[dynarmic] Refactoring to reduce latency hit from recompilation #358

Open
Lizzie wants to merge 22 commits from liz-dynarmic-latency-improvments into master
Showing only changes of commit 623a732ef7 - Show all commits

View file

@ -4,7 +4,7 @@ The main way of accessing memory in JITed programs is via an invoked function, s
The solution? Delegate invalid accesses to a dedicated arena, similar to a swap. The main idea behind such mechanism is to allow the OS to transmit page faults from invalid accesses into the JIT translator directly, bypassing address space calls, while this sacrifices i-cache coherency, it allows for smaller code-size and "faster" throguhput.
Many kernels however, do not support fast signal dispatching (Solaris, *BSD). Only Linux and Windows support relatively "fast" signal dispatching. Hence this feature is better suited for them only.
Many kernels however, do not support fast signal dispatching (Solaris, OpenBSD, FreeBSD). Only Linux and Windows support relatively "fast" signal dispatching. Hence this feature is better suited for them only.
![Host to guest translation](./HostToGuest.svg)