| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
fs_inst::overwrites_reg is rather easy to misuse because it cannot
tell how large the register region starting at 'reg' is, so in cases
where the destination region starts after 'reg' it may give a
misleading result. regions_overlap() is somewhat more verbose to use
but handles arbitrary overlap correctly so it should generally be used
instead.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous regs_written field can be recovered by rewriting each
rvalue reference of regs_written like 'x = i.regs_written' to 'x =
DIV_ROUND_UP(i.size_written, reg_unit)', and each lvalue reference
like 'i.regs_written = x' to 'i.size_written = x * reg_unit'.
For the same reason as in the previous patches, this doesn't attempt
to be particularly clever about simplifying the result in the interest
of keeping the rather lengthy patch as obvious as possible. I'll come
back later to clean up any ugliness introduced here.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is in preparation for dropping fs_inst::regs_read and
::regs_written in favor of more accurate alternatives expressed in
byte units. The main reason these wrappers are useful is that a
number of optimization passes implement dataflow analysis with
register granularity, so these helpers will come in handy once we've
switched register offsets and sizes to the byte representation. The
wrapper functions will also make sure that GRF misalignment (currently
neglected by most of the back-end) is taken into account correctly in
the calculation of regs_read and regs_written.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The fs_reg::offset field in byte units introduced in this patch is a
more straightforward alternative to the current register offset
representation split between fs_reg::reg_offset and ::subreg_offset.
The split representation makes it too easy to forget about one of the
offsets while dealing with the other, which has led to multiple
back-end bugs in the past. To make the matter worse the unit
reg_offset was expressed in was rather inconsistent, for uniforms it
would be expressed in either 4B or 16B units depending on the
back-end, and for most other things it would be expressed in 32B
units.
This encodes reg_offset as a new offset field expressed consistently
in byte units. Each rvalue reference of reg_offset in existing code
like 'x = r.reg_offset' is rewritten to 'x = r.offset / reg_unit', and
each lvalue reference like 'r.reg_offset = x' is rewritten to
'r.offset = r.offset % reg_unit + x * reg_unit'.
Because the change affects a lot of places and is rather non-trivial
to verify due to the inconsistent value of reg_unit, I've tried to
avoid making any additional changes other than applying the rewrite
rule above in order to keep the patch as simple as possible, sometimes
at the cost of introducing obvious stupidity (e.g. algebraic
expressions that could be simplified given some knowledge of the
context) -- I'll clean those up later on in a second pass.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
|
|
|
| |
We weren't checking the fs_inst::target field when comparing whether
two instructions are equal. For FB writes it doesn't matter because
they aren't CSE-able anyway, but this would have become a problem with
FB reads which are expression-like instructions.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
| |
This will prevent some shader-db regressions when we start plumbing
logical sends through the optimizer.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
| |
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
|
|
|
|
|
|
| |
This generalizes the current fs_inst::force_sechalf flag to allow
specifying channel enable groups other than 0 or 8. At some point it
will likely make sense to fix the vec4 generator to support arbitrary
execution groups and then move the definition of fs_inst::group into
backend_instruction (e.g. so we can do FP64 in the VEC4 back-end).
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
|
|
|
| |
These can be easily represented in the IR as a MOV instruction with
strided source so they seem rather redundant.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
|
|
|
|
| |
This will allow the SIMD lowering pass to split 32-wide varying pull
constant loads (not natively supported by the hardware) into 16-wide
instructions.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If the LOAD_PAYLOAD instruction only has header sources it's possible
for the number of registers written to be less than or equal to the
SIMD component size, in which case it would take the single-MOV path
at the bottom which would cause the channel enable masks to be applied
incorrectly to the header contents and/or cause it to write past the
end of the allocated temporary. If the instruction is either
LOAD_PAYLOAD or doesn't write exactly one component the MOV path is
going to mess up the program so just don't use it.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
|
|
|
|
|
|
|
|
| |
v2 (Sam):
- Fix line width (Topi).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It's not correct to CSE these multiplies
mul.sat dst1, -a, b
mul.sat dst2, a, b
by emitting a negated MOV from dst1 to dst2:
mul.sat dst1, -a, b
mov dst2, -dst1
Take 2.0*2.0 for example. The first multiply would produce 0.0 and the
second would produce 1.0.
Fixes bad generated code in 18 to 22 shaders:
instructions in affected programs: 432 -> 464 (7.41%)
helped: 4
HURT: 18
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
|
|
|
|
| |
The vec4 backend will lower it.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
| |
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The geometry and tessellation control shader stages both read from
multiple URB entries (one per vertex). The thread payload contains
several URB handles which reference these separate memory segments.
In GLSL, these inputs are represented as per-vertex arrays; the
outermost array index selects which vertex's inputs to read. This
array index does not necessarily need to be constant.
To handle that, we need to use indirect addressing on GRFs to select
which of the thread payload registers has the appropriate URB handle.
(This is before we can even think about applying the pull model!)
This patch introduces a new opcode which performs a MOV from a
source using VxH indirect addressing (which allows each of the 8
SIMD channels to select distinct data.)
Based on a patch by Jason Ekstrand.
v2: Rename from INDIRECT_THREAD_PAYLOAD_MOV to MOV_INDIRECT; make it
a bit more generic. Use regs_read() instead of hacking up the
register allocator. (Suggested by Jason Ekstrand.)
v3: Fix regs_read() to be more accurate for small unaligned regions.
Also rebase on Matt's work.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com> [v3]
Reviewed-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> [v1]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
HW_REGs are (were!) kind of awful. If the file was HW_REG, you had to
look at different fields for type, abs, negate, writemask, swizzle, and
a second file. They also caused annoying problems like immediate sources
being considered scheduling barriers (commit 6148e94e2) and other such
nonsense.
Instead use ARF/FIXED_GRF/MRF for fixed registers in those files.
After a sufficient amount of time has passed since "GRF" was used, we
can rename FIXED_GRF -> GRF, but doing so now would make rebasing awful.
Reviewed-by: Emil Velikov <emil.velikov@collabora.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
| |
The 2-bit hardware register file field is ARF, GRF, MRF, IMM.
Rename GRF to VGRF (virtual GRF) so that we can reuse the GRF name to
mean an assigned general purpose register.
Reviewed-by: Emil Velikov <emil.velikov@collabora.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
In addition to combining another field, we get replace silliness like
"reg.reg" with something that actually makes sense, "reg.nr"; and no one
will ever wonder again why dst.reg isn't a dst_reg.
Moving the now 16-bit nr field to a 16-bit boundary decreases code size
by about 3k.
Reviewed-by: Emil Velikov <emil.velikov@collabora.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
| |
Reviewed-by: Emil Velikov <emil.velikov@collabora.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Generated by
sed -i -e 's/\.bits\././g' *.c *.h *.cpp
sed -i -e 's/dw1\.//g' *.c *.h *.cpp
and then reverting changes to comments in gen7_blorp.cpp and
brw_fs_generator.cpp.
There wasn't any utility offered by forcing the programmer to list these
to access their fields. Removing them will reduce churn in future
commits.
This is C11 (and gcc has apparently supported it for sometime
"compatibility with other compilers")
See https://gcc.gnu.org/onlinedocs/gcc/Unnamed-Fields.html
Reviewed-by: Emil Velikov <emil.velikov@collabora.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An untyped surface read is volatile because it might be affected by a
write.
In the ES31-CTS.compute_shader.resources-max test, two back to back
read/modify/writes of an SSBO variable looked something like this:
r1 = untyped_surface_read(ssbo_float)
r2 = r1 + 1
untyped_surface_write(ssbo_float, r2)
r3 = untyped_surface_read(ssbo_float)
r4 = r3 + 1
untyped_surface_write(ssbo_float, r4)
And after CSE, we had:
r1 = untyped_surface_read(ssbo_float)
r2 = r1 + 1
untyped_surface_write(ssbo_float, r2)
r4 = r1 + 1
untyped_surface_write(ssbo_float, r4)
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This instruction will translate to the MUL/MACH sequence that computes
the high 32-bits of the result of a 64-bit multiply. Before Gen8
integer operations that used the accumulator were limited to 8-wide,
but the SIMD lowering pass can easily be hooked up to sidestep this
limitation, we just need a virtual opcode to represent the MUL/MACH
sequence in the IR.
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
| |
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
is on.
This assertion was meant to catch code inadvertently escaping the
control flow jail determined by the group of channel enable signals
selected by some caller, however it seems useful to be able to
increase the default execution size as long as force_writemask_all is
enabled, because force_writemask_all is an explicit indication that
there is no longer a one-to-one correspondence between channels and
SIMD components so the restriction doesn't apply.
In addition reorder the calls to fs_builder::group and ::exec_all in a
couple of places to make sure that we don't temporarily break this
invariant in the future for instructions with exec_size higher than
the dispatch width.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As of now, the width field is no longer used for anything. The width field
"seemed like a good idea at the time" but is actually entirely redundant
with the instruction's execution size. Initially, it gave us the ability
to easily set the instructions execution size based entirely on register
widths. With the builder, we can easiliy set the sizes explicitly and the
width field doesn't have as much purpose. At this point, it's just
redundant information that can get out of sync so it really needs to go.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Acked-by: Francisco Jerez <currojerez@riseup.net>
|
|
|
|
|
|
|
|
| |
There are a variety of places where we use dst.width / 8 to compute the
size of a single logical channel. Instead, we should be using exec_size.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Acked-by: Francisco Jerez <currojerez@riseup.net>
|
|
|
|
|
| |
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Acked-by: Francisco Jerez <currojerez@riseup.net>
|
|
|
|
| |
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
temporary.
LOAD_PAYLOAD instructions need the same treatment as any other
generator instructions, at least FB writes and typed surface messages
will need a payload built with non-zero execution controls.
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
|
|
|
|
| |
Most of these fields affect the behaviour of the instruction so it
could actually break the program if we CSE a pair of otherwise
matching instructions with different values of these fields.
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The newly reworked instruction is far more straightforward than the
original. Before, the LOAD_PAYLOAD instruction was lowered by a the
complicated and broken-by-design pile of heuristics to try and guess
force_writemask_all, exec_size, and a number of other factors on the
sources.
Instead, we use the header_size on the instruction to denote which sources
are "header sources". Header sources are required to be a single physical
hardware register that is copied verbatim. The registers that follow are
considered the actual payload registers and have a width that correspond's
to the LOAD_PAYLOAD's exec_size and are treated as being per-channel. This
gives us a fairly straightforward lowering:
1) All header sources are copied directly using force_writemask_all and,
since they are guaranteed to be a single register, there are no
force_sechalf issues.
2) All non-header sources are copied using the exact same force_sechalf
and force_writemask_all modifiers as the LOAD_PAYLOAD operation itself.
3) In order to accommodate older gens that need interleaved colors,
lower_load_payload detects when the destination is a COMPR4 register
and automatically interleaves the non-header sources. The
lower_load_payload pass does the right thing here regardless of whether
or not the hardware actually supports COMPR4.
This patch commit itself is made up of a bunch of smaller changes squashed
together. Individual change descriptions follow:
i965/fs: Rework fs_visitor::LOAD_PAYLOAD
We rework LOAD_PAYLOAD to verify that all of the sources that count as
headers are, indeed, exactly one register and that all of the non-header
sources match the destination width. We then take the exec_size for
LOAD_PAYLOAD directly from the destination width.
i965/fs: Make destinations of load_payload have the appropreate width
i965/fs: Rework fs_visitor::lower_load_payload
v2: Don't allow the saturate flag on LOAD_PAYLOAD instructions
i965/fs_cse: Support the new-style LOAD_PAYLOAD
i965/fs_inst::is_copy_payload: Support the new-style LOAD_PAYLOAD
i965/fs: Simplify setup_color_payload
Previously, setup_color_payload was a a big helper function that did a
lot of gen-specific special casing for setting up the color sources of
the LOAD_PAYLOAD instruction. Now that lower_load_payload is much more
sane, most of that complexity isn't needed anymore. Instead, we can do
a simple fixup pass for color clamps and then just stash sources
directly in the LOAD_PAYLOAD. We can trust lower_load_payload to do the
right thing with respect to COMPR4.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
| |
This commit adds a new is_copy_payload helper to fs_inst that takes the
place of the similarly named functions in cse and register coalesce. The
two is_copy_payload functions in CSE and register coalesce were subtly
different and potentially subtly broken. The new version unifies the two
and should be more correct.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
| |
v2: Get rid of the block parameter and make src a const reference
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
| |
v2: Save some CPU cycles by doing 'return progress' rather than
'depth++' in the discard jump special case.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
| |
v2: Style fixes.
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
|
|
|
|
|
| |
a * b is equivalent to -a * -b, and the previous code was failing at
that.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89961
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
mul x, -y is equivalent to mul -x, y; and mul x, y is the negation of
mul x, -y.
With NIR:
total instructions in shared programs: 6167779 -> 6161193 (-0.11%)
instructions in affected programs: 983511 -> 976925 (-0.67%)
helped: 4106
HURT: 16
GAINED: 18
LOST: 7
Without NIR:
total instructions in shared programs: 6192323 -> 6185299 (-0.11%)
instructions in affected programs: 987875 -> 980851 (-0.71%)
helped: 4146
HURT: 16
GAINED: 16
LOST: 0
|
|
|
|
|
|
|
|
| |
Create a backend_inst::is_commutative() method to replace two static
functions that did the exact same thing.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Right now virtual GRF book-keeping and allocation is performed in each
visitor class separately (among other hundred different things),
leading to duplicated logic in each visitor and preventing layering as
it forces any code that manipulates i965 IR and needs to allocate
virtual registers to depend on the specific visitor that happens to be
used to translate from GLSL IR.
v2: Use realloc()/free() to allocate VGRF book-keeping arrays (Connor).
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When performing common subexpression elimination on instructions with
non-null destinations we emit a MOV to copy the result to a new
register that must have no other uses. In the case of:
cmp.g.f0.0(8) null:D, vgrf43:F, 0.500000f
...
cmp.g.f0.0(8) vgrf113:D, vgrf43:F, 0.500000f
we put the first instruction in the AEB and decided that we could reuse
its result when we found the second. Unfortunately, that meant that we'd
emit a MOV from the first's destination, which is null.
Don't do anything if the entry's destination is null and the
instruction's destination is non-null.
Tested-by: Tapani Pälli <tapani.palli@intel.com>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Safe from causing optimization loops, since we don't constant propagate
VF arguments.
(for this and the previous patch):
total instructions in shared programs: 4289075 -> 4271932 (-0.40%)
instructions in affected programs: 1616779 -> 1599636 (-1.06%)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
|
|
|
|
|
|
| |
texture_offset was only used by some texturing operations, and offset
was only used by spill/unspill and some URB operations. These fields are
never used at the same time.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
|
|
|
|
|
|
|
|
| |
Multiplication is commutative.
instructions in affected programs: 48314 -> 47954 (-0.75%)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Due to the implicit move-from-GRF, unary math looks a lot like the Gen6+
math instruction: it's a single instruction (SEND) with a GRF source.
The difference is that it also implicitly clobbers a message register.
The only visible effect is that CSE will remove the MRF-clobbering from
later math operations. This should be fine; compute_to_mrf and
remove_redundant_mrf_writes don't look at the values populated by
implied writes, so they can't rely on those values being present.
Less interference may actually help those passes make more progress.
Binary math is still problematic, since it involves a separate MOV
instruction to load the second operand. We continue disallowing CSE for
binary math operations.
total instructions in shared programs: 3340303 -> 3340100 (-0.01%)
instructions in affected programs: 26927 -> 26724 (-0.75%)
Nothing hurt, gained, or lost. ~6% reduction on a few shaders.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
|