit
to zahiraam/llvm-project
that referenced
this pull request
Nov 20, 2023
… of resources (llvm#71991) When serializing to bytecode, users can select the option to elide resources from the bytecode file. This will instruct the bytecode writer to serialize only the key and resource kind, while skipping serialization of the data buffer. At parsing, the IR is built in memory with valid (but empty) resource handlers.
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.
When serializing to bytecode, users can select the option to elide resources from the bytecode file. This will instruct the bytecode writer to serialize only the key and resource kind, while skipping serialization of the data buffer. At parsing, the IR is built in memory with valid (but empty) resource handlers.