研讨会 软件中断.ppt

* * * 54 Any structure can added to a Q. The only requirement is that the FIRST element in the Q is a QUE_Elem. * * * 54 You have to use some external means to synchronize threads which communicate with Q’s. It is NOT built in like PIPs. For multiple buffers, you can use two Q, a full Q and an empty Q. This is more complex than using one PIP. * STUDY THE BUFFER MANAGEMENT TECHNIQUES CAREFULLY BEFORE TEACHING!!WHILE NOT A BIOS DETAIL, BEING ABLE TO PRESENT THE CONCEPTS OF THE BUFFERS WILL MAKE AN UNDERSTANDING OF THE LABS MORE COMPLETE AND – HOPEFULLY – FASTER. DIFFERENT LABS USE DIFFERENT BUFFER MANAGEMENT TECHNIQUES. BE CAREFUL TO NOT CONFUSE OR MIX CONCEPTS FROM EACH. LAB 3 WAS EASY – NO BUFFER WAS USED, SINCE THE PROCESSING WAS PER SAMPLE. CHAPTER 4 USES TWO DIFFERENT BUFFER METHODS: LAB04.C IS A ==SINGLE BUFFER== EXAMPLE. NOT A GOOD REAL-TIME CHOICE, BUT A PLACE TO START OUR STUDIES. after each buffer is FIR processed, the end of the data buffer is copied to the start of the data buffer. this allows the next block of data to use the end of the prior block as it’s “initial conditions” or “priming” the buffer size needs to be the size of the collected block of data PLUS the size of the FIR filter (minus 1) new data is placed into the buffer below the initial condition values LAB04A.C IS A ==DOUBLE BUFFER== EXAMPLE. IT IS ALSO A **MONO** SOLUTION. with 2 buffers alternating, the initial conditions are copied from the end of 1 buffer to the start of the other data is again collected below the initial condition area through the end of the buffer LAB04C.C is a **STEREO** SOLUTION, using yet a third approach to buffering the HWI reads a single 32 bit input word from the McBSP : 16MSB = Left data, 16LSB = Right data each HWI run we collect a sample each for left and right buffers each block completion, we call the SWI that: begins processing a FIR on the buffer specified in the SWI.mailbox (as written to by the HWI using SWI_or) updates the