[#6447] assembler should reject "[--SP] = SP;"
Submitted By: Mike Frysinger
2011-01-14 00:31:04 Close Date
N/A Silicon Revision:
Fixed Fixed In Release:
Host Operating System:
Closed Found In Release:
Is this bug repeatable?:
Summary: assembler should reject "[--SP] = SP;"
you cant push the stack pointer onto the stack, but our assembler happily accepts it and outputs the correct encoding for it
we should reject this:
$ cat test.s
[--SP] = SP;
$ bfin-elf-as test.s
--- David Gibson 2011-01-14 03:03:01
I don't understand why should we reject the code?
I believe that it's a legal instruction for Blackfin. As you say, the assembler
accepts and outputs the correct encoding. Is there another reason for rejecting
--- Mike Frysinger 2011-01-14 10:47:14
because the hardware rejects it
--- Mike Frysinger 2011-01-14 14:00:50
sorry, i was remembering wrong. when Robin & i were doing heavy sim
development, we came up with a bunch of opcodes that the hardware should reject
but doesnt actually. this is one of those. design agreed.
it is also my understanding that the VDSP assembler rejects this ? our
disassembler certainly flags it as illegal ...
--- David Gibson 2011-01-14 16:56:14
You're right. The VDSP Assembler accepts it also, which is why I questioned it.
I'll open a TAR for the VDSP assembler - and check that the compiler doesn't
emit this instruction. We also need to check if the GNU compiler can emit this
--- Stuart Henderson 2011-01-28 13:29:22
Fixed on trunk.
File Name File Type File Size Posted By
No Files Were Found