On 09/08/2015, Sam Ravnborg <sam@xxxxxxxxxxxx> wrote: > When browsing sparc64 assembler I can see most > handwritten assembler routines are aligned with ".align 32". > > But there is also a bunch of routines that uses "ENTRY(foo) - ENDPROC(foo). > They are all aligned with ".align 4,0x90" (see include/linux/linkage.h). > > I have tried to find something that mandates the 32 byte alignment - so far > no luck. > Are there any good reasosn for the use of the 32 bytes alignment? > > I also tried to decode 0x90 in the SPARC world - it is NOP in x86. > I did not manage to decode it - but then I did not try very hard either. > > I would expect that .align 4,0x40 would be a better choice for > sparc because 0x40 is sethi 0, %g0 aka nop. > See https://en.wikipedia.org/wiki/NOP > And 4 seems to be the necessary aligmnet for both sparc32 and sparc64. > > But I am left with an impression that I am missing something since > so many routines uses 32 bytes. > > Sam Hi, I tried to decode 0x90909090 in the V8/V9 manual, and it came back as ORcc %g2, %l0, %o0 (0x90908010). Doesn't make much sense? I guess aligning with 0x00 would be better as it would generate a illegal instruction trap if executed. If the alignment is executed something is already wrong and the first instructions from the next function might be executed depending alignment of functions in a given compile. Regards, Kjetil Oftedal -- To unsubscribe from this list: send the line "unsubscribe sparclinux" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
