Cc: Jason Ekstrand
---
Seems like we missed fixing this issue during review. Jason, can you commit
this?
src/glsl/nir/nir_lower_vars_to_ssa.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/glsl/nir/nir_lower_vars_to_ssa.c
b/src/glsl/nir/nir_lower_vars_to_ssa.c
index e3
gets included into nir.h like before. In
addition to solving the above problems, using Python and Mako to generate
everything means that it's much easier to add keep information centralized as we
add new things like constant propagation that require per-opcode information.
Signed-off-by: Con
taken from
ir_constant_expressions.cpp.
v2: use Python formatting and get rid of regex's
Signed-off-by: Connor Abbott
---
src/glsl/Makefile.am | 10 +-
src/glsl/Makefile.sources| 3 +-
src/glsl/nir/.gitignore | 1 +
src/glsl/nir/nir_constant_expre
e had before for getting the
constant expressions to a state where they could be compiled as C code.
Connor Abbott (5):
nir: add generated file to .gitignore
nir: use Python to autogenerate opcode information
nir: add new constant folding infrastructure
nir/constant_folding: use the new con
Signed-off-by: Connor Abbott
---
src/glsl/nir/nir_opt_constant_folding.c | 179
1 file changed, 21 insertions(+), 158 deletions(-)
diff --git a/src/glsl/nir/nir_opt_constant_folding.c
b/src/glsl/nir/nir_opt_constant_folding.c
index 6c02582..496aa41 100644
--- a
Reviewed-by: Jason Ekstrand
Signed-off-by: Connor Abbott
---
src/glsl/nir/nir_lower_vars_to_ssa.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/glsl/nir/nir_lower_vars_to_ssa.c
b/src/glsl/nir/nir_lower_vars_to_ssa.c
index 4df9bdd..cdf0023 100644
--- a/src/glsl/nir
Signed-off-by: Connor Abbott
---
src/glsl/nir/.gitignore | 1 +
1 file changed, 1 insertion(+)
create mode 100644 src/glsl/nir/.gitignore
diff --git a/src/glsl/nir/.gitignore b/src/glsl/nir/.gitignore
new file mode 100644
index 000..6d954fe
--- /dev/null
+++ b/src/glsl/nir/.gitignore
Oh, and I forgot... the series is also available at
https://github.com/cwabbott0/mesa nir-opcodes-cleanup
On Fri, Jan 16, 2015 at 4:46 PM, Connor Abbott wrote:
> Hi,
>
> This is a series I had floating around a while. The idea is to have all the
> opcode stuff, including cons
Hi Dylan,
On Fri, Jan 16, 2015 at 7:01 PM, Dylan Baker wrote:
> Hi Conner, I have a couple of things you should change, and a suggestion
> for you below, hopefully it all makes sense.
>
> On Friday, January 16, 2015 04:46:07 PM Connor Abbott wrote:
>> Before, we used a s
On Fri, Jan 16, 2015 at 10:48 PM, Dylan Baker wrote:
> On Friday, January 16, 2015 10:18:43 PM Connor Abbott wrote:
>> Hi Dylan,
>>
>> On Fri, Jan 16, 2015 at 7:01 PM, Dylan Baker wrote:
>> > Hi Conner, I have a couple of things you should change, and a suggestion
On Sat, Jan 17, 2015 at 11:42 AM, ahmad wrote:
> hi.
>
> "#! /usr/bin/env python" corresponds python 3.x series for some major distro
> (arch,fedora ...) and python 2.x for some others.
>
> python 2.x and python 3.x are not source compatible each other.
>
> python 3.x not contains "xrange" funci
way
> it worked out. A lot of the codegen stuff is very confusing and its not at
> all obvious what's going on. I'll give it some thought and see if I can
> come up with a good way to clean it up.
>
> On Jan 16, 2015 3:46 PM, "Connor Abbott" wrote:
>>
&
Shouldn't the subject say "exec_list: Add a
foreach_in_list_reverse_safe macro."?
On Mon, Jan 19, 2015 at 6:31 PM, Matt Turner wrote:
> ---
> src/glsl/list.h | 6 ++
> 1 file changed, 6 insertions(+)
>
> diff --git a/src/glsl/list.h b/src/glsl/list.h
> index 330c17e..85368a4 100644
> --- a/s
> That makes sense. However, in the vec4 backend, we do emit scratch reads
> and writes prior to optimization - in move_grf_array_access_to_scratch().
> I've been meaning to port that over to the scalar backend. At which
> point, SHADER_OPCODE_GEN4_SCRATCH_WRITE would be a problem.
Sort of off-t
On Tue, Jan 20, 2015 at 1:15 PM, Matt Turner wrote:
> On Tue, Jan 20, 2015 at 6:16 AM, Connor Abbott wrote:
>> Shouldn't the subject say "exec_list: Add a
>> foreach_in_list_reverse_safe macro."?
>
> Meh? Shortlog on this file shows glsl2, exec_list, glsl, gls
Assuming you grepped for uses of foreach_list* with nir_phi_src and
made sure there were no more,
Reviewed-by: Connor Abbott
On Tue, Jan 20, 2015 at 7:34 PM, Jason Ekstrand wrote:
> ---
> src/glsl/nir/nir.c | 4 ++--
> src/glsl/nir/nir.h | 3 +
Reviewed-by: Connor Abbott
On Wed, Jan 21, 2015 at 2:14 PM, Jason Ekstrand wrote:
> ---
> src/glsl/nir/nir.h | 37 +
> 1 file changed, 37 insertions(+)
>
> diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
> index 5ebfc5a..7b5794d 10064
Assuming this actually compiles with GCC 4.4...
Reviewed-by: Connor Abbott
On Wed, Jan 21, 2015 at 2:14 PM, Jason Ekstrand wrote:
> Designated initializers with anonymous unions don't work in MSVC or
> GCC < 4.6. With a couple of constructor methods, we don't need them any
Reviewed-by: Connor Abbbott
On Wed, Jan 21, 2015 at 8:25 PM, Jason Ekstrand wrote:
> ---
> src/glsl/nir/nir_search.c | 24 ++--
> 1 file changed, 10 insertions(+), 14 deletions(-)
>
> diff --git a/src/glsl/nir/nir_search.c b/src/glsl/nir/nir_search.c
> index a7bd051..e69fdfd
Reviewed-by: Connor Abbott
On Wed, Jan 21, 2015 at 11:23 PM, Matt Turner wrote:
> Fixes a couple of warnings in the process.
> ---
> src/glsl/nir/glsl_to_nir.cpp | 57
> --
> src/glsl/nir/nir.h | 3 +-
On Tue, Dec 18, 2018 at 11:35 AM Samuel Iglesias Gonsálvez <
sigles...@igalia.com> wrote:
> v2:
> - Refactor conditions and shared function (Connor)
> - Move code to nir_eval_const_opcode() (Connor)
> - Don't flush to zero on fquantize2f16
> From Vulkan spec, VK_KHR_shader_float_controls section
I think maybe it would be better if we put all this in nir_opcodes.py
instead. Again, I'd like to make sure that for every opcode, what it does
is right next to the definition instead of buried in
nir_constant_expressions.py.
Also, I don't see anywhere in the series where you handle different
roun
On Wed, Feb 6, 2019 at 11:46 AM Samuel Iglesias Gonsálvez <
sigles...@igalia.com> wrote:
> According to Vulkan spec, the new execution modes affect only
> correctly rounded SPIR-V instructions, which includes fadd,
> fsub and fmul.
>
> Signed-off-by: Samuel Iglesias Gonsálvez
> ---
> src/compile
Reviewed-by: Connor Abbott
I'm a bit surprised it's that slow... do you have any idea what's going on?
I've made flamegraphs in the past on i965 to see where most of the time is
spent.
On Wed, Feb 13, 2019 at 9:00 AM Timothy Arceri
wrote:
> Currently the radeonsi N
Reviewed-by: Connor Abbott
I agree that if we ever need to bring this back, we should just check for
both branches empty and no phis afterwards.
On Thu, Feb 14, 2019 at 2:38 AM Timothy Arceri
wrote:
> This was probably useful when it was first written, however it
> looks to be no
On Tue, Feb 19, 2019 at 5:33 PM Emil Velikov
wrote:
> On Fri, 15 Feb 2019 at 21:39, Alyssa Rosenzweig
> wrote:
> >
> > > - about 1/5 of the patch seems to be white space changes
> >
> > ...Oops. Any tips for avoiding this type of diff churn in the future? I
> > suppose it's not inherently harmfu
On Wed, Feb 20, 2019 at 4:29 AM Marek Olšák wrote:
> On Tue, Feb 19, 2019 at 7:57 PM Rob Clark wrote:
>
>> On Tue, Feb 19, 2019 at 6:49 PM Marek Olšák wrote:
>> >
>> > st_link_shader takes 55% of CPU time with NIR, and 9% with TGSI.
>> >
>> > nir_validate_shader 49%
>> >
>> > nir_validate_shade
On Thu, Feb 21, 2019 at 3:01 PM Rob Clark wrote:
> On Thu, Feb 21, 2019 at 1:19 AM Alyssa Rosenzweig
> wrote:
> >
> > For reasons that are still unclear (speculation included in the comment
> > added in this patch), the tiler? metadata has a fast path that we were
> > not enabling; there looks t
On Thu, Feb 21, 2019 at 5:07 PM Alyssa Rosenzweig
wrote:
> > Yes, there is a buffer for holding the results of the tiler. The way it
> > works is that the userspace driver allocates a very large buffer with a
> > "grow on page fault" bit, so the kernel will allocate more memory as the
> > tiler a
On Fri, Mar 15, 2019 at 3:46 PM Ilia Mirkin wrote:
>
> On Fri, Mar 15, 2019 at 10:29 AM Alyssa Rosenzweig
wrote:
> >
> > > This is needed if you can only handle point sprites on certain
> > > varyings but not others. This is the case for nv30- and nvc0-based
> > > GPUs, which is why the cap was i
On Fri, Mar 15, 2019 at 5:00 PM Ilia Mirkin wrote:
>
> On Fri, Mar 15, 2019 at 11:52 AM Connor Abbott
wrote:
> >
> > On Fri, Mar 15, 2019 at 3:46 PM Ilia Mirkin
wrote:
> > >
> > > On Fri, Mar 15, 2019 at 10:29 AM Alyssa Rosenzweig <
aly...@rosenzweig.io&g
On Thu, Sep 3, 2015 at 2:32 PM, Kenneth Graunke wrote:
> Prevents regressions in ~128 tests when fixing unlink_block_successors
> in the next commit.
>
> XXX: Zero thought has been put into whether this is the right solution
I'm confused as to how this can happen. The fake link is only for the
si
cks(block, block->successors[0]);
> if (block->successors[1] != NULL)
>unlink_blocks(block, block->successors[1]);
> + if (block->successors[0] != NULL)
> + unlink_blocks(block, block->successors[0]);
> }
>
> static void
> --
> 2.5.0
>
Patches
On Sat, Sep 5, 2015 at 2:31 AM, Kenneth Graunke wrote:
> On Friday, September 04, 2015 11:56:29 AM Connor Abbott wrote:
>> On Thu, Sep 3, 2015 at 2:32 PM, Kenneth Graunke
>> wrote:
>> > Prevents regressions in ~128 tests when fixing unlink_block_successors
>> >
On Wed, Sep 9, 2015 at 1:37 AM, Kenneth Graunke wrote:
> On Saturday, September 05, 2015 11:10:58 AM Connor Abbott wrote:
>> On Sat, Sep 5, 2015 at 2:31 AM, Kenneth Graunke
>> wrote:
>> > On Friday, September 04, 2015 11:56:29 AM Connor Abbott wrote:
>> >> I&
On Thu, Sep 10, 2015 at 10:54 PM, Rob Clark wrote:
>
> On Sep 10, 2015 7:39 PM, "Jason Ekstrand" wrote:
>>
>> On Thu, Sep 10, 2015 at 2:39 PM, Rob Clark wrote:
>> > From: Rob Clark
>> >
>> > So this is basically working as a lowering pass for handling user-clip-
>> > planes, and frag-shader emu
On Fri, Sep 11, 2015 at 11:52 AM, Jason Ekstrand wrote:
> Fortunately, nir_constant_expr already auto-splats if "dst" never shows up
> in the constant expression field so we don't need to do anything there.
>
> Cc: Connor Abbott
> ---
> src/glsl/nir/nir.h
On Sun, Sep 13, 2015 at 11:51 AM, Rob Clark wrote:
> From: Rob Clark
>
> Signed-off-by: Rob Clark
> ---
> src/glsl/nir/nir_print.c | 73
> ++--
> 1 file changed, 59 insertions(+), 14 deletions(-)
>
> diff --git a/src/glsl/nir/nir_print.c b/src/glsl/n
On Tue, Sep 15, 2015 at 7:33 PM, Rob Clark wrote:
> From: Rob Clark
>
> Rename print_var_state to print_state, and stuff FILE ptr into the state
> object. This avoids passing around an extra parameter everywhere.
>
> Signed-off-by: Rob Clark
> ---
> src/glsl/nir/nir_print.c | 95
> +++
On Tue, Sep 15, 2015 at 7:33 PM, Rob Clark wrote:
> From: Rob Clark
>
> v2: split out moving of FILE *fp into state structure into it's own
> (more complete patch) to reduce the noise in this one
>
> Signed-off-by: Rob Clark
Rev
On Tue, Sep 15, 2015 at 9:42 PM, Rob Clark wrote:
> On Tue, Sep 15, 2015 at 8:48 PM, Connor Abbott wrote:
>> On Tue, Sep 15, 2015 at 7:33 PM, Rob Clark wrote:
>>> From: Rob Clark
>>>
>>> Rename print_var_state to print_state, and stuff FILE ptr into the st
instead of
> FILE ptr, and convert nir_print_instr() to use state as well
>
> Signed-off-by: Rob Clark
Reviewed-by: Connor Abbott
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
Except for the thing Ilia pointed out,
Reviewed-by: Connor Abbott
On Fri, Sep 18, 2015 at 1:24 PM, Rob Clark wrote:
> From: Rob Clark
>
> Signed-off-by: Rob Clark
> ---
> src/glsl/nir/nir_builder.h| 18 ++
> src/glsl/nir/nir_lower_load
cf_node_as_block(last);
> + /* insert fake link */
I was a little terse here with this comment, and based on how long it
took to actually fix this, there seems to be a lot going on that one
could easily miss from reading the code. The commit message has a
pretty good summary of
> Signed-off-by: Kenneth Graunke
Reviewed-by: Connor Abbott
> ---
> src/glsl/nir/nir_control_flow.c | 3 +++
> src/glsl/nir/nir_opt_dead_cf.c | 7 ---
> 2 files changed, 7 insertions(+), 3 deletions(-)
>
> New patch!
>
> diff --git a/src/glsl/nir/nir_control_flow.
On Tue, Sep 22, 2015 at 11:01 PM, Kenneth Graunke wrote:
> Signed-off-by: Kenneth Graunke
> ---
> src/glsl/Makefile.am | 14 +++
> src/glsl/nir/tests/control_flow_tests.cpp | 155
> ++
> 2 files changed, 169 insertions(+)
> create mode 100644 s
ink. In the above example, removing the break would cause block_1 to
> have itself as a successor (as it becomes an infinite loop), so adding
> the fake link won't cause a duplicate successor.
>
> v2: Add comments (requested by Connor Abbott) and fix commit message.
>
> Signe
On Tue, Aug 18, 2015 at 1:47 PM, Jason Ekstrand wrote:
> On Fri, May 22, 2015 at 11:24 AM, Connor Abbott wrote:
>> This can be used for both CSE and value numbering.
>>
>> Signed-off-by: Connor Abbott
>> ---
>> src/glsl/Makefile.sources | 2 +
>&g
On Tue, Aug 18, 2015 at 1:29 AM, Jason Ekstrand wrote:
>
> On May 22, 2015 11:25 AM, "Connor Abbott" wrote:
>>
>> We'll want to use these outside of nir_opt_cse.c. Rather than adding yet
>> another thing to the mess that is nir.c, create a new file
On Mon, Sep 28, 2015 at 4:47 AM, Iago Toral Quiroga wrote:
> NIR is typeless so this is the only way to keep track of the
> type to select the proper atomic to use.
> ---
> I decided to squash the i965 changes in because otherwise we would break
> the build between the nir and i965 patches. Let me
on texture support.
Signed-off-by: Connor Abbott
---
src/glsl/Makefile.sources| 1 +
src/glsl/nir/nir.c | 26 --
src/glsl/nir/nir_instr_set.c | 206 +++
src/glsl/nir/nir_instr_set.h | 29 ++
src/glsl/nir/nir_opt_cse
f the dominance tree so that the only instructions in the set
are the instructions that dominate the current block.
No piglit regressions. No shader-db changes.
XXX perf numbers
v2:
- rebase on start_block removal
- remove useless state struct
- change commit message
Signed-off-by: Connor A
v2: rebase, don't constify nir_srcs_equal() as it's pass-by-value
anyways
Signed-off-by: Connor Abbott
---
src/glsl/nir/nir_instr_set.c | 6 +++---
src/glsl/nir/nir_instr_set.h | 2 +-
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/glsl/nir/nir_instr_set.c b/sr
This was previously tied to CSE, since it would only work for
instructions where nir_can_cse() (now instr_can_rewrite()) returned
true. Now that CSE uses the instruction set abstraction which only uses
this internally, we can make it local to nir_instr_set.c.
Signed-off-by: Connor Abbott
e the chance
that one would handle less than the other without us noticing.
The main missing thing is updated performance numbers from compiling
shader-db, but I don't expect that they'll be too different.
The series is also available at
git://people.freedesktop.org/~cwabbott0/mesa nir-cse-
Signed-off-by: Connor Abbott
---
src/glsl/nir/nir.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
index 4f45770..58a86f7 100644
--- a/src/glsl/nir/nir.h
+++ b/src/glsl/nir/nir.h
@@ -722,7 +722,7 @@ nir_alu_instr_channel_used
since it's really more
about whether its uses can be rewritten, and it's implicitly used by
nir_instrs_equal() as well.
- Rename nir_instr_set_add() to nir_instr_set_add_or_rewrite() (Jason).
- Make the HASH() macro less magical (Topi).
- Rewrite the commit message.
Signed-off-by: Connor A
tions that use an index, so we can turn the check into
an assert. This also means that now nir_instrs_equal(instr, instr) will
always return true unless it assert-fails.
Signed-off-by: Connor Abbott
---
src/glsl/nir/nir_instr_set.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --
(CC'ing the list)
On Wed, Sep 30, 2015 at 4:37 PM, Michael Schellenberger
wrote:
> Hi,
>
> Am 30.09.2015 um 17:11 schrieb Connor Abbott:
>> Right now nir_instrs_equal() is tied pretty tightly to CSE, but we're
>> going to introduce the idea of an instruction se
On Wed, Sep 30, 2015 at 12:35 PM, Jason Ekstrand wrote:
> On Wed, Sep 30, 2015 at 8:11 AM, Connor Abbott wrote:
>> This will replace direct usage of nir_instrs_equal() in the CSE pass,
>> which reduces an O(n^2) algorithm with an effectively O(n) one. It'll
>> also be u
dead, and this patch will cause the loop to be deleted entirely.
No piglit regressions.
Shader-db results on bdw:
instructions in affected programs: 5824 -> 5664 (-2.75%)
helped:32
HURT: 0
Signed-off-by: Connor Abbott
---
S
On Fri, Oct 2, 2015 at 12:56 PM, Matt Turner wrote:
> On Fri, Oct 2, 2015 at 9:28 AM, Connor Abbott wrote:
>> Some loops may have phi nodes that look like:
>>
>> foo = ...
>> loop {
>> bar = phi(foo, bar)
>> ...
>> }
>>
>> in whic
On Fri, Oct 2, 2015 at 1:26 PM, Connor Abbott wrote:
> On Fri, Oct 2, 2015 at 12:56 PM, Matt Turner wrote:
>> On Fri, Oct 2, 2015 at 9:28 AM, Connor Abbott wrote:
>>> Some loops may have phi nodes that look like:
>>>
>>> foo = ...
&
into account, and it can be more
aggressive than the pre-RA scheduler since it doesn't have to worry
about register pressure at all.
XXX perf data
Signed-off-by: Connor Abbott
---
src/mesa/drivers/dri/i965/brw_fs.cpp | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git
time. The old code had it backwards, and so we
would underestimate the total time whenever we thought there would be a
pipeline stall by up to the issue time of the instruction.
Signed-off-by: Connor Abbott
---
src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp | 15 ---
1 file ch
The heuristic we're using is rather lame, since it assumes everything is
non-uniform and loops execute 50 times, but it should be enough for
measuring improvements in the scheduler that don't result in a change in
the number of instructions.
Signed-off-by: Connor Abbott
---
src/mesa/d
ned-off-by: Connor Abbott
---
src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp
b/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp
index 76d58e2..1652261 10
n affected programs: 11338636 -> 11276736 (-0.55%)
helped: 876
HURT: 873
LOST: 8
GAINED: 0
Signed-off-by: Connor Abbott
---
The results are a wash, but this is needed for a lot of the more
experimental things I want to do. I can drop this if there are any
objections.
ver
each measurement and report helped/hurt before reporting the gained/lost
programs.
Signed-off-by: Connor Abbott
---
report.py | 140 ++
1 file changed, 67 insertions(+), 73 deletions(-)
diff --git a/report.py b/report.py
index 4c06714.
ranch with some even-more-experimental patches,
which range from "this is a good idea but it hurts some stuff" to "I'm not
sure about this at all." I have some ideas about how to improve them so that
they don't regress as much, but I want to get this series out of the way
We'll need this for the scheduler too, since it wants to know when the
live ranges of payload registers end in order to model them in our
register pressure calculations.
Signed-off-by: Connor Abbott
---
src/mesa/drivers/dri/i965/brw_fs.h| 2 +
src/mesa/drivers/dri
Signed-off-by: Connor Abbott
---
report.py | 7 ++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/report.py b/report.py
index bc3a640..f9445dd 100755
--- a/report.py
+++ b/report.py
@@ -48,14 +48,19 @@ def get_result_string(p, b, a):
p = p + ' '
return
On Fri, Oct 2, 2015 at 5:37 PM, Connor Abbott wrote:
> The heuristic we're using is rather lame, since it assumes everything is
> non-uniform and loops execute 50 times, but it should be enough for
> measuring improvements in the scheduler that don't result in a change
On Mon, Oct 5, 2015 at 11:36 AM, Emil Velikov wrote:
> Hi all,
>
> I am looking at ARB_shader_clock with i965 in mind.
>
> So far I've got the most of the infra/plumbing, and a fancy a new intrinsic :)
>
> On the hardware side, I was thinking about using the Observability
> Architecture (OA) count
On Mon, Oct 5, 2015 at 1:06 PM, Emil Velikov wrote:
> On 5 October 2015 at 17:11, Connor Abbott wrote:
>> On Mon, Oct 5, 2015 at 11:36 AM, Emil Velikov
>> wrote:
>>> Hi all,
>>>
>>> I am looking at ARB_shader_clock with i965 in mind.
>>>
>
On Wed, Oct 7, 2015 at 7:59 AM, Emil Velikov wrote:
> On 7 October 2015 at 12:27, Boyan Ding wrote:
>> 2015-10-07 18:11 GMT+08:00 Emil Velikov :
>>> We will reuse it in later commits, to check if other headers are
>>> included from within nir.h.
>>>
>>> Signed-off-by: Emil Velikov
>>> ---
>>> s
On Wed, Oct 7, 2015 at 7:51 AM, Emil Velikov wrote:
> XXX: commit message, comment in nir_intrinsics.h
>
> Signed-off-by: Emil Velikov
> ---
> src/glsl/nir/glsl_to_nir.cpp | 6 ++
> src/glsl/nir/nir_intrinsics.h | 2 ++
> 2 files changed, 8 insertions(+)
>
> diff --git a/src/glsl/nir/glsl_t
On Wed, Oct 7, 2015 at 4:48 PM, Emil Velikov wrote:
> On 7 October 2015 at 18:04, Connor Abbott wrote:
>> On Wed, Oct 7, 2015 at 7:51 AM, Emil Velikov
>> wrote:
>>> XXX: commit message, comment in nir_intrinsics.h
>>>
>>> Signed-off-by: Emil Velikov
On Wed, Oct 7, 2015 at 7:51 AM, Emil Velikov wrote:
> Signed-off-by: Emil Velikov
> ---
> src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 9 +
> src/mesa/drivers/dri/i965/brw_vec4_nir.cpp | 10 ++
> 2 files changed, 19 insertions(+)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_fs
On Thu, Oct 8, 2015 at 6:54 AM, Emil Velikov wrote:
> On 7 October 2015 at 23:50, Connor Abbott wrote:
>> On Wed, Oct 7, 2015 at 4:48 PM, Emil Velikov
>> wrote:
>>> On 7 October 2015 at 18:04, Connor Abbott wrote:
>>>> On Wed, Oct 7, 2015 at 7:51 AM,
On Thu, Oct 8, 2015 at 8:54 AM, Emil Velikov wrote:
> On 8 October 2015 at 11:54, Emil Velikov wrote:
>> On 7 October 2015 at 23:50, Connor Abbott wrote:
>>> On Wed, Oct 7, 2015 at 4:48 PM, Emil Velikov
>>> wrote:
>>>> On 7 October 2015 at 18:04, Co
On Thu, Oct 8, 2015 at 10:09 AM, Emil Velikov wrote:
> On 8 October 2015 at 14:20, Connor Abbott wrote:
>> On Thu, Oct 8, 2015 at 6:54 AM, Emil Velikov
>> wrote:
>>> On 7 October 2015 at 23:50, Connor Abbott wrote:
>>>> On Wed, Oct 7, 2015 at 4:48 PM, E
rces, use a VLA, store pointers to the sources, and
compare the predecessor pointer directly (Jason).
Signed-off-by: Connor Abbott
---
Jason, does this look good to you? Still no piglit regressions -- I'm
working on getting perf numbers.
src/glsl/nir/n
On Mon, Dec 28, 2015 at 1:35 PM, Rob Clark wrote:
> On Mon, Dec 28, 2015 at 12:37 PM, Connor Abbott wrote:
>> On Mon, Dec 28, 2015 at 10:13 AM, Rob Clark wrote:
>>> On Tue, Dec 22, 2015 at 10:11 PM, Connor Abbott wrote:
>>>> On Tue, Dec 22, 2015 at 9:55 PM, Rob C
t; + return false;
> +
> + if (last_instr->type != nir_instr_type_jump)
> + return false;
> +
> + nir_jump_instr *jump = nir_instr_as_jump(last_instr);
> + if (jump->type != nir_jump_return)
> + return false;
> +
> + nir_builder *b = &state->bu
On Mon, Dec 28, 2015 at 3:31 PM, Rob Clark wrote:
> On Mon, Dec 28, 2015 at 1:58 PM, Connor Abbott wrote:
>> On Mon, Dec 28, 2015 at 1:35 PM, Rob Clark wrote:
>>> On Mon, Dec 28, 2015 at 12:37 PM, Connor Abbott wrote:
>>>> On Mon, Dec 28, 2015 at 10:13 AM, Rob C
gt; >
>>> >
>>> > On Mon, Dec 28, 2015 at 9:37 AM, Connor Abbott
>>> > wrote:
>>> >>
>>> >> On Mon, Dec 28, 2015 at 10:13 AM, Rob Clark
>>> >> wrote:
>>> >> > On Tue, Dec 22, 2015 at 10:11 PM,
On Mon, Dec 28, 2015 at 4:21 PM, Jason Ekstrand wrote:
>>
>>
>> but I hope you can get loops working soon so you can test this (and
>> hey, loops are useful for other things too :P).
>
>
> Yeah, I'm starting on adding a real CFG to spirv_to_nir today. It's the
> task that I've had on my todo list
On Mon, Dec 28, 2015 at 5:02 PM, Jason Ekstrand wrote:
>
>
> On Mon, Dec 28, 2015 at 1:28 PM, Connor Abbott wrote:
>>
>> On Mon, Dec 28, 2015 at 4:21 PM, Jason Ekstrand
>> wrote:
>> >>
>> >>
>> >> but I hope you can get loops wor
well
> as tgsi_to_nir and prog_to_nir.
Reviewed-by: Connor Abbott
> ---
> src/gallium/auxiliary/nir/tgsi_to_nir.c | 12
> src/glsl/nir/nir_builder.h| 11 +++
> src/glsl/nir/tests/control_flow_tests.cpp | 26 ++
> src
On Thu, Dec 31, 2015 at 10:16 AM, Rob Clark wrote:
> On Tue, Dec 29, 2015 at 10:32 AM, Rob Clark wrote:
> If you do this, you'll be back to always needing a mutable copy. Most
> lowering and optimization passes die the moment they see a register.
> You'll
> either have to go fi
On Mon, Jan 11, 2016 at 11:44 PM, Ilia Mirkin wrote:
> Actually this is still a little bogus... We'd have to ensure that
> *all* of the defs are used in this bb or later. I'll try to come up
> with something that still maintains some of the potential benefit but
> is correct.
I don't know about t
On Mon, Jan 11, 2016 at 9:21 AM, Jose Fonseca wrote:
> FWIW, I updated SCons to build NIR, both with GCC and MSVC:
>
> http://cgit.freedesktop.org/~jrfonseca/mesa/log/?h=scons-nir
>
> It was actually simpler than I anticipated.
>
> But I hit a wall -- there's actually no way to get NIR used with
On Tue, Jan 12, 2016 at 12:54 AM, Matt Turner wrote:
> On Mon, Jan 11, 2016 at 6:58 PM, Jason Ekstrand wrote:
>>
>> On Jan 11, 2016 2:47 PM, "Matt Turner" wrote:
>>>
>>> The OpenGL specifications for bitfieldInsert() and bitfieldExtract() says:
>>>
>>>The result will be undefined if or is
On Mon, Jan 11, 2016 at 7:20 PM, Ilia Mirkin wrote:
> On Mon, Jan 11, 2016 at 7:12 PM, Matt Turner wrote:
>> From: Kenneth Graunke
>>
>> We would like to be able to combine
>>
>>result.x = bitfieldInsert(src0.x, src1.x, src2.x, src3.x);
>>result.y = bitfieldInsert(src0.y, src1.y, src2.y,
On Mon, Jan 11, 2016 at 6:52 PM, Ian Romanick wrote:
> On 01/11/2016 02:48 PM, Matt Turner wrote:
>> NIR's bfm, like Intel/AMD's hardware instructions and GLSL IR's
>> ir_binop_bfm takes as src0 and as src1.
>
> All the questions...
>
> Is the ordering of the operands documented anywhere? I was
Both are
Reviewed-by: Connor Abbott
On Wed, Jan 13, 2016 at 2:25 PM, Matt Turner wrote:
> The OpenGL specifications for bitfieldInsert() says:
>
>The result will be undefined if or is negative, or if
>the sum of and is greater than the number of bits
>used to st
A few things:
- How much work would it take to actually deduplicate this with the
other vec4 type_size implementations out there? I'm pretty sure no one
would object to it, and if we just leave this as a TODO it probably
won't actually get done.
- Having to include nir_types.h in something that ha
On Sun, Jan 31, 2016 at 3:16 PM, Rob Clark wrote:
> From: Rob Clark
>
> Signed-off-by: Rob Clark
> ---
> src/compiler/Makefile.sources| 1 +
> src/compiler/nir/nir.h | 12 +
> src/compiler/nir/nir_lower_wpos_ytransform.c | 317
>
This doesn't really seem to be following NIR's convention/style of
having separate passes/files for each thing that needs to be lowered.
This might've made sense for TGSI, where there was a lot of
boilerplate that each thing needed since TGSI isn't designed for doing
transforms on it, but with NIR
501 - 600 of 1264 matches
Mail list logo
