performance - What's the best way of implementing a buffer of fixed size when using fread in C++? -

performance - What's the best way of implementing a buffer of fixed size when using fread in C++? -

suppose have file of integers , want read them 1 one.

you have 2 options buffering.

declare array buffer of size n , utilize setvbuf tell fread buffer use. when calling function fread read integer write fread(&myint, sizeof(myint), 1, inputfile);

declare same array buffer time don't utilize function setvbuf. instead work on buffering yourself. phone call fread(buffer, buffersize*sizeof(int), 1, inputfile)

from understanding setvbuf exists create life easier, come @ cost? method utilize in terms of performance?

i utilize neither of examples. don't think part of i/o performance bottleneck.

the vbuf area input routine place info before putting destination. used cache or preformatting buffer.

most of time, i/o bottlenecks related quantity of info fetched , number of fetches. example, reading 1 byte @ time less efficient reading block of bytes.

another i/o related bottleneck duration between input requests. i/o devices prefer maintain streaming data, non-stop. input devices, hard drives, have overhead time between when request received , when info starts transmitting. hard drives, disk speed time.

your best performance not waste development time messing c or c++ libraries. need utilize hardware assist. platforms have device called direct memory access controller (dma). device can take info input source , deliver memory without using cpu. cpu can executing instructions while dma transferring data. in order utilize hardware assistance, need write code @ os driver level, or access os drivers directly.

the c , c++ i/o libraries designed platform independent concept called streams. there may execution overhead associated (such buffering). if don't care different platforms, access os drivers directly.

don't waste time messing c , c++ libraries. not much performance gain there. more performance lies in accessing os drivers straight (or using own). how , when access i/o show bigger performance gains tweaking c , c++ libraries.

lastly, using processors info cache gain performance too.

c++ performance fread