优化http/rtsp推流器/rtmp推流器内存占用

This commit is contained in:
xiongziliang
2019-04-23 12:16:14 +08:00
parent ebd2ed2d97
commit 74621618ff
3 changed files with 13 additions and 0 deletions

View File

@@ -115,6 +115,9 @@ void HttpClient::onConnect(const SockException &ex) {
return;
}
//先假设http客户端只会接收一点点数据只接受http头节省内存
_sock->setReadBuffer(std::make_shared<BufferRaw>(1 * 1024));
_totalBodySize = 0;
_recvedBodySize = 0;
HttpRequestSplitter::reset();
@@ -155,6 +158,9 @@ int64_t HttpClient::onRecvHeader(const char *data, uint64_t len) {
}
if(_parser["Transfer-Encoding"] == "chunked"){
//我们认为这种情况下后面应该有大量的数据过来,加大接收缓存提高性能
_sock->setReadBuffer(std::make_shared<BufferRaw>(256 * 1024));
//如果Transfer-Encoding字段等于chunked则认为后续的content是不限制长度的
_totalBodySize = -1;
_chunkedSplitter = std::make_shared<HttpChunkedSplitter>([this](const char *data,uint64_t len){
@@ -179,6 +185,8 @@ int64_t HttpClient::onRecvHeader(const char *data, uint64_t len) {
//但是由于我们没必要等content接收完毕才回调onRecvContent(因为这样浪费内存并且要多次拷贝数据)
//所以返回-1代表我们接下来分段接收content
_recvedBodySize = 0;
//根据_totalBodySize设置接收缓存大小
_sock->setReadBuffer(std::make_shared<BufferRaw>(MAX(_totalBodySize + 1,256 * 1024)));
return -1;
}