| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit switches us from the current setup of using hash sets for
use/def sets to using linked lists. Doing so should save us quite a bit of
memory because we aren't carrying around 3 hash sets per register and 2 per
SSA value. It should also save us CPU time because adding/removing things
from use/def sets is 4 pointer manipulations instead of a hash lookup.
Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists:
GLSL IR Only: 586.4 +/- 1.653833
NIR with hash sets: 675.4 +/- 2.502108
NIR + use/def lists: 641.2 +/- 1.557043
I also ran a memory usage experiment with Ken's patch to delete GLSL IR and
keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over
and above what we gained by deleting the GLSL IR on the same dota trace.
On the code complexity side of things, some things are now much easier and
others are a bit harder. One of the operations we perform constantly in
optimization passes is to replace one source with another. Due to the fact
that an instruction can use the same SSA value multiple times, we had to
iterate through the sources of the instruction and determine if the use we
were replacing was the only one before removing it from the set of uses.
With this patch, uses are per-source not per-instruction so we can just
remove it safely. On the other hand, trying to iterate over all of the
instructions that use a given value is more difficult. Fortunately, the
two places we do that are the ffma peephole where it doesn't matter and GCM
where we already gracefully handle duplicates visits to an instruction.
Another aspect here is that using linked lists in this way can be tricky to
get right. With sets, things were quite forgiving and the worst that
happened if you didn't properly remove a use was that it would get caught
in the validator. With linked lists, it can lead to linked list corruption
which can be harder to track. However, we do just as much validation of
the linked lists as we did of the sets so the validator should still catch
these problems. While working on this series, the vast majority of the
bugs I had to fix were caught by assertions. I don't think the lists are
going to be that much worse than the sets.
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
|
| |
Acked-by: Connor Abbott <[email protected]>
Reviewed-by: Rob Clark <[email protected]>
|
|
|
|
|
|
|
| |
v2: Don't use C99 when iterating over the list
Acked-by: Connor Abbott <[email protected]>
Reviewed-by: Rob Clark <[email protected]>
|
|
|
|
|
|
|
| |
v2: Use LIST_ENTRY instead of container_of in iterators
Acked-by: Connor Abbott <[email protected]>
Reviewed-by: Rob Clark <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The linked list in gallium is pretty much the kernel list and we would like
to have a C-based linked list for all of mesa. Let's not duplicate and
just steal the gallium one.
Acked-by: Connor Abbott <[email protected]>
Reviewed-by: Rob Clark <[email protected]>
|
|
|
|
|
| |
Acked-by: Connor Abbott <[email protected]>
Reviewed-by: Rob Clark <[email protected]>
|
|
|
|
|
|
|
| |
We were rolling our own rewrite_src variant in copy-propagation. Let's
stop doing that and use the ones in core NIR.
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
| |
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
| |
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The out-of-SSA pass was one of the first passes written when getting SSA
up-and-going (for obvious reasons). As such, it came before a lot of the
nifty SSA-based helpers were introduced. This commit modernizes it so that
we're no longer doing nearly as much manual banging on use/def sets.
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
| |
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
The former logic would copy the saturate up to any mul with an immediate
if there was a subsequent mul with a saturate. However we only want to
do that if we collapsed 2 muls by multiplying their immediates (or were
able to put the immediate in as a post-multiplier).
Signed-off-by: Ilia Mirkin <[email protected]>
Cc: [email protected]
|
|
|
|
|
|
|
|
| |
Nothing produces it, and nothing can consume it.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
| |
All paths that produce GLSL IR for NIR lower ir_unop_log. All paths
that consume NIR will explode if they geta nir_op_flog.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
| |
Nothing produces it, and nothing can consume it.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
| |
All paths that produce GLSL IR for NIR lower ir_unop_exp. All paths
that consume NIR will explode if they geta nir_op_fexp.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
| |
It's a weird thing that provides some values related to 2**x. It's also
already handled by a case in the switch.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
| |
Originally I wrote that removing the first parameter doesn't work but
I didn't know why. I now found a mention of this in the PRM so it's
probably worthing adding it to the comment.
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
v2: Fix the name of the entry point in the error messages.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
This parameter was added in OpenGL 4.3 and GL_ARB_direct_state_access.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This is needed to implement glGetVertexArrayIndexediv and
glGetVertexArrayIndexed64iv.
v2: Make the vao parameter const.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
This is needed to implement VertexArrayBindingDivisor.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
This is needed to implement VertexArrayAttribBinding.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
This is needed to implement VertexArrayAttrib*Format.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
| |
The only difference between these functions is the legal types and
sizes, so consolidate the code into a single vertex_attrib_format()
function and call it from all three entry points.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
| |
This is needed to implement VertexArrayVertexBuffer and
VertexArrayVertexBuffers.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
v2: Add a doxygen comment.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This saves the cost of repeated hash table lookups when the same
vertex array object is referenced in a sequence of calls such as:
glVertexArrayAttribFormat(vao, ...);
glVertexArrayAttribBinding(vao, ...);
glEnableVertexArrayAttrib(vao, ...);
...
Note that VAO's are container objects that are not shared between
contexts.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
| |
This is a convenience function that generates GL_INVALID_OPERATION
when the array object doesn't exist.
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
| |
v2: Update the documentation for gen_vertex_arrays().
Reviewed-by: Laura Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
opt_sampler_eot enables a direct write to framebuffer from a sample.
In order to do this the sample message needs to have a message header
so if there wasn't one already then the function adds one. In addition
the function sets the destination register to null because it's no
longer used. However it was only doing this in cases where it was
adding a message header. This patch just moves setting the destination
so that it happens even if there's a messge header. In practice this
doesn't seem to make any difference but it's a bit cleaner.
Reviewed-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Commit 94ee908448 added a header size parameter to the function to
create the LOAD_PAYLOAD instruction. However this broke
opt_sampler_eot which manually constructs the instruction and so
wasn't setting the header_size. This ends up making the parameters for
the send message all have the wrong location and it all falls apart.
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
| |
Suggested-by: Axel Davy <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
Signed-off-by: Martin Peres <[email protected]>
|
|
|
|
|
|
| |
Add to GL3.txt and release notes.
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This takes a different approach to previously, we cannot index into the
inputMapping with anything but the mesa attribute index, so we can't use
the just add one to index trick, we need more info to add one to it
after we've mapped the input.
(Fixed copy propgation and cleaned up a little)
v2: drop float64 format check, just attr->Doubles.
merge enable patch.
v3: cleanup code a bit.
v3.1: minor review fixups (comment, newline) (Ilia)
Reviewed-by: Ilia Mirkin <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds support in the vbo and array code to handle
double vertex attributes.
v0.2: merge code to handle doubles in vbo layer.
v1: don't use v0, merge api_array elt code.
Acked-by: Ilia Mirkin <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The spec is vague all over the place about this, but this seems
to be the intent, we can probably make this optional later if
someone makes hw that cares and writes a driver.
Basically we need to double count some of the d types but
only for totalling not for slot number assignment.
Reviewed-by: Ilia Mirkin <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
| |
instead of doing the attempts at dual slot handling here,
let the backend do it.
Reviewed-by: Ilia Mirkin <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Just more boilerplate stuff.
v2:
bad fallthrough on versioning,
this is my ugly but self contained solution (Ian)
Reviewed-by: Ilia Mirkin <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
Just add the boilerplate bits.
v2: add to version.c
Reviewed-by: Ilia Mirkin <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
| |
This just adds the glapi bits.
Reviewed-by: Ilia Mirkin <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|