Incorrect encoding of vmovddup xmm16, [rax+64]

[herumi]

When assembling vmovddup xmm16, [rax+64] with NASM version 3.01 (and version 3.01rc9), the output encoding is 62E1FF08124004.
However, when this byte sequence is disassembled with Intel XED, it is decoded as vmovddup xmm16, qword ptr [rax+0x20].

The correct encoding should be 62E1FF08124008.
This instruction was encoded correctly with nasm2.

Reproduction

; test.asm
vmovddup xmm16, [rax+64]

Assembling with nasm3:

% nasm3 -l test.lst -f elf64 test.asm
% cat test.lst
     1 00000000 62E1FF08124004          vmovddup xmm16, [rax+64]

Disassembling with Intel XED:

% xed -64 -d 62E1FF08124004
62E1FF08124004
ICLASS:     VMOVDDUP
CATEGORY:   DATAXFER
EXTENSION:  AVX512EVEX
IFORM:      VMOVDDUP_XMMf64_MASKmskw_MEMf64_AVX512
ISA_SET:    AVX512F_128
ATTRIBUTES: DISP8_MOVDDUP MASKOP_EVEX
SHORT:      vmovddup xmm16, qword ptr [rax+0x20]

This suggests that the displacement (disp8) is incorrectly encoded in NASM 3.01 for this instruction.