gtirb-pprinter emits invalid instruction `fmov v0,x3`

[avncharlie]

When running gtirb pprinter on an arm64 ls binary (Attached: ls_arm64.zip) using the following command:

$ docker run --platform linux/amd64 --rm -v "$(pwd)":/workspace grammatech/ddisasm ddisasm /workspace/ls_arm64 --asm /workspace/out_ls_identity/ls_arm64.S

The generated assembly contains fmov v0,x3 instructions:

$ cat ./out_ls_identity/ls_arm64.S | grep "fmov v0"
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3
            fmov v0,x3

Trying to assemble this code results in this assembler error:

$ gcc -c -o ./ls_arm64.o ./ls_arm64.S -nodefaultlibs -nostartfiles
./ls_arm64.S: Assembler messages:
./ls_arm64.S:29839: Error: the top half of a 128-bit FP/SIMD register is expected at operand 1 -- `fmov v0,x3'
...

I believe the instruction it should emit is: fmov v0.d[1], x3 (this appears to be the corresponding instruction in the binary to the fmov v0,x3 instruction gtirb-pprinter emits).

[avncharlie]

Doing some more testing with AI, there are a few more instructions causing similar errors:

fmov — drops .d[1]

• pprinter emits fmov vN,xN when it should emit fmov vN.d[1],xN
• pprinter emits fmov xN,vN when it should emit fmov xN,vN.d[1]

mov (umov alias) — drops lane qualifier

• pprinter emits mov xN,vN when it should emit mov xN,vN.d[0]
• pprinter emits mov wN,vN when it should emit mov wN,vN.s[0]

dup — drops source lane qualifier

• pprinter emits dup vN.2d,vN when it should emit dup vN.2d,vN.d[1]