* 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.
*/
* \file all_gather_matmul_v2_def.cpp
* \brief
*/
#include "register/op_def_registry.h"
namespace ops {
class AllGatherMatmulV2 : public OpDef {
public:
explicit AllGatherMatmulV2(const char *name) : OpDef(name) {
this->Input("x1")
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_HIFLOAT8, ge::DT_HIFLOAT8, ge::DT_HIFLOAT8,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2})
.FormatList({ge::FORMAT_ND})
.AutoContiguous();
this->Input("x2")
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_HIFLOAT8, ge::DT_HIFLOAT8, ge::DT_HIFLOAT8,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN})
.FormatList({ge::FORMAT_ND})
.IgnoreContiguous();
this->Input("bias")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.FormatList({ge::FORMAT_ND})
.AutoContiguous();
this->Input("x1_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0})
.FormatList({ge::FORMAT_ND})
.AutoContiguous();
this->Input("x2_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0,
ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0, ge::DT_FLOAT8_E8M0})
.FormatList({ge::FORMAT_ND});
this->Input("quant_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.FormatList({ge::FORMAT_ND})
.AutoContiguous();
this->Output("y")
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16,
ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16,
ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT,
ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16,
ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT, ge::DT_FLOAT16})
.FormatList({ge::FORMAT_ND});
this->Output("gather_out")
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_HIFLOAT8, ge::DT_HIFLOAT8, ge::DT_HIFLOAT8,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN, ge::DT_FLOAT8_E4M3FN,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2,
ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2, ge::DT_FLOAT8_E5M2})
.FormatList({ge::FORMAT_ND});
this->Output("amax_out")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.FormatList({ge::FORMAT_ND});
this->Attr("group").AttrType(REQUIRED).String();
this->Attr("is_trans_a").AttrType(OPTIONAL).Bool(false);
this->Attr("is_trans_b").AttrType(OPTIONAL).Bool(false);
this->Attr("gather_index").AttrType(OPTIONAL).Int(0);
this->Attr("comm_turn").AttrType(OPTIONAL).Int(0);
this->Attr("rank_size").AttrType(OPTIONAL).Int(0);
this->Attr("block_size").AttrType(OPTIONAL).Int(0);
this->Attr("group_size").AttrType(OPTIONAL).Int(0);
this->Attr("is_gather_out").AttrType(OPTIONAL).Bool(true);
this->Attr("is_amax_out").AttrType(OPTIONAL).Bool(false);
this->Attr("y_dtype").AttrType(OPTIONAL).Int(static_cast<int>(ge::DT_UNDEFINED));
this->Attr("comm_mode").AttrType(REQUIRED).String("aicpu");
OpAICoreConfig aicore_config;
aicore_config.DynamicCompileStaticFlag(true)
.DynamicFormatFlag(true)
.DynamicRankSupportFlag(true)
.DynamicShapeSupportFlag(true)
.NeedCheckSupportFlag(false)
.PrecisionReduceFlag(true)
.ExtendCfgInfo("aclnnSupport.value", "support_aclnn")
.ExtendCfgInfo("jitCompile.flag", "static_false")
.ExtendCfgInfo("multiKernelSupportDynamicGraph.value", "multi_kernel")
.ExtendCfgInfo("opFile.value", "all_gather_matmul_v2_apt");
this->AICore().AddConfig("ascend950", aicore_config);
this->MC2().HcclGroup("group");
OpAICoreConfig aicore_config_910b;
aicore_config_910b.Input("x1")
.ParamType(REQUIRED)
.DataType({ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT8, ge::DT_INT8,
ge::DT_INT8, ge::DT_INT8, ge::DT_INT8, ge::DT_INT8,
ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT4, ge::DT_INT4})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Input("x2")
.ParamType(REQUIRED)
.DataType({ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT8, ge::DT_INT8,
ge::DT_INT8, ge::DT_INT8, ge::DT_INT8, ge::DT_INT8,
ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT4, ge::DT_INT4})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_FRACTAL_NZ, ge::FORMAT_FRACTAL_NZ, ge::FORMAT_ND, ge::FORMAT_FRACTAL_NZ,
ge::FORMAT_FRACTAL_NZ, ge::FORMAT_FRACTAL_NZ, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_FRACTAL_NZ, ge::FORMAT_FRACTAL_NZ, ge::FORMAT_ND, ge::FORMAT_FRACTAL_NZ,
ge::FORMAT_FRACTAL_NZ, ge::FORMAT_FRACTAL_NZ, ge::FORMAT_ND, ge::FORMAT_ND})
.IgnoreContiguous();
aicore_config_910b.Input("bias")
.ParamType(OPTIONAL)
.DataType({ge::DT_BF16, ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Input("x1_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Input("x2_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_INT64, ge::DT_FLOAT,
ge::DT_INT64, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Input("quant_scale")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Output("y")
.ParamType(REQUIRED)
.DataType({ge::DT_BF16, ge::DT_FLOAT16, ge::DT_FLOAT16, ge::DT_BF16,
ge::DT_FLOAT16, ge::DT_BF16, ge::DT_FLOAT16, ge::DT_FLOAT16,
ge::DT_BF16, ge::DT_FLOAT16, ge::DT_FLOAT16, ge::DT_BF16})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Output("gather_out")
.ParamType(REQUIRED)
.DataType({ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT8, ge::DT_INT8,
ge::DT_INT8, ge::DT_INT8, ge::DT_INT8, ge::DT_INT8,
ge::DT_BF16, ge::DT_FLOAT16, ge::DT_INT4, ge::DT_INT4})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.Output("amax_out")
.ParamType(OPTIONAL)
.DataType({ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT,
ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT, ge::DT_FLOAT})
.Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND,
ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
aicore_config_910b.DynamicCompileStaticFlag(true)
.DynamicFormatFlag(true)
.DynamicRankSupportFlag(true)
.DynamicShapeSupportFlag(true)
.NeedCheckSupportFlag(false)
.PrecisionReduceFlag(true)
.ExtendCfgInfo("aclnnSupport.value", "support_aclnn")
.ExtendCfgInfo("jitCompile.flag", "static_false")
.ExtendCfgInfo("multiKernelSupportDynamicGraph.value", "multi_kernel");
this->AICore().AddConfig("ascend910b", aicore_config_910b);
this->AICore().AddConfig("ascend910_93", aicore_config_910b);
this->MC2().HcclGroup("group");
}
};
OP_ADD(AllGatherMatmulV2);
}