torch-mlir/tools/npcomp-run-mlir/npcomp-run-mlir.cpp

213 lines
7.8 KiB
C++
Raw Normal View History

//===------------------------------------------------------------*- C++ -*-===//
//
// This file is licensed 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
//
//===----------------------------------------------------------------------===//
//
// Utility binary for compiling and running code through the npcomp
// compiler/runtime stack.
//
//===----------------------------------------------------------------------===//
#include "llvm/Support/InitLLVM.h"
#include "mlir/Dialect/StandardOps/IR/Ops.h"
#include "mlir/IR/AsmState.h"
#include "mlir/InitAllDialects.h"
#include "mlir/InitAllPasses.h"
#include "mlir/Parser.h"
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
#include "mlir/Pass/PassManager.h"
#include "npcomp-c/InitLLVM.h"
#include "npcomp/InitAll.h"
#include "npcomp/RefBackend/JITHelpers/JITModule.h"
using namespace mlir;
using llvm::Error;
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
using llvm::Expected;
using llvm::StringError;
using llvm::Twine;
/// Wrap a string into an llvm::StringError.
static Error make_string_error(const Twine &message) {
return llvm::make_error<StringError>(message.str(),
llvm::inconvertibleErrorCode());
}
static Expected<refbackrt::Ref<refbackrt::Tensor>>
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
convertAttrToTensor(Attribute attr) {
auto type = attr.getType().dyn_cast<RankedTensorType>();
if (!type)
return make_string_error("unhandled argument type; must be a tensor type");
auto extents = llvm::to_vector<6>(llvm::map_range(
type.getShape(), [](int64_t x) { return static_cast<std::int32_t>(x); }));
auto elementType = type.getElementType();
if (auto denseFp = attr.dyn_cast<DenseFPElementsAttr>()) {
if (elementType.isF32()) {
auto values = llvm::to_vector<100>(llvm::map_range(
denseFp, [](APFloat f) { return f.convertToFloat(); }));
return refbackrt::Tensor::create(
refbackrt::ArrayRef<std::int32_t>(extents.data(), extents.size()),
refbackrt::ElementType::F32, static_cast<void *>(values.data()));
}
}
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
return make_string_error("unhandled argument");
}
static Expected<SmallVector<refbackrt::Ref<refbackrt::Tensor>, 6>>
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
createInputs(ArrayRef<StringRef> argValues) {
MLIRContext context;
SmallVector<refbackrt::Ref<refbackrt::Tensor>, 6> ret;
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
for (auto argValue : argValues) {
auto attr = parseAttribute(argValue, &context);
if (!attr)
return make_string_error(Twine("could not parse arg value: ") + argValue);
auto expectedTensor = convertAttrToTensor(attr);
if (!expectedTensor)
return expectedTensor.takeError();
ret.push_back(std::move(*expectedTensor));
}
return ret;
}
static Type convertToMLIRType(refbackrt::ElementType type, Builder &builder) {
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
switch (type) {
case refbackrt::ElementType::F32:
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
return builder.getF32Type();
}
}
static RankedTensorType getCorrespondingMLIRTensorType(refbackrt::Tensor &tensor,
Builder &builder) {
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
auto elementType = convertToMLIRType(tensor.getElementType(), builder);
SmallVector<int64_t, 6> extents;
for (int i = 0, e = tensor.getRank(); i < e; i++)
extents.push_back(tensor.getExtent(i));
return RankedTensorType::get(extents, elementType);
}
static Attribute convertToMLIRAttribute(refbackrt::Tensor &tensor,
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
Builder &builder) {
RankedTensorType type = getCorrespondingMLIRTensorType(tensor, builder);
switch (tensor.getElementType()) {
case refbackrt::ElementType::F32: {
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
SmallVector<float, 100> values;
auto *basePtr = tensor.getData<float>();
for (int i = 0, e = type.getNumElements(); i < e; i++)
values.push_back(basePtr[i]);
return DenseFPElementsAttr::get(type, values);
}
}
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
}
static void printOutput(refbackrt::Tensor &tensor, llvm::raw_ostream &os) {
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
MLIRContext context;
Builder builder(&context);
auto attr = convertToMLIRAttribute(tensor, builder);
attr.print(os);
}
static void printOutputs(ArrayRef<refbackrt::Ref<refbackrt::Tensor>> outputs,
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
llvm::raw_ostream &os) {
for (auto output : llvm::enumerate(outputs)) {
os << "output #" << output.index() << ": ";
printOutput(*output.value(), os);
os << "\n";
}
}
Error compileAndRun(std::string mlirFile, mlir::DialectRegistry &registry,
std::string invokeFunction, ArrayRef<StringRef> argValues,
ArrayRef<StringRef> sharedLibs, bool optimize) {
MLIRContext context;
registry.loadAll(&context);
OwningModuleRef moduleRef = parseSourceFile(mlirFile, &context);
if (!moduleRef)
return make_string_error(Twine("could not open ") + mlirFile);
ModuleOp module = *moduleRef;
// Compile.
PassManager pm(module.getContext(), /*verifyPasses=*/true);
applyPassManagerCLOptions(pm);
refback::JITModule::buildBackendCompilationPipeline(pm, optimize);
if (failed(pm.run(module))) {
return make_string_error(Twine("error compiling to jit backend"));
}
auto expectedJitModule =
refback::JITModule::fromCompiledModule(module, sharedLibs);
Rework e2e flow to use new "npcomprt" This ~totally reworks the existing "runtime" stuff to be more principled and usable, such as from Python. It's still not fully production-quality, mainly in the department of memory management (e.g. it currently leaks memory; we need to figure out "who frees memrefs" + the analysis and transformation needed to do that (maybe use upstream buffer allocation pass?)). The user API is in include/npcomp/runtime/UserAPI.h, though include/npcomp/JITRuntime/JITModule.h is a friendlier wrapper. The stuff under {include,lib}/runtime is totally firewalled from the compiler and tiny (<6kB, though no attention has gone into optimizing that size). For example, we don't link in libSupport into the runtime, instead having our own bare bones replacements for basics like ArrayRef (the JITRuntime helps with bridging that gap, since it *can* depend on all common LLVM utilities). The overall features of npcomprt is that it exposes a module that with multiple function entry points. Each function has arguments and results that are tensor-valued, and npcomprt::Tensor is the runtime type that is used to interact with that (and a npcomprt::Ref<T> reference-counting wrapper is provided to wrap npcomprt::Tensor in the common case). From an implementation perspective, an npcomprt module at the LLVM/object/binary level exposes a single module descriptor struct that has pointers to other metadata (currently just a list of function metadata descriptors). All interactions with the npcomp runtime are keyed off of that module descriptor, including function lookups and dispatching. This is done to dodge platform ABI issues and also allow enough reflection to e.g. verify provided arguments. Most of the compiler-side work here was in LowerToNpcomprtABI and LowerToLLVM. Also, - Rename npcomp_rt/NpcompRt to npcomprt/Npcomprt; it was getting annoying to type the underscores/caps. - misc improvements to bash_helpers.sh
2020-07-09 08:15:40 +08:00
if (!expectedJitModule)
return expectedJitModule.takeError();
auto jitModule = std::move(*expectedJitModule);
auto expectedInputs = createInputs(argValues);
if (!expectedInputs)
return expectedInputs.takeError();
auto expectedOutputs = jitModule->invoke(invokeFunction, *expectedInputs);
if (!expectedOutputs)
return expectedOutputs.takeError();
auto outputs = std::move(*expectedOutputs);
printOutputs(outputs, llvm::outs());
llvm::outs() << "SUCCESS\n";
return Error::success();
}
//===----------------------------------------------------------------------===//
// Main-related init and option parsing.
//===----------------------------------------------------------------------===//
namespace {
namespace cl = llvm::cl;
struct Options {
cl::opt<std::string> inputFile{
cl::Positional, cl::desc("the input .mlir file"), cl::init("-")};
cl::opt<std::string> invokeFunction{"invoke", cl::Required,
cl::desc("function to invoke")};
cl::list<std::string> argValues{"arg-value", cl::ZeroOrMore,
cl::desc("Arguments to the called function")};
cl::list<std::string> sharedLibs{"shared-libs", cl::ZeroOrMore,
cl::MiscFlags::CommaSeparated,
cl::desc("Libraries to link dynamically")};
cl::opt<bool> optimize{
"optimize", cl::Optional,
cl::desc("whether the refback pass pipeline should run optimizations"),
cl::init(false)};
};
} // namespace
int main(int argc, char **argv) {
mlir::DialectRegistry registry;
mlir::registerAllDialects(registry);
mlir::registerAllPasses();
mlir::NPCOMP::registerAllDialects(registry);
mlir::NPCOMP::registerAllPasses();
llvm::InitLLVM y(argc, argv);
npcompInitializeLLVMCodegen();
mlir::registerAsmPrinterCLOptions();
mlir::registerPassManagerCLOptions();
Options options;
llvm::cl::ParseCommandLineOptions(argc, argv, "npcomp compile+run utility\n");
SmallVector<StringRef, 6> sharedLibs(options.sharedLibs.begin(),
options.sharedLibs.end());
SmallVector<StringRef, 6> argValues(options.argValues.begin(),
options.argValues.end());
Error error =
compileAndRun(options.inputFile, registry, options.invokeFunction,
argValues, sharedLibs, options.optimize);
int exitCode = EXIT_SUCCESS;
llvm::handleAllErrors(std::move(error),
[&exitCode](const llvm::ErrorInfoBase &info) {
llvm::errs() << "Error: ";
info.log(llvm::errs());
llvm::errs() << '\n';
exitCode = EXIT_FAILURE;
});
return exitCode;
}