tANS decoding from a bit stream (x86_64 assembly)

tans_assembly_x86_64

LBB226_3:
  movq  1072(%rdi), %r11       load state_idx
  movq  1088(%rdi), %rbx       load ptr to ANS nodes
  movl  %eax, %ecx             (copy bits_past_byte to do arithmetic later)
  shrl  $3, %ecx               bytes_to_advance = bits_past_byte / 8
  addq  %rcx, %r8              byte_idx += bytes_to_advance
  andl  $7, %eax               bit_idx = bit_idx % 8
  movq  (%r9,%r8), %r14        word = bytes[byte_idx]
  movl  %eax, %ecx             (copy bit_idx to do arithmetic later)
  shrq  %cl, %r14              word >>= bit_idx
  shlq  $4, %r11               (multiply node index due to struct size)
  movl  12(%rbx,%r11), %ecx    bits_to_read = node[state_idx].bits_to_read
  addl  %ecx, %eax             bits_past_byte = bit_idx + bits_to_read
  negb  %cl                    excess_bits = word_size -    
  shlq  %cl, %r14              word <<= excess_bits (with next line, masking out msb at end of buffer)
  shrq  %cl, %r14              word >>= excess_bits
  addq  (%rbx,%r11), %r14      state_idx = node[state_idx].next_base_state_idx + word     
  movq  %r14, 1072(%rdi)       decoder.state_idx = state_idx
  movl  8(%rbx,%r11), %ecx     token = tokens[token_idx]
  movl  %ecx, 40(%rdi,%r10,4)  dst[i] = token
  leaq  1(%r10), %rcx          i += 1
  movq  %rcx, %r10             (copy i for next iteration; honestly idk why rcx was used here)
  cmpq  %rcx, %rdx             check if i != n...
  jne   LBB226_3               ... and loop again if so