| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
| |
Currently we were two restrictive, and would insert an output move in
cases like: MOV OUT[0], IN[0].xyzw
Loosen the restriction to allow the current instruction to appear in the
neighbor list but only at it's current possition.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When we have something like:
MOV OUT[n], IN[m].wzyx
the existing grouping code was missing a potential conflict. Due to
input needing to be sequential scalar regs, we have:
IN: x <-> y <-> z <-> w
which would be grouped to:
OUT: w <-> z2 <-> y2 <-> x (where the 2 denotes a copy/mov)
but that can't actually work. We need to realize that x and w are
already in the same chain, not just that they aren't both already in
new chain being built.
With this fixed, we probably no longer need the hack from f68f6c0.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
| |
No longer used, so drop the extra arg to ir3_instr_create()
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
| |
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
| |
This shows up with a glamor shader, which does a TXF and uses the result
for conditional kill. Before we wouldn't group the fanin (collect)
neighbors which need to be allocated adjacently at RA, resulting in
badness.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This shuffles things around to allow the shader to have multiple basic
blocks. We drop the entire CFG structure from nir and just preserve the
blocks. At scheduling we know whether to schedule conditional branches
or unconditional jumps at the end of the block based on the # of block
successors. (Dropping jumps to the following instruction, etc.)
One slight complication is that variables (load_var/store_var, ie.
arrays) are not in SSA form, so we have to figure out where to put the
phi's ourself. For this, we use the predecessor set information from
nir_block. (We could perhaps use NIR's dominance frontier information
to help with this?)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
| |
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
| |
These belong in the shader, rather than the block. Mostly a lot of
churn and nothing too interesting. But splitting this out from the
rest of ir3_block reshuffling to cut down the noise in the later
patch.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
| |
Also remove ir3_flatten which was only used by tgsi f/e.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
| |
Use ir3_MOV() builder in a couple of spots, rather than open-coding the
instruction construction. Also add ir3_NOP() builder and use that
instead of open coding.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Be smarter about propagating copies from const or immed, or with abs/neg
modifiers. Also, realize that absneg.s and absneg.f are really "fancy"
mov instructions.
This opens up the possibility to remove more copies. It helps the TGSI
frontend a bit, but will be really needed for the NIR f/e which builds
everything up in SSA form (ie. will *always* insert a mov from const or
immediate).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
| |
I remembered that we are using c99.. which makes some sugary iterator
macros easier. So introduce iterator macros to iterate all src
registers and all SSA src instructions. The _n variants also return
the src #, since there are a handful of places that need this.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Turns out there are scenarios where we need to insert mov's in "front"
of an input. Triggered by shaders like:
VERT
DCL IN[0]
DCL IN[1]
DCL OUT[0], POSITION
DCL OUT[1], GENERIC[9]
DCL SAMP[0]
DCL TEMP[0], LOCAL
0: MOV TEMP[0].xy, IN[1].xyyy
1: MOV TEMP[0].w, IN[1].wwww
2: TXF TEMP[0], TEMP[0], SAMP[0], 1D_ARRAY
3: MOV OUT[1], TEMP[0]
4: MOV OUT[0], IN[0]
5: END
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
If, for example, only the x/y/w components of in.xyzw are actually used,
we still need to have a group of four registers and assign all four
components. The hardware can't write in.xy and in.w to discontiguous
registers. To handle this, pad with a dummy NOP instruction, to keep
the neighbor chain contiguous.
This fixes a problem noticed with firefox OMTC.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|
|
Group inputs/outputs, in addition to fanin/fanout, as they must also
exist in sequential scalar registers. This lets us simplify RA by
working in terms of neighbor groups.
NOTE: has the slight problem that it can't optimize out mov's for things
like:
MOV OUT[n], IN[m]
To avoid this, instead of trying to figure out what mov's we can
eliminate, we first remove all mov's prior to grouping, and then
re-insert mov's as needed while grouping inputs/outputs/fanins.
Eventually we'd prefer the frontend to not insert extra mov's in the
first place (so we don't have to bother removing them). This is the
plan for an eventual NIR based frontend, so separate out the instr
grouping (which will still be needed for NIR frontend) from the mov
elimination (which won't).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
|