This new development release adds support for Intel MPX instructions.

Because few of these new instructions apply special meaning to components of the address of memory operand, there is a special syntax added (after NASM) to allow full level of control over these fields. The classic memory operand can also be used, but it is reliable only in basic cases. The new syntax uses "," to separate address within square brackets into two parts, the meaning of these parts depends on the instruction.

In case of BNDMK, first sub-address is the base that will be loaded into lower bound (either 0 or a register of mode-appropriate size), and the second one is the remaining part of the address for upper bound (it can be interpreted as the length of bounded area). The two instructions shown below demonstrate how this syntax allows to distinguish functionally different encodings of the same effective address:
In case of BNDLDX and BNDSTX, first sub-address of memory operand is any address without an index register, and the second sub-address should be either 0 or an index register without a multiplier other than 1. There are some inconsistencies in Intel manuals that make it unclear whether it is possible to use displacement without base register to form the base address for these instructions, so I decided to allow such form just in case. I will test if it actually works when I have an opportunity.

i downloaded the new fasmw.dll and Get_Version() still returns 1.67

fasm.dll is regularly updated here. Perhaps you downloaded it from some other source that I do not maintain.