在永续合约交易所系统中，交易撮合是一项核心功能。它通过匹配买方和卖方的需求，实现交易的达成。一般来说，交易撮合算法会考虑价格、交易量、交易时间等多个因素，以寻找最合适的交易对手。这一过程需要确保交易的公平性和有效性，防止市场出现不正常的波动。
交割合约是一种有固定到期日的合约，交割合约开发技术与永续合约开发技术类似，但需要考虑到期日的处理和交割机制。
合约量化跟单系统开发是指开发一个用于自动化执行交易策略的系统。该系统可以根据预先设定的交易策略，自动执行交易操作，实现量化交易。需要考虑策略的编写、数据的获取和处理、交易的执行等问题。
Secondly,the basic architecture and functional modules of the perpetual contract exchange system should be briefly introduced.Usually,the system consists of three parts:front-end business,back-end operations,and system support.The front-end business mainly includes functions such as user interface,transaction execution,and transaction query;The backend operation is responsible for core businesses such as contract management,risk control,clearing and delivery;And system support covers aspects such as system maintenance,security guarantees,and technical support.
One of the core technologies in the perpetual contract exchange system is contract encoding.In order to meet the needs of different investors,the perpetual contract exchange system provides diversified contract coding schemes.These coding schemes differ in terms of trading variety,leverage ratio,trading time,etc.Investors can choose appropriate contracts to trade based on their own needs.
Another key technology is smart contracts.Smart contracts are automated contracts based on blockchain technology,which can automatically execute transactions when specific conditions are met.In the perpetual contract exchange system,smart contracts are widely used in transaction matching,clearing and delivery processes.Through smart contracts,transaction efficiency has been improved and the risk of human intervention has been effectively reduced.
int SymmetricQuantizeWeight(const floatweight,const int size,int8_tquantizedWeight,float*scale,
const int channels,float weightClampValue){
/**对参数进行量化
*weight为乘上scale后的权重，
*quantizedWeight用于存放量化后的参数
*/
DCHECK((size%channels)==0)<<"weight size error!";
const int channelStride=size/channels;
const int quantizedMaxValue=weightClampValue;//127
for(int c=0;c<channels;++c){//对每个channel分别量化
const auto weightChannelStart=weight+c*channelStride;
auto quantizedWeightChannelStart=quantizedWeight+c*channelStride;
//获取该channel内最大最小值
auto minmaxValue=std::minmax_element(weightChannelStart,weightChannelStart+channelStride);
const float dataAbsMax=std::fmax(std::fabs(minmaxValue.first),std::fabs(minmaxValue.second));
float scaleDataToInt8=1.0f;
if(dataAbsMax==0){
scale[c]=0.0f;
}else{
//用于逆量化时对用的scale
scale[c]=dataAbsMax/quantizedMaxValue;
//映射到int8空间上的scale
scaleDataToInt8=quantizedMaxValue/dataAbsMax;
}
for(int i=0;i<channelStride;++i){
//将输入权重乘上scale映射到int8上之后，对不在[-127，127]区间的都截断设置为-127或者127.
const int32_t quantizedInt8Value=static_cast<int32_t>(roundf(weightChannelStart*scaleDataToInt8));
quantizedWeightChannelStart=
std::min(quantizedMaxValue,std::max(-quantizedMaxValue,quantizedInt8Value));
}
}
return 0;
}
