The body was lagging behind the text. This was fixed by calling an updated version of init() on settings change.
This introduced spikes in the frametime graph, so the data had to become members and m_force_update was adjusted to not interfere with data aquisition.
The frametime_timer was restarted at the end of the update method, instead of immediately after reading its value.
This means that the frametime was lower than expected when an expensive task was executed in-between.
Some DMA block entries are stubs whose parents are DMA_block_EXT
entries.
Performing load() in this case becomes a memcpy(address,
same_address_again, length) which wastes performance and introduces
bugs.
* Use atomic waitables instead instead of global thread wait as often as possible.
* Add ::is_stopped() and and ::is_paued() which can be used in atomic loops and with atomic wait. (constexpr cpu flags test functions)
* Fix notification bug of sys_spu_thread_group_exit/terminate. (old bug, enhanced by #9117)
* Function time statistics at Emu.Stop() restored. (instead of current "X syscall failed with 0x00000000 : 0")
* Multi-thread shader compilation
This offers a huge improvement in startup performance. With around 13,000 shaders we go from ~1:30 to under 10 seconds. It looks like this was the original intention of the author given the outer scope recompile variable.
Signed-off-by: Alex Saveau <saveau.alexandre@gmail.com>
Disable features still unimplemented by lavapipe when using it:
- samplerAnisotropy
- shaderStorageBufferArrayDynamicIndexing
- wideLines
as of mesa 21.1.0-dev (aea36ee05e9, 2020-02-10)
- Avoids a silly situation where a texture is discarded and an identical copy created immediately afterward.
Unfortunately allocating memory blocks is really slow so avoid it as much as possible.
* Set the ground for RSX modes of register editor and insttruction editor, do not use shared ptrs directly.
* Make register editor and instruction editor modeless to allow to copypaste values from thread context etc in the background.
- The driver seems to cache page mapping as long as allocation has not been removed
- This is undesirable as we cannot stop the emulator to remove stale allocations every time a page is unmapped
