> Sorry, I don't know mips assembler at all, or gas. > I run this into gdb, I 'stepi', and I don't see the proper values loaded > into the floating point registers. > > Is it me doing something wrong? It looks like those ldc1 go compiled into > totally different instructions, not referring to 'a' or 'b'.
mips is such a strange beast that its assembly language is composed of many macro instructions that the assembler understand and replace with other, simpler ones. Also, the assembler frequently reorder instructions. This is very annoying when you do some serious programming in assembler (because if you wanted the help of a program to find out the instructions to output then you wouldn't use assembly language in the first place), but I suppose this was designed to help higher level languages compilers to deal with mips singularities. So, no worries if the code does not look like the above one when singlestepping in gdb. The final values loaded into the registers should be the same. Now for our problem (madd.d raising an exception for some values and some registers) I cannot investigate further until monday evening, having no mips at hand for now. But it would be interresting to know what exception is raised precisely, because IIRC that could also be an unimplemented instruction exception that the fpu is allowed to raise for some values in some circumstances if it feels like it, according to the manual which does not states which circumstances exactly. Certainly there is a debug printk between the trap handler and when the process is signaled that could tell us this? -- You received this message because you are subscribed to the Google Groups "loongson-dev" group. To post to this group, send email to loongson-dev@googlegroups.com. To unsubscribe from this group, send email to loongson-dev+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/loongson-dev?hl=en.
