Messages in this thread | | | Subject | Re: [PATCH] dsp56k: use request_firmware | From | David Woodhouse <> | Date | Thu, 10 Jul 2008 09:53:39 +0100 |
| |
On Thu, 2008-07-10 at 09:38 +0200, Fredrik Noring wrote: > Hi David, > > 9 jul 2008 kl. 15.05 skrev David Woodhouse: > > It is in fact identical. Disassembly with the tools found on > > sourceforge > > was fun, since they have endianness bugs, don't recognise the 'JMP' > > which was the first instruction, and need each 3-byte instruction in > > the > > binary to be prefixed with a zero byte to make it 4 bytes. But after > > sorting that out and going through it instruction by instruction, it > > really does seem to match. > > Heh. Amazing. > > [ I think the code can be simplified: It currently copies itself to a > high program memory location, but this is not necessary because DSP > soft reset was never implemented (the reset would have been trapped by > the code, in order to load a new program). The driver instead power- > cycles the DSP every time a new program is loaded. ]
I did wonder about that. But since we can't test, we're not going to touch. Better to keep the assembler code which precisely matches the known-good binary.
-- dwmw2
| |