Current ABI differences between DMD and LDC

[solidstate1991]

I'm planning to write a few functions purely in x86 Assembly 
(later adding ARM and other platforms), and I'm planning to 
support both DMD and LDC. As I'm running into massive issues when 
compiling and running very long assembly codes in LDC (see this: 
https://github.com/ZILtoid1991/CPUblit/blob/master/src/CPUblit/composing.d), I came up with the solution of using an external assembler like NASM, then add its build commands to DUB's prebuild commands. Should I go with C or D API for function calls? Or issue a D improvement proposal on fixing the SIMD support?