target -> The evaluated target of the load or the store, 0 if the opcode is not a memory access
load -> The evaluated target of a load, is 0 if the opcode is not loading
store -> The evaluated target of a store, is 0 if the opcode is not storing
Fixes the disassembly widget mouse click handler
Fixes the breakpoint dialog not setting the breakpoint address or condition cpu (oops)
The Azure Ubuntu package server, in particular, seems to be very
unreliable. Hopefully within 10 attempts it'll complete the package
download in such cases.
3 is a nice compromise if you have 4 core or more but for some games higher numbers like 8 or even higher would work, however for a lot of games it would perhaps even reduce the performance compared to 3 software rendering threads or in more likelihood have diminishing returns.
Keep in mind that this is just a general situation, in actual use it might be less extra performance (if at all) as you have expected.
Lowering 100 ms audio latency to 60 ms.
Makes 100 mixing latency with 20 output latency (120) into a new default of 80 ms which is 50% lower but still is enough headroom as 40 ms even for less capable machines is too high.
Adding tooltip for Affinity Control
Changing tooltip for Cycle Skip from None to Disabled as that is a valid option.
PCSX2 does not even boot on SSE2 or SSE3 systems and we do not use SSE4a so we should remove it SSE4.1 / 4.2 is a requirement so no use being here either.
Consistent with the other models, and saves some awkward usage in the cpu widget where we have to interpret the variant as a string, to then convert it to a uint. I like this better as it provides an easy interface to get the 'raw' breakpoint data out of the model.
* VU: Make _x86regs.reg signed, as -1 is supposed to indicate that it is unassigned.
* VU: Convert a few defines to enums.
* microVU: Set bool to true rather than 1.
* xmm: Use std:array in a few spots, and make initializing some spots less verbose.
* iCore: Remove comment, as it's inaccurate. Leaving as s8 for consistency with xmm.