On 9/14/20 11:07 PM, Richard Henderson wrote: > On 9/13/20 11:23 PM, Philippe Mathieu-Daudé wrote: >> The old disassembler is easier to follow: >> >> ---------------- >> IN: >> -0x4000d214: lda [ %g3 ] #ASI_M_FLUSH_PROBE, %g6 >> -0x4000d218: sta %g6, [ %g4 ] #ASI_M_FLUSH_PROBE >> +0x4000d214: cc80c060 .byte 0xcc, 0x80, 0xc0, 0x60 >> +0x4000d218: cca10060 .byte 0xcc, 0xa1, 0x00, 0x60 >> >> ^ lda/sta opcodes not supported (there might be more). >> >> ---------------- >> IN: >> -0x4000d22c: sethi %hi(0xf01f0000), %g1 >> -0x4000d230: mov %g1, %g1 ! 0xf01f0000 >> -0x4000d234: jmp %g1 >> -0x4000d238: nop >> +0x4000d22c: 033c07c0 sethi 0x3c07c0, %g1 >> +0x4000d230: 82106000 or %g1, 0, %g1 >> +0x4000d234: 81c04000 jmp %g1 >> +0x4000d238: 01000000 nop >> >> ^ hi()/lo() macros not expanded (easier to read!) > > Hmm, yes. I'm going to drop this for now. > Maybe revisit if this gets fixed upstream. > >> Can we restrict dumping the encoded hex content for debug >> profile only? > > Why? Including the hex content was in fact requested when I first added the > capstone code.
Ah, OK. Now I wonder if I hadn't asked you the same when you introduced capstone :) Maybe we could include the hex content in the current disas.c as a first step, that would ease diffing for missing opcodes.
