Platform.AArch64.HostTarget

1module Platform.AArch64.HostTarget
2
3import Runtime.NativeHostTarget
4import Platform.AArch64.HostLowering
5
6-- Linux AArch64 as a native host target. Platform.AArch64.HostLowering
7-- translates the program's system calls and machine routines to their
8-- AArch64 forms and Platform.AArch64.NativePhysicalRuntime interprets it; a
9-- command without an AArch64 realization refuses the whole executable.
10-- Floating-point commands are separate binary64 FADD/FSUB/FMUL (never a
11-- fused multiply-add) under the FPCR Linux gives a new process: round to
12-- nearest even, no flush-to-zero. The x86-64 runtime's SSE2 arithmetic under
13-- the default MXCSR implements the same IEEE 754 contract, so every finite
14-- result has the same bits on either host. NaN encodings are not shared (the
15-- two architectures' default NaNs differ in sign); programs that publish
16-- state admit finite values only and refuse a nonfinite one before it can
17-- reach a peer or a checkpoint.
18def nativeHostTargetAArch64Linux : (family NativeHostTarget) =
19  (constructor NativeHostTarget NativeHostTargetOf aarch64HostELFBytes)