experimental/ModelBuilder/test/TestSimpleJIT.cpp

// Copyright 2020 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//      https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// clang-format off

// NOLINTNEXTLINE
// RUN: test-simple-jit -runtime-support=$(dirname %s)/runtime-support.so 2>&1 | IreeFileCheck %s

// clang-format on

#include "experimental/ModelBuilder/MemRefUtils.h"
#include "experimental/ModelBuilder/ModelBuilder.h"
#include "experimental/ModelBuilder/ModelRunner.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/InitLLVM.h"
#include "mlir/EDSC/Builders.h"
#include "mlir/EDSC/Intrinsics.h"
#include "mlir/IR/Function.h"
#include "mlir/IR/StandardTypes.h"

using namespace mlir;  // NOLINT

static llvm::cl::opt<std::string> runtimeSupport(
    "runtime-support", llvm::cl::desc("Runtime support library filename"),
    llvm::cl::value_desc("filename"), llvm::cl::init("-"));

// Flush the different output streams. Needed to ensure FileCheck
// sees the various buffered streams in a reasonable order.
static void flush() {
  fflush(stderr);
  fflush(stdout);
  llvm::errs().flush();
  llvm::outs().flush();
}

template <unsigned M, unsigned N>
void testVectorAdd(StringLiteral funcName, unsigned kNumElements) {
  ModelBuilder modelBuilder;

  auto f32 = modelBuilder.f32;
  auto mnVectorType = modelBuilder.getVectorType({M, N}, f32);
  auto typeA = modelBuilder.getMemRefType({kNumElements}, mnVectorType);
  auto typeB = modelBuilder.getMemRefType({kNumElements}, mnVectorType);
  auto typeC = modelBuilder.getMemRefType({kNumElements}, mnVectorType);

  // 1. Build a simple vector_add.
  {
    auto f = modelBuilder.makeFunction(funcName, {}, {typeA, typeB, typeC});
    OpBuilder b(&f.getBody());
    ScopedContext scope(b, f.getLoc());

    StdIndexedValue A(f.getArgument(0)), B(f.getArgument(1)),
        C(f.getArgument(2));
    auto last = std_constant_index(kNumElements - 1);
    C(last) = A(last) + B(last);

    (vector_print(*A(last)));
    (vector_print(*B(last)));
    (vector_print(*C(last)));

    std_ret();
  }

  modelBuilder.getModuleRef()->dump();
  flush();

  // 2. Compile the function, pass in runtime support library
  //    to the execution engine for vector.print.
  ModelRunner runner(modelBuilder.getModuleRef());
  runner.compile(/*llvmOptLevel=*/3, /*llcOptLevel=*/3, runtimeSupport);

  // 3. Allocate data within data structures that interoperate with the MLIR ABI
  // conventions used by codegen.
  auto oneInit = [](unsigned idx, Vector2D<M, N, float> *ptr) {
    float *p = reinterpret_cast<float *>(ptr + idx);
    for (unsigned i = 0; i < M * N; ++i) p[i] = 1.0f;
  };
  auto incInit = [](unsigned idx, Vector2D<M, N, float> *ptr) {
    float *p = reinterpret_cast<float *>(ptr + idx);
    for (unsigned i = 0; i < M * N; ++i) p[i] = 1.0f + i;
  };
  auto zeroInit = [](unsigned idx, Vector2D<M, N, float> *ptr) {
    float *p = reinterpret_cast<float *>(ptr + idx);
    for (unsigned i = 0; i < M * N; ++i) p[i] = 0.0f;
  };
  auto A = makeInitializedStridedMemRefDescriptor<Vector2D<M, N, float>, 1>(
      {kNumElements}, oneInit);
  auto B = makeInitializedStridedMemRefDescriptor<Vector2D<M, N, float>, 1>(
      {kNumElements}, incInit);
  auto C = makeInitializedStridedMemRefDescriptor<Vector2D<M, N, float>, 1>(
      {kNumElements}, zeroInit);

  // 5. Call the funcOp named `funcName`.
  const std::string kFuncAdapterName =
      (llvm::Twine("_mlir_ciface_") + funcName).str();
  auto *bufferA = A.get();
  auto *bufferB = B.get();
  auto *bufferC = C.get();
  void *args[3] = {&bufferA, &bufferB, &bufferC};

  auto err =
      runner.engine->invoke(kFuncAdapterName, MutableArrayRef<void *>{args});
  flush();
  if (err) llvm_unreachable("Error running function.");

  llvm::outs() << "\nSUCCESS\n\n";
  flush();
}

template <unsigned M, unsigned N, unsigned K>
void testMatmulOnVectors(StringLiteral funcName) {
  ModelBuilder modelBuilder;

  auto f32 = modelBuilder.f32;
  auto mkVectorType = modelBuilder.getVectorType({M, K}, f32);
  auto typeA = modelBuilder.getMemRefType({-1, -1}, mkVectorType);
  auto knVectorType = modelBuilder.getVectorType({K, N}, f32);
  auto typeB = modelBuilder.getMemRefType({-1, -1}, knVectorType);
  auto mnVectorType = modelBuilder.getVectorType({M, N}, f32);
  auto typeC = modelBuilder.getMemRefType({-1, -1}, mnVectorType);

  auto func = modelBuilder.makeFunction(funcName, {}, {typeA, typeB, typeC});

  OpBuilder b(&func.getBody());
  ScopedContext scope(b, func.getLoc());
  ValueHandle A(func.getArgument(0)), B(func.getArgument(1)),
      C(func.getArgument(2));
  auto contractionBuilder = [](ArrayRef<BlockArgument> args) {
    assert(args.size() == 3 && "expected 3 block arguments");
    (linalg_yield(vector_matmul(args[0], args[1], args[2])));
  };

  linalg_matmul(A, B, C, contractionBuilder);
  std_ret();

  modelBuilder.getModuleRef()->dump();
  flush();
}

int main(int argc, char **argv) {
  // Allow LLVM setup through command line and parse the
  // test specific option for a runtime support library.
  llvm::InitLLVM y(argc, argv);
  llvm::cl::ParseCommandLineOptions(argc, argv, "TestSimpleJIT\n");

  // CHECK-LABEL: func @test_vector_add_4_4
  //      CHECK: ( ( 1, 1, 1, 1 ), ( 1, 1, 1, 1 ), ( 1, 1, 1, 1 ),
  // CHECK-SAME: ( 1, 1, 1, 1 ) )
  //      CHECK: ( ( 1, 2, 3, 4 ), ( 5, 6, 7, 8 ), ( 9, 10, 11, 12 ),
  // CHECK-SAME: ( 13, 14, 15, 16 ) )
  //      CHECK: ( ( 2, 3, 4, 5 ), ( 6, 7, 8, 9 ), ( 10, 11, 12, 13 ),
  // CHECK-SAME: ( 14, 15, 16, 17 ) )
  //      CHECK: SUCCESS
  testVectorAdd<4, 4>("test_vector_add_4_4", /*kNumElements=*/13);

  // CHECK-LABEL: func @test_vector_add_9_7
  // CHECK: SUCCESS
  // TODO(ntv): fix padding?
  testVectorAdd<9, 7>("test_vector_add_9_7", /*kNumElements=*/5);

  // CHECK-LABEL: func @test_vector_add_17_19
  // CHECK: SUCCESS
  // TODO(ntv): fix padding?
  testVectorAdd<17, 19>("test_vector_add_17_19", /*kNumElements=*/3);

  // CHECK-LABEL: func @test_vector_matmul(
  //  CHECK-SAME:   %[[A:.*]]: memref<?x?xvector<4x16xf32>>,
  //  CHECK-SAME:   %[[B:.*]]: memref<?x?xvector<16x8xf32>>,
  //  CHECK-SAME:   %[[C:.*]]: memref<?x?xvector<4x8xf32>>)
  // Fill its body.
  //      CHECK:   linalg.generic {{.*}} %[[A]], %[[B]], %[[C]]
  //      CHECK:     vector.contract {{.*}} : vector<4x16xf32>, vector<16x8xf32>
  // CHECK-SAME:       into vector<4x8xf32>
  testMatmulOnVectors<4, 8, 16>("test_vector_matmul");
}