|
Hello Martin Probably the first and formost thing to mention is that the bootloader is a program running on the DSP. When the VC33 starts in bootloader mode the bootloader code (in the internal ROM) polls the INTx lines to see which ones are active in this order. INT3 (serial) INT0, INT1 and then INT2. The first value that is read tells the bootloader the chunk size, followed by some initialization and header information. The bootloader then reads data into the DSP in what I like to call 'chunks' of 8, 16 or 32b width. These in turn are concatenated into 32 bit values that will become the instructions and data. Boot information at 0x400000 is where the external VC33 DSK ROM is, so it had better work! More Bootloader Notes --------------------- The first read simply tells the bootloader the width and is only 1 byte. If you are feeding the boot information via a FIFO, the DSK interface, or your logic you need to know this! The bootloader is a program running from the internal ROM. It makes calls to functions and therefor needs to have a small stack space of 1 word. That word is located at 0x809801 --> IN OTHER WORDS DO NOT TRY TO BOOTLOAD OVER THIS LOCATION! Extra L/A Hardware ------------------ As long as the loading is light (and it should be) no buffering should not be a problem. Anytime you hang something on a signal you have the potential of creating signal integrity problems. Sometimes this is a care, sometimes not. Consider for example the address or data lines. Poor signal quality will affect the setup and hold times, but these are qualified by the PAGEx, STRB and R/W control signals. As long as these signals are clean and stable on the clock edges you can get away with really crappy signals. In other words, simply slowing down the clock will fix these kinds of problems. Peripherals like the serial ports and timers also fall into this catagory. Now consider what happens if you corrupt a control signal! Here noise and other junk can cause bogus reads, writes or bus conflicts and really mess things up. RPTS ---- When the RPTS instruction is fetched (and has not been decoded yet) it will be followed by one more instruction fetch that will ultimately be repeated. Since the next instruction is now within a register the CPU can re-issue that instruction N times without having to go back and re-read the instruction from memory. This frees up the bus, but the CPU also happens to be 'locked in' and executing this one instruction. The CPU therefor is not servicing interrupts. If you need interrupts, use the RPTB instruction as it is valid to have a block length of 1. The penalty of using RPTB is that the opcode is constantly fetched potentially leading to bus conflicts and slowdowns. Here are some more interesting ways to repeat chunks of code ;---------------- ; Adds 2*2*1 to R0 ; the call pushes the next address creating a loop ;---------------- Func call Next ; syntactically== $+1 Next call $+1 ;... this address+1 addi 1,R0 ; rets ; ;--------------- ; the RPTB registers, RS, RE and RC can be preloaded ; well outside the innermost loop. An example would ; be an FFT where the innermost butterfly needs to ; start quickly ;--------------- ldi @Addr_Loop1,RS ; Preload start ldi @Addr_Loop2,RE ; Preload end ldi 7,RC ; repeat 8x :: ; or RM,ST ; fast start repeat mode Loop1 addi 1,R0 ; starts innermost loop addi 1,R1 ; with only 1 cycle lost Loop2 addi 1,R2 ; :: rets ; ;---------------------------- Addr_Loop1 .word Loop1 Addr_Loop2 .word Loop2 Hope this helps Best regards, Keith Larson =================================== At 01:55 PM 10/28/03 -0000, you wrote: Hello I'm Martin Argoit from Argentina, I'm work with a board that include the DSP TMS320vc33-150, all bypass capacitors, the 2 supply plane (1.8 & 3.3V) and the ground plane. The doubts I have its because I'm new in this and I don't want do something wrong . I'm using an logic analyzer with a pattern generator to load the program. For example: -It's ok that I place the program I want to try in the memory 400000, when I'm in the boot load mode? -Must I use some hardware between the analyzer & the vc33, which one? -How das it work the repeats (RPTS) function? I hope that someone could help me. Martin Argoit +-----------+ |Keith Larson | |Member Group Technical Staff | |Texas Instruments Incorporated | | | | 281-274-3288 | | | | www.micro.ti.com/~klarson | |-----------+ | TMS320C3x/C4x/VC33 Applications | | | | TMS320VC33 | | The lowest cost and lowest power 500 w/Mflop | | floating point DSP on the planet! | +-----------+ |
Re:
Started by ●October 28, 2003
