If we are flushing multiple sequential guest GPRs then we can store two in a single STP instruction.
Ikaruga does this quite a bit in their blocks where they do an lmw at the very end and then we have to flush them all.
Typically cuts 16 STR instructions down to 8 STP instructions there.
This instruction is fairly heavily used by Ikaruga to load a bunch of registers from the stack.
In particular at the start of the second stage is a block that takes up ~20% CPU time that includes a usage of lmw to load half of the guest
registers.
Basic thing optimized here is changing from a single 32bit LDR to potentially a single 128bit LDR.
a single 32bit LDR is fairly slow, so we can optimize a few ways.
If we have four or more registers to load, do a 64bit LDP in to two host registers, byteswap, and then move the high 32bits of the host registers in
to the correct mapped guest register locations.
If we have two registers to load then do a 32bit LDP which will load two guest registers in a single instruction.
and then if we have only one register left to load, load it as before.
This saves quite a bit of cycles since the Cortex-A57 and A72's LDR instruction takes a few cycles.
Each 32bit LDR takes 4 cycles latency, plus 1 cycle for post-index(which typically happens in parallel.
Both the 32bit and 64bit LDP take the same amount of latency.
So we are improving latencies and reducing code bloat here.
This is a bug that crops if BindToRegister() is called multiple times in a row without a R() function call between them.
How to reproduce the bug:
1) Have a completely filled cache with no host register remaining
2) Call BindToRegister() with different guest registers
3) Don't call R() between the BindToRegister() calls.
This issue typically wouldn't be seen for a couple of reasons. Typically we have /plenty/ of registers in the cache, and in most cases we only call
BindToRegister() once per instruction. In the off chance that it is called multiple times, it wouldn't update the last used counts and would flush the
same register as the previous call to it.
Samsung updated the video drivers on the SGS6 which introduced a bug when disabling vsync.
Both the driver versions are r5p0, but the md5sums of the blob differ.
To work around the issue, make sure to never disable vsync by calling eglSwapInterval.
We can't actually determine the driver version on Android yet.
So until the driver version lands that displays the driver version string in the GL_VERSION string
we will need to keep this workaround enabled at all times, which is a bit annoying.
Current mali drivers return the video driver version in one of the EGL strings you can query.
The issue with that is that Android eats all of those strings, so we can't query it.
OpenGL ES 3.2 adds a few things we care about supporting in core. In particular:
- GL_{ARB,EXT,OES}_draw_elements_base_vertex
- KHR_Debug
- Sample Shading
- GL_{ARB,EXT,OES,NV}_copy_image
- Geometry shaders
- Geometry shader instancing (If they support GL_{EXT,OES}_geometry_point_size)
Nvidia was the first to release an OpenGL ES 3.2 driver which I uesd to test this on.
This also enables GS Instancing on GLES 3.1 hardware if it supports all of the required extensions.
In particular this optimizes the case where a 32bit float is loaded via lfs, and then used in double operations.
This happens very often in Gekko based code because the best way to load a 32bit value as a double is lfs since it automatically turns in to a double value.
There are a few other implications of this in practice as well. Like if both of the paired registers are loaded via psq_l and then used in double
operations it would be improved.
Also if we implement a double register we've got to be careful to make sure we understand if it is in "lower" register or the full 128bit register.
Ryan Houdek