* Copyright (c) 2025 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
#include "gtest/gtest.h"
#include "node_utils_ex.h"
#include "graph_utils.h"
#include "ascendc_ir.h"
#include "ascir_ops.h"
#include "ascir_ops_utils.h"
#include "codegen_kernel.h"
#include "graph/ascendc_ir/utils/asc_tensor_utils.h"
#include "common_utils.h"
#include "utils/api_call_factory.h"
#include "elewise/leaky_relu_api_call.h"
#include "elewise/cast_api_call.h"
using namespace ge;
using namespace af::ops;
using namespace af::ascir_op;
using namespace codegen;
TEST(CodegenKernel, CastApiCall_Zero_Stride) {
af::AscGraph graph("test_graph");
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto s2 = graph.CreateSizeVar("s2");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
auto z2 = graph.CreateAxis("z2", s2);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id, z2.id};
*load_op.y.axis = {z0.id, z1.id, z2.id};
*load_op.y.repeats = {s0, One, s2};
*load_op.y.strides = {s2, Zero, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id, z2.id};
*cast_op.y.repeats = {s0, One, s2};
*cast_op.y.strides = {s2, Zero, One};
auto load = graph.FindNode("load");
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
auto size = ge::GetSizeByDataType(ge::DT_FLOAT16);
load->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
load->outputs[0].attr.vectorized_strides = {Zero, One};
load->outputs[0].attr.dtype = ge::DT_FLOAT;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
cast->outputs[0].attr.vectorized_strides = {Zero, One};
cast->outputs[0].attr.dtype = ge::DT_INT16;
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddAxis(z2);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
tiler.AddSizeVar(af::SizeVar(s2));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
EXPECT_EQ(result, std::string{"CastExtend(local_1[0], local_0[0], tmp_buf_0, local_0_actual_size);\n"});
}
TEST(CodegenKernel, CastApiCall) {
af::AscGraph graph("test_graph");
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id};
*load_op.y.axis = {z0.id, z1.id};
*load_op.y.repeats = {s0, s1};
*load_op.y.strides = {s1, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id};
*cast_op.y.repeats = {s0, s1};
*cast_op.y.strides = {s1, One};
auto load = graph.FindNode("load");
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
load->outputs[0].attr.vectorized_axis = {z1.id};
load->outputs[0].attr.vectorized_strides = {One};
load->outputs[0].attr.dtype = ge::DT_FLOAT;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z1.id};
cast->outputs[0].attr.vectorized_strides = {One};
cast->outputs[0].attr.dtype = ge::DT_INT16;
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
EXPECT_EQ(result, std::string{
"CastExtend(local_1[0], local_0[0], tmp_buf_0, local_0_actual_size);\n"
});
}
TEST(CodegenKernel, CastApiCallTwoDimension) {
af::AscGraph graph("test_graph");
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id};
*load_op.y.axis = {z0.id, z1.id};
*load_op.y.repeats = {s0, s1};
*load_op.y.strides = {s1, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id};
*cast_op.y.repeats = {s0, s1};
*cast_op.y.strides = {s1 + s1, One};
auto load = graph.FindNode("load");
auto size = ge::GetSizeByDataType(ge::DT_FLOAT16);
auto stride = af::sym::Align(z1.size, 32 / size);
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
load->outputs[0].attr.vectorized_axis = {z0.id, z1.id};
load->outputs[0].attr.vectorized_strides = {stride, One};
load->outputs[0].attr.dtype = ge::DT_FLOAT16;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
auto size1 = ge::GetSizeByDataType(ge::DT_FLOAT);
auto stride1 = af::sym::Align(z1.size, 32 / size1);
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z0.id, z1.id};
cast->outputs[0].attr.vectorized_strides = {stride1, One};
cast->outputs[0].attr.dtype = ge::DT_FLOAT;
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
EXPECT_EQ(result, std::string{
"CastExtend(local_1[0], local_0[0], tmp_buf_0, t->s0, t->s1, ((16 * Ceiling((Rational(1 , 16) * t->s1))))/(1), ((8 * Ceiling((Rational(1 , 8) * t->s1))))/(1), 4);\n"
});
}
TEST(CodegenKernel, CastApiCallThreeDimension) {
af::AscGraph graph("test_graph");
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto s2 = graph.CreateSizeVar("s2");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
auto z2 = graph.CreateAxis("z2", s2);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id, z2.id};
*load_op.y.axis = {z0.id, z1.id, z2.id};
*load_op.y.repeats = {s0, s1, s2};
*load_op.y.strides = {s1 * s2, s2, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id, z2.id};
*cast_op.y.repeats = {s0, s1, s2};
*cast_op.y.strides = {s1 * s2 + s1 * s2 + s1 * s2 , s2 + s2, One};
auto load = graph.FindNode("load");
auto size = ge::GetSizeByDataType(ge::DT_FLOAT16);
auto stride = af::sym::Align(z2.size, 32 / size);
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
load->outputs[0].attr.vectorized_axis = {z0.id, z1.id, z2.id};
load->outputs[0].attr.vectorized_strides = {stride * z1.size, stride, One};
load->outputs[0].attr.dtype = ge::DT_FLOAT16;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
auto size1 = ge::GetSizeByDataType(ge::DT_FLOAT);
auto stride1 = af::sym::Align(z2.size, 32 / size1);
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z0.id, z1.id, z2.id};
cast->outputs[0].attr.vectorized_strides = {stride1 * z1.size + stride1 * z1.size, stride1, One};
cast->outputs[0].attr.dtype = ge::DT_FLOAT;
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddAxis(z2);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
tiler.AddSizeVar(af::SizeVar(s2));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
EXPECT_EQ(result, std::string{
"for(int outer_for_0 = 0; outer_for_0 < t->s0; outer_for_0++) {\nCastExtend(local_1[outer_for_0 * ((16 * Ceiling((Rational(1 , 8) * t->s2)) * t->s1))/(1)], local_0[outer_for_0 * ((16 * Ceiling((Rational(1 , 16) * t->s2)) * t->s1))/(1)], tmp_buf_0, t->s1, t->s2, ((16 * Ceiling((Rational(1 , 16) * t->s2))))/(1), ((8 * Ceiling((Rational(1 , 8) * t->s2))))/(1), 4);\n\n}\n"
});
}
TEST(CodegenKernel, CastApiCall_Offset) {
af::AscGraph graph("test_graph");
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto s2 = graph.CreateSizeVar("s2");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
auto z2 = graph.CreateAxis("z2", s2);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id, z2.id};
*load_op.y.axis = {z0.id, z1.id, z2.id};
*load_op.y.repeats = {s0, s1, s2};
*load_op.y.strides = {s1 * s2, s2, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id, z2.id};
*cast_op.y.repeats = {s0, s1, s2};
*cast_op.y.strides = {s1 * s2, s2, One};
auto load = graph.FindNode("load");
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
auto size = ge::GetSizeByDataType(ge::DT_FLOAT16);
auto stride = af::sym::Align(z2.size, 32 / size);
load->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
load->outputs[0].attr.vectorized_strides = {stride, One};
load->outputs[0].attr.dtype = ge::DT_FLOAT;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
cast->outputs[0].attr.vectorized_strides = {stride, One};
cast->outputs[0].attr.dtype = ge::DT_INT16;
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddAxis(z2);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
tiler.AddSizeVar(af::SizeVar(s2));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
EXPECT_EQ(result, std::string{
"CastExtend(local_1[0], local_0[0], tmp_buf_0, t->s1, t->s2, ((16 * Ceiling((Rational(1 , 16) * t->s2))))/(1), ((16 * Ceiling((Rational(1 , 16) * t->s2))))/(1), 4);\n"
});
}
TEST(CodegenKernel, CastApiCall_WithMaskMode) {
std::vector<std::tuple<ge::DataType, ge::DataType, std::string>> x_y_max_size_list = {
std::make_tuple(ge::DT_UINT8, ge::DT_FLOAT16, "2"),
std::make_tuple(ge::DT_INT64, ge::DT_FLOAT, "8"),
std::make_tuple(ge::DT_INT64, ge::DT_INT32, "8"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_FLOAT, "4"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_INT32, "4"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_INT16, "2"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_INT8, "2"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_UINT8, "2"),
std::make_tuple(ge::DT_FLOAT, ge::DT_FLOAT16, "4"),
std::make_tuple(ge::DT_FLOAT, ge::DT_INT64, "8"),
std::make_tuple(ge::DT_FLOAT, ge::DT_INT32, "4"),
std::make_tuple(ge::DT_FLOAT, ge::DT_INT16, "4"),
std::make_tuple(ge::DT_FLOAT, ge::DT_BF16, "4"),
std::make_tuple(ge::DT_INT16, ge::DT_FLOAT16, "2"),
std::make_tuple(ge::DT_INT16, ge::DT_FLOAT, "4"),
std::make_tuple(ge::DT_INT32, ge::DT_FLOAT, "4"),
std::make_tuple(ge::DT_INT32, ge::DT_INT64, "8"),
std::make_tuple(ge::DT_INT32, ge::DT_INT16, "4"),
std::make_tuple(ge::DT_INT32, ge::DT_FLOAT16, "4"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_FLOAT, "4"),
std::make_tuple(ge::DT_FLOAT16, ge::DT_INT32, "4"),
};
for (const auto &t : x_y_max_size_list) {
af::AscGraph graph("test_graph");
std::string max_size = std::get<2>(t);
auto s0 = graph.CreateSizeVar("s0");
auto s1 = graph.CreateSizeVar("s1");
auto s2 = graph.CreateSizeVar("s2");
auto z0 = graph.CreateAxis("z0", s0);
auto z1 = graph.CreateAxis("z1", s1);
auto z2 = graph.CreateAxis("z2", s2);
Data x_op("x", graph);
Load load_op("load");
af::ascir_op::Cast cast_op("cast");
graph.AddNode(load_op);
graph.AddNode(cast_op);
load_op.x = x_op.y;
load_op.attr.sched.axis = {z0.id, z1.id, z2.id};
*load_op.y.axis = {z0.id, z1.id, z2.id};
*load_op.y.repeats = {s0, s1, s2};
*load_op.y.strides = {s1 * s2, s2, One};
cast_op.x = load_op.y;
*cast_op.y.axis = {z0.id, z1.id, z2.id};
*cast_op.y.repeats = {s0, s1, s2};
*cast_op.y.strides = {s1 * s2, s2, One};
auto load = graph.FindNode("load");
load->attr.api.compute_type = af::ComputeType::kComputeLoad;
load->attr.api.type = af::ApiType::kAPITypeCompute;
load->attr.api.unit = af::ComputeUnit::kUnitMTE2;
load->attr.sched.loop_axis = z0.id;
auto size = ge::GetSizeByDataType(ge::DT_FLOAT16);
auto stride = af::sym::Align(z2.size, 32 / size);
load->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
load->outputs[0].attr.vectorized_strides = {stride, One};
load->outputs[0].attr.dtype = std::get<0>(t);
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.tensor_id = 0;
load->outputs[0].attr.mem.position = af::Position::kPositionVecIn;
load->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
load->outputs[0].attr.que.id = 1;
load->outputs[0].attr.opt.merge_scope = af::kIdNone;
auto cast = graph.FindNode("cast");
cast->attr.api.compute_type = af::ComputeType::kComputeElewise;
cast->attr.api.type = af::ApiType::kAPITypeCompute;
cast->attr.api.unit = af::ComputeUnit::kUnitVector;
cast->attr.sched.loop_axis = z0.id;
cast->attr.tmp_buffers = {{{af::Symbol(8192), -1}, af::MemAttr(), 0}};
cast->outputs[0].attr.vectorized_axis = {z1.id, z2.id};
cast->outputs[0].attr.vectorized_strides = {stride, One};
cast->outputs[0].attr.dtype = std::get<1>(t);
cast->outputs[0].attr.mem.position = af::Position::kPositionVecOut;
cast->outputs[0].attr.mem.tensor_id = 1;
cast->outputs[0].attr.mem.alloc_type = af::AllocType::kAllocTypeQueue;
cast->outputs[0].attr.que.id = 2;
cast->outputs[0].attr.opt.merge_scope = af::kIdNone;
codegen::Tiler tiler;
codegen::TPipe tpipe("tpipe", tiler);
tpipe.AddTensor(load->outputs[0]);
tpipe.AddTensor(cast->outputs[0]);
tiler.AddAxis(z0);
tiler.AddAxis(z1);
tiler.AddAxis(z2);
tiler.AddSizeVar(af::SizeVar(s0));
tiler.AddSizeVar(af::SizeVar(s1));
tiler.AddSizeVar(af::SizeVar(s2));
std::vector<af::AxisId> current_axis;
current_axis.push_back(z0.id);
codegen::ApiTensor x1;
x1.id = load->outputs[0].attr.mem.tensor_id;
codegen::CastApiCall call("Cast");
EXPECT_EQ(call.Init(cast), 0);
call.inputs.push_back(&x1);
std::string result;
call.Generate(tpipe, current_axis, result);
std::string expect = "CastExtend(local_1[0], local_0[0], tmp_buf_0, t->s1, t->s2, ((16 * Ceiling((Rational(1 , 16) * t->s2))))/(1), ((16 * Ceiling((Rational(1 , 16) * t->s2))))/(1), " + max_size + ");\n";
EXPECT_EQ(result, expect);
}
}
