dynarmic: Inline exclusive memory accesses

Inlines implementation of exclusive instructions into JITted code,
improving performance of applications relying heavily on these
instructions.

We also fastmem these instructions for additional speed, with
support for appropriate recompilation on fastmem failure.

An unsafe optimization to disable the intercore global_monitor is also
provided, should one wish to rely solely on cmpxchg semantics for
safety.

See also: merryhime/dynarmic#664
This commit is contained in:
merry
2022-02-27 19:40:05 +00:00
parent 96d90be59f
commit 16784e5bb3
17 changed files with 114 additions and 8 deletions

View File

@ -174,6 +174,14 @@ cpuopt_reduce_misalign_checks =
# 0: Disabled, 1 (default): Enabled
cpuopt_fastmem =
# Enable Host MMU Emulation for exclusive memory instructions (faster guest memory access)
# 0: Disabled, 1 (default): Enabled
cpuopt_fastmem_exclusives =
# Enable fallback on failure of fastmem of exclusive memory instructions (faster guest memory access)
# 0: Disabled, 1 (default): Enabled
cpuopt_recompile_exclusives =
# Enable unfuse FMA (improve performance on CPUs without FMA)
# Only enabled if cpu_accuracy is set to Unsafe. Automatically chosen with cpu_accuracy = Auto-select.
# 0: Disabled, 1 (default): Enabled
@ -199,6 +207,11 @@ cpuopt_unsafe_inaccurate_nan =
# 0: Disabled, 1 (default): Enabled
cpuopt_unsafe_fastmem_check =
# Enable faster exclusive instructions
# Only enabled if cpu_accuracy is set to Unsafe. Automatically chosen with cpu_accuracy = Auto-select.
# 0: Disabled, 1 (default): Enabled
cpuopt_unsafe_ignore_global_monitor =
[Renderer]
# Which backend API to use.
# 0 (default): OpenGL, 1: Vulkan