2019-06-24 11:32:41 +08:00
|
|
|
//
|
|
|
|
// GLBinary.cpp
|
|
|
|
// MNN
|
|
|
|
//
|
|
|
|
// Created by MNN on 2019/01/31.
|
|
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
|
|
//
|
|
|
|
|
2020-01-15 13:33:47 +08:00
|
|
|
#include "backend/opengl/GLBinary.hpp"
|
2019-06-24 11:32:41 +08:00
|
|
|
#include <sstream>
|
2020-01-15 13:33:47 +08:00
|
|
|
#include "AllShader.hpp"
|
2019-12-27 22:16:57 +08:00
|
|
|
#include "backend/opengl/GLBackend.hpp"
|
|
|
|
#include "core/Macro.h"
|
|
|
|
#include "core/TensorUtils.hpp"
|
2019-06-24 11:32:41 +08:00
|
|
|
|
|
|
|
namespace MNN {
|
|
|
|
namespace OpenGL {
|
|
|
|
GLBinary::GLBinary(const std::vector<Tensor *> &inputs, const Op *op, Backend *bn) : Execution(bn) {
|
|
|
|
mType = op->main_as_BinaryOp()->opType();
|
|
|
|
}
|
|
|
|
|
|
|
|
GLBinary::~GLBinary() {
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-06-24 11:32:41 +08:00
|
|
|
}
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-06-24 11:32:41 +08:00
|
|
|
ErrorCode GLBinary::onResize(const std::vector<Tensor *> &inputs, const std::vector<Tensor *> &outputs) {
|
|
|
|
std::vector<std::string> prefix;
|
|
|
|
setLocalSize(prefix, mLocalSize, 8, 8, 1);
|
|
|
|
if (BinaryOpOperation_ADD == mType) {
|
|
|
|
prefix.push_back("#define ADD");
|
|
|
|
mProgram = ((GLBackend *)backend())->getProgram("binary_add", glsl_binary_glsl, prefix);
|
|
|
|
}else if(BinaryOpOperation_MUL == mType) {
|
|
|
|
prefix.push_back("#define MUL");
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
mProgram = ((GLBackend *)backend())->getProgram("binary_mul", glsl_binary_glsl, prefix);
|
|
|
|
}else if(BinaryOpOperation_SUB == mType) {
|
|
|
|
prefix.push_back("#define SUB");
|
|
|
|
mProgram = ((GLBackend *)backend())->getProgram("binary_sub", glsl_binary_glsl, prefix);
|
|
|
|
}else if(BinaryOpOperation_REALDIV == mType) {
|
|
|
|
prefix.push_back("#define REALDIV");
|
|
|
|
mProgram = ((GLBackend *)backend())->getProgram("binary_realdiv", glsl_binary_glsl, prefix);
|
2019-06-24 11:32:41 +08:00
|
|
|
}else{
|
|
|
|
MNN_PRINT("Not Supported Binary Operation: %d\n", mType);
|
|
|
|
}
|
|
|
|
return NO_ERROR;
|
|
|
|
}
|
|
|
|
|
|
|
|
ErrorCode GLBinary::onExecute(const std::vector<Tensor *> &inputs, const std::vector<Tensor *> &outputs) {
|
|
|
|
MNN_ASSERT(2 == inputs.size());
|
|
|
|
MNN_ASSERT(1 == outputs.size());
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-06-24 11:32:41 +08:00
|
|
|
auto input0 = inputs[0];
|
|
|
|
auto input1 = inputs[1];
|
|
|
|
auto output = outputs[0];
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-07-02 18:01:08 +08:00
|
|
|
std::vector<int> inputShape = tensorShapeFormat(input0);
|
|
|
|
std::vector<int> outputShape = tensorShapeFormat(output);
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-07-02 18:01:08 +08:00
|
|
|
int ib = inputShape.at(0);
|
|
|
|
int ih = inputShape.at(1);
|
|
|
|
int iw = inputShape.at(2);
|
|
|
|
int ic = inputShape.at(3);
|
|
|
|
int ic_4 = UP_DIV(ic, 4);
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-07-02 18:01:08 +08:00
|
|
|
int ob = outputShape.at(0);
|
|
|
|
int oh = outputShape.at(1);
|
|
|
|
int ow = outputShape.at(2);
|
|
|
|
int oc = outputShape.at(3);
|
|
|
|
int oc_4 = UP_DIV(oc, 4);
|
2019-12-27 22:16:57 +08:00
|
|
|
|
2019-07-02 18:01:08 +08:00
|
|
|
MNN_ASSERT(input0->getType().code == halide_type_float);
|
2019-06-24 11:32:41 +08:00
|
|
|
MNN_ASSERT(input0->dimensions() == input1->dimensions());
|
|
|
|
|
|
|
|
const auto intputFormat = TensorUtils::getDescribe(input0)->dimensionFormat;
|
2019-12-27 22:16:57 +08:00
|
|
|
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
mProgram->useProgram();
|
|
|
|
glBindImageTexture(0, output->deviceId(), 0, GL_TRUE, 0, GL_WRITE_ONLY, ((GLBackend *)backend())->getTextrueFormat());
|
|
|
|
{
|
|
|
|
int texId = 0;
|
|
|
|
glActiveTexture(GL_TEXTURE0 + texId);
|
|
|
|
glUniform1i(1, texId);
|
|
|
|
glBindTexture(GL_TEXTURE_3D, input0->deviceId());
|
|
|
|
OPENGL_CHECK_ERROR;
|
|
|
|
}
|
|
|
|
{
|
|
|
|
int texId = 1;
|
|
|
|
glActiveTexture(GL_TEXTURE0 + texId);
|
|
|
|
glUniform1i(2, texId);
|
|
|
|
glBindTexture(GL_TEXTURE_3D, input1->deviceId());
|
2019-06-24 11:32:41 +08:00
|
|
|
OPENGL_CHECK_ERROR;
|
|
|
|
}
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
glUniform4i(3, iw, ih, ic_4, 1);
|
|
|
|
OPENGL_CHECK_ERROR;
|
|
|
|
((GLBackend *)backend())->compute(UP_DIV(iw, mLocalSize[0]), UP_DIV(ih, mLocalSize[1]), UP_DIV(ic_4, mLocalSize[2]));
|
2019-07-02 18:01:08 +08:00
|
|
|
|
2019-06-24 11:32:41 +08:00
|
|
|
return NO_ERROR;
|
|
|
|
}
|
2019-12-27 22:16:57 +08:00
|
|
|
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
class BinaryCreator : public GLBackend::Creator {
|
|
|
|
public:
|
|
|
|
virtual ~BinaryCreator() = default;
|
|
|
|
virtual Execution *onCreate(const std::vector<Tensor *> &inputs, const std::vector<Tensor *> &outputs,
|
|
|
|
const MNN::Op *op, Backend *backend) const override {
|
2019-12-27 22:16:57 +08:00
|
|
|
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
MNN_ASSERT(inputs.size() > 1);
|
|
|
|
auto input0 = inputs[0];
|
|
|
|
// Don't support broatcast
|
|
|
|
for (int i = 1; i < inputs.size(); ++i) {
|
|
|
|
auto input = inputs[i];
|
|
|
|
if (input0->dimensions() != input->dimensions()) {
|
|
|
|
MNN_PRINT("dimensions : [%d, %d] \n", input0->dimensions(), input->dimensions());
|
|
|
|
MNN_PRINT("opengl binary don't support broatcast !!! \n");
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
auto dim = input0->dimensions();
|
|
|
|
for (int l = 0; l < dim; ++l) {
|
|
|
|
if (input0->length(l) != input->length(l)) {
|
|
|
|
MNN_PRINT("length : [%d, %d] \n", input0->length(l), input->length(l));
|
|
|
|
MNN_PRINT("opengl binary don't support broatcast !!! \n");
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2019-12-27 22:16:57 +08:00
|
|
|
|
- dynamic computation graph (beta)
- add supports (/express)
- add tests
- add benchmarks with it (/benchmark/exprModels)
- Python
- MNN engine and tools were submitted to pip
- available on Windows/macOS/Linux
- Engine/Converter
- add supports for each op benchmarking
- refactor optimizer by separating steps
- CPU
- add supports for Conv3D, Pool3D, ELU, ReverseSequence
- fix ArgMax, Permute, Scale, BinaryOp, Slice, SliceTf
- OpenCL
- add half transform in CPU
- add broadcast supports for binary
- optimize Conv2D, Reshape, Eltwise, Gemm, etc.
- OpenGL
- add sub, real div supports for binary
- add supports for unary
- optimize Conv2D, Reshape
- Vulkan
- add max supports for eltwise
- Metal
- fix metallib missing problem
- Train/Quantization
- use express to refactor training codes
2019-09-26 21:02:07 +08:00
|
|
|
return new GLBinary(inputs, op, backend);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
GLCreatorRegister<BinaryCreator> __binary_op(OpType_BinaryOp);
|
2019-06-24 11:32:41 +08:00
|
|
|
} // namespace OpenGL
|
|
|
|
} // namespace MNN
|