bit today. Maybe we should just go ahead and
> implement that? For unary ALU operations, phi distribution is a clear win
> and mov is definitely unary.
Yeah, if we implement that then we can just get rid of the whole thing.
>
> On Fri, Sep 2, 2016 at 4:28 PM, Connor Abbott wro
Reviewed-by: Connor Abbott
On Sat, Sep 3, 2016 at 2:59 PM, Jason Ekstrand wrote:
> In aad4f1550, we removed the concept of "fake" edges from NIR. Now, if you
> have a block at the end of an infinite loop it really has no predecessors.
> This updates the unit tests to match
Since demoted inputs become global variables, wouldn't it be more
efficient to simply turn it into a straight assignment? Then the
backend wouldn't have to initialize it to a constant (since it would
be undefined).
On Fri, Sep 9, 2016 at 10:06 AM, Marek Olšák wrote:
> From: Marek Olšák
>
> This
On Tue, Oct 25, 2016 at 8:59 PM, Ian Romanick wrote:
> From: Ian Romanick
>
> The shift operations are a little weird. NIR expects both operands to
> have the same size, but the shift count operand is always 32-bits in
> GLSL. Upconvert to make the rest of NIR happy, and we'll assume the
> driv
On Tue, Oct 25, 2016 at 8:59 PM, Ian Romanick wrote:
> From: Ian Romanick
>
> Signed-off-by: Ian Romanick
> ---
> src/compiler/glsl/glsl_to_nir.cpp | 30 +++
> src/compiler/nir/nir_lower_alu_to_scalar.c | 1 +
> src/compiler/nir/nir_opcodes.py| 33
On Tue, Oct 25, 2016 at 8:59 PM, Ian Romanick wrote:
> Whew! Here it is! There are four main groups to this patch series.
> All of this is available at:
>
> https://cgit.freedesktop.org/~idr/mesa/log/?h=arb_gpu_shader_int64
>
> There are some gaps in our current piglit tests. For example, I
On Wed, Oct 26, 2016 at 10:52 PM, Ian Romanick wrote:
> On 10/26/2016 03:04 PM, Connor Abbott wrote:
>> On Tue, Oct 25, 2016 at 8:59 PM, Ian Romanick wrote:
>>> From: Ian Romanick
>>>
>>> The shift operations are a little weird. NIR expects both operand
imilar for the others. If that doesn't work, I could take a look
if you push what you tried to fdo.
>
> Signed-off-by: Ian Romanick
> Cc: Connor Abbott
> Cc: Jason Ekstrand
> ---
> src/compiler/nir/nir_builder.h | 4 +---
> src/compiler/nir/nir_validate.c | 7 ---
&g
2:25 assert src_class == src_type_bits
> 03:12:25 AssertionError
It's pretty great that Jason wrote this validation pass, since
otherwise this would've been a rare and hard-to-diagnose failure
condition at runtime. Anyways, much better! This patch is
Reviewed-by: Connor Abbot
I'm a little worried about this since it seems like the
load/store_scratch intrinsics are basically doing the same thing as
registers were originally intended to do. Either we should use the
existing register lowering, and make it conditional on the size like
you've done here, or we should just gut
On Mon, Dec 5, 2016 at 3:22 PM, Matt Turner wrote:
> On 12/05, Matt Turner wrote:
>>
>> On 11/28, Ian Romanick wrote:
>>>
>>> From: Ian Romanick
>>>Patches 42 through 50 enable the extension on BDW+.
>>
>>
>> 42-48 are
>>
>> Reviewed-by: Matt Turner
>>
>> I don't understand the 64-bit CMP is
On Mon, Dec 5, 2016 at 5:09 PM, Connor Abbott wrote:
> On Mon, Dec 5, 2016 at 3:22 PM, Matt Turner wrote:
>> On 12/05, Matt Turner wrote:
>>>
>>> On 11/28, Ian Romanick wrote:
>>>>
>>>> From: Ian Romanick
>>>>Patches 42
On Mon, Dec 5, 2016 at 5:48 PM, Matt Turner wrote:
> On Mon, Dec 5, 2016 at 2:20 PM, Connor Abbott wrote:
>> On Mon, Dec 5, 2016 at 5:09 PM, Connor Abbott wrote:
>>> On Mon, Dec 5, 2016 at 3:22 PM, Matt Turner wrote:
>>>> On 12/05, Matt Turner wrote:
>>&g
On Thu, Jul 7, 2016 at 12:57 PM, Erik Faye-Lund wrote:
> On Thu, Jul 7, 2016 at 2:12 AM, Eric Anholt wrote:
>> This was appearing in vc4 VS/CS in mupen64, due to vertex attrib lowering
>> producing some constants that were getting compared.
>>
>> total instructions in shared programs: 112276 -> 1
On Sat, Jul 23, 2016 at 3:32 PM, tournier.elie wrote:
> Hello,
>
> I hope you are well and that you enjoy these beautiful summer days.
>
> You can found a new post on my blog : https://hopetech.github.io/ .
> During these few days, I implemented eq_fp64, le_fp64, lt_fp64, mul_fp64,
> add_fp64 and
On Thu, Jul 28, 2016 at 2:25 AM, Iago Toral wrote:
> On Thu, 2016-07-28 at 15:49 +1000, Timothy Arceri wrote:
>> On Tue, 2016-07-19 at 12:40 +0200, Iago Toral Quiroga wrote:
>> >
>> > From: Connor Abbott
>> >
>> > ---
>> > src/mesa/drivers/d
On Thu, Jul 28, 2016 at 7:58 PM, Roland Scheidegger wrote:
> Am 29.07.2016 um 00:35 schrieb Matt Turner:
>> ---
>> src/gallium/auxiliary/Makefile.sources | 1 +
>> src/gallium/auxiliary/util/u_bitcast.h | 57
>> ++
>> 2 files changed, 58 insertions(+)
>> create
ieve "real"
softfloat libraries use some kind of integer-based higher-precision
thing, but you don't need an exact result so you don't need to be that
fancy (again, a good reason why our implementation is probably better
to base off of!).
Connor
>
>
> Elie
>
>
>
On Tue, Jun 10, 2014 at 3:34 PM, Matt Turner wrote:
> On Wed, May 28, 2014 at 4:44 PM, Connor Abbott wrote:
>> On Tue, May 27, 2014 at 9:47 PM, Matt Turner wrote:
>>> Here's a respin of my load_payload series from mid-April with some
>>> feedback from K
On Jun 26, 2014 5:30 PM, "Jason Ekstrand" wrote:
>
> Right now, the Intel driver is the only driver in mesa that implements
this extension. Is anyone using this? Is it ok if we purge it? I'm doing
some work on the texture format code and DUDV8 is an ugly special case.
> Thanks,
> --Jason Ekstra
On Sat, Jun 28, 2014 at 9:34 PM, Kenneth Graunke wrote:
> Apparently INTEL_DEBUG=fs has crashed on Broadwell for anything using
> ARB_fragment_program since commit 9cee3ff5. We need to NULL-check the
> right field.
>
> Signed-off-by: Kenneth Graunke
> Cc: "10.2"
> ---
> src/mesa/drivers/dri/i9
This series adds a couple things I need to exec_list for my work, and
does some cleanups made possible. Only compile tested on i965.
Connor Abbott (3):
exec_list: add a prepend function
exec_list: add a function to count the size of a list
exec_list: make various places use the new get_size
This complements the existing append function. It's implemented in a
rather simple way right now; it could be changed if performance is a
concern.
Signed-off-by: Connor Abbott
---
src/glsl/list.h | 20 +++-
1 file changed, 19 insertions(+), 1 deletion(-)
diff --git a/src
Signed-off-by: Connor Abbott
---
src/glsl/list.h | 20
1 file changed, 20 insertions(+)
diff --git a/src/glsl/list.h b/src/glsl/list.h
index ca6ee9d..68ab3fd 100644
--- a/src/glsl/list.h
+++ b/src/glsl/list.h
@@ -324,6 +324,8 @@ struct exec_list {
const exec_node
Instead of hand-rolling it.
Signed-off-by: Connor Abbott
---
src/glsl/ast_to_hir.cpp | 4 +---
src/glsl/ir_reader.cpp| 7 +++
src/glsl/opt_function_inlining.cpp| 7 ++-
src/mesa/drivers/dri/i965/brw_fs.cpp
On Wednesday, July 09, 2014 03:56:41 PM Ian Romanick wrote:
> On 07/08/2014 12:20 PM, Connor Abbott wrote:
> > Signed-off-by: Connor Abbott
> > ---
> >
> > src/glsl/list.h | 20
> > 1 file changed, 20 insertions(+)
> >
> > d
optimization, in that if deref->array_index
were a constant but it got wrapped in an i2u expressoin then we
would've gone and generated an add and a mul instruction when that
isn't really necessary.
In any case, this patch is
Reviewed-by: Connor Abbott
>
7;s useful for codepaths that want to distinguish between
intrinsics and non-intrinsics without using strcmp.
Signed-off-by: Connor Abbott
---
src/glsl/link_functions.cpp | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/glsl/link_functions.cpp b/src/glsl/link_functions.cpp
index 2ce960
.
No difference in shader-db run times, but I'm keeping this in because
the q total that it calculates will also be used in the next commit.
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 35 +++
1 file changed, 27 insertions(+), 8 deletion
s the summary:
total instructions in shared programs: 4545447 -> 4545411 (-0.00%)
instructions in affected programs: 1353 -> 1317 (-2.66%)
GAINED:124
LOST: 6
Connor Abbott (3):
ra: cleanup the public API
ra: make the p,
we're at it rename ra_allocate_no_spills() to
ra_allocate() since there's no equivalent "with spills," because the
backend is supposed to handle spilling.
Signed-off-by: Connor Abbott
---
src/gallium/drivers/r300/compiler/radeon_pair_regalloc.c | 2 +-
src/m
4545411 (-0.00%)
instructions in affected programs: 1353 -> 1317 (-2.66%)
GAINED:124
LOST: 6
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 111 +--
1 file changed, 29 inserti
On Tue, Jul 29, 2014 at 5:53 PM, Connor Abbott wrote:
> This patch series contains some improvements to the register allocator
> used by the i965 fs and vec4 backends and r300g. The most important
> patch, and the only one with an intended functional change, is the last
> one. Fu
Signed-off-by: Connor Abbott
---
src/glsl/list.h | 6 ++
1 file changed, 6 insertions(+)
diff --git a/src/glsl/list.h b/src/glsl/list.h
index 3ee6cda..c402f48 100644
--- a/src/glsl/list.h
+++ b/src/glsl/list.h
@@ -644,6 +644,12 @@ inline void exec_node::insert_before(exec_list *before
we're at it rename ra_allocate_no_spills() to
ra_allocate() since there's no equivalent "with spills," because the
backend is supposed to handle spilling.
Reviewed-by: Eric Anholt
Signed-off-by: Connor Abbott
---
src/gallium/drivers/r300/compiler/radeon_pair_regalloc.c | 2
.
No difference in shader-db run times, but I'm keeping this in because
the q total that it calculates will also be used in the next few commits.
Reviewed-by: Eric Anholt
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 33 ++---
1 file c
124
LOST: 6
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 57 ++--
1 file changed, 22 insertions(+), 35 deletions(-)
diff --git a/src/mesa/program/register_allocate.c
b/src/mesa/program/register_allocate.c
0
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 52 +---
1 file changed, 6 insertions(+), 46 deletions(-)
diff --git a/src/mesa/program/register_allocate.c
b/src/mesa/program/register_allocate.c
index 73d3b1a..f2f5910 100644
--- a/src/mes
ix trailing whitespace, split the last patch in two
Connor Abbott (4):
ra: cleanup the public API
ra: make the p, q test more efficient
ra: consider all spillable nodes for spilling
ra: optimistically color only one node at a time
.../drivers/r300/compiler/radeon_pair_regalloc.c | 2 +
On Wed, Jul 30, 2014 at 10:50 PM, Eric Anholt wrote:
> Connor Abbott writes:
>
>> Before, we would only consider nodes for spilling if they were
>> optimistically pushed onto the stack. But the logic for this was
>> complicated, and duplicated some things; also, we migh
On Thu, Jul 31, 2014 at 1:05 PM, Eric Anholt wrote:
> Connor Abbott writes:
>> That's not necessarily true - you could want to spill a trivially
>> colored register that interferes with a non trivially colored
>> register, especially if the spill cost of the non trivia
s in affected programs: 1353 -> 1307 (-3.40%)
GAINED:124
LOST: 6
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 57 ++--
1 file changed, 22 insertions(+), 35 deletions(
6
v2: fix trailing whitespace, split the last patch in two
v3: rewrite patch 3, removing regressions in number of instructions
Connor Abbott (4):
ra: cleanup the public API
ra: make the p, q test more efficient
ra: don't consider nodes for spilling we don't need to
ra: optimistically c
we're at it rename ra_allocate_no_spills() to
ra_allocate() since there's no equivalent "with spills," because the
backend is supposed to handle spilling.
Reviewed-by: Eric Anholt
Signed-off-by: Connor Abbott
---
src/gallium/drivers/r300/compiler/radeon_pair_regalloc.c | 2
7;s).
Signed-off-by: Connor Abbott
---
Eric, does this patch seem OK to you? There are some comments indicating
the deleted code is necessary, but based on our conversation and my
tests it doesn't seem so...
src/mesa/program/register_allocate.c | 51
1 fi
.
No difference in shader-db run times, but I'm keeping this in because
the q total that it calculates will also be used in the next few commits.
Reviewed-by: Eric Anholt
Signed-off-by: Connor Abbott
---
src/mesa/program/register_allocate.c | 33 ++---
1 file c
On Fri, Aug 1, 2014 at 11:51 AM, Eric Anholt wrote:
> Connor Abbott writes:
>
>> On Thu, Jul 31, 2014 at 1:05 PM, Eric Anholt wrote:
>>> Connor Abbott writes:
>>>> That's not necessarily true - you could want to spill a trivially
>>>> col
On Fri, Aug 1, 2014 at 12:25 PM, Eric Anholt wrote:
> Connor Abbott writes:
>
>> On Fri, Aug 1, 2014 at 11:51 AM, Eric Anholt wrote:
>>> Connor Abbott writes:
>>>
>>>> On Thu, Jul 31, 2014 at 1:05 PM, Eric Anholt wrote:
>>>>> Connor
Hi,
While looking at the GLSL IR code, I noticed that different parts of the
code use 2 different hash table implementations. ir_loop_analysis,
ir_variable_refcount, ir_clone, etc. use struct hash_table under
program/hash_table.h, whereas ir_variable_refcount and link_uniform_blocks
use struct has
I'm not sure if this is relevant to the i965 fs backend in particular, but
Briggs in his thesis [1] recommended simply ignoring the liveness of any
register read by a copy instruction immediately after said instruction when
computing the interference graph in order to solve this exact problem (see
27, 2013 at 7:38 PM, Connor Abbott wrote:
> I'm not sure if this is relevant to the i965 fs backend in particular, but
> Briggs in his thesis [1] recommended simply ignoring the liveness of any
> register read by a copy instruction immediately after said instruction when
> computi
Like swizzle_x(), swizzle_y(), etc., except the index is programmable.
---
src/glsl/ir_builder.cpp | 6 ++
src/glsl/ir_builder.h | 1 +
2 files changed, 7 insertions(+)
diff --git a/src/glsl/ir_builder.cpp b/src/glsl/ir_builder.cpp
index 7f41ed6..2229cd2 100644
--- a/src/glsl/ir_builder.cpp
--git a/src/glsl/ir_dead_branches.cpp b/src/glsl/ir_dead_branches.cpp
new file mode 100644
index 000..f86f009
--- /dev/null
+++ b/src/glsl/ir_dead_branches.cpp
@@ -0,0 +1,226 @@
+/*
+ * Copyright © 2013 Connor Abbott (con...@abbott.cx)
+ *
+ * Permission is hereby granted, free of charge, to any
\
$(GLSL_SRCDIR)/ir_reader.cpp \
$(GLSL_SRCDIR)/ir_rvalue_visitor.cpp \
diff --git a/src/glsl/ir_loop_jumps.cpp b/src/glsl/ir_loop_jumps.cpp
new file mode 100644
index 000..1386340
--- /dev/null
+++ b/src/glsl/ir_loop_jumps.cpp
@@ -0,0 +1,129 @@
+/*
+ * Copyright © 2013 Connor Abbott (con
0..c1044f6
--- /dev/null
+++ b/src/glsl/opt_to_ssa.cpp
@@ -0,0 +1,1155 @@
+/*
+ * Copyright © 2013 Connor Abbott (con...@abbott.cx)
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
We forgot to handle the case where the base type was a boolean.
---
src/glsl/ir_constant_expression.cpp | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/glsl/ir_constant_expression.cpp
b/src/glsl/ir_constant_expression.cpp
index f811fd1..9edc378 100644
--- a/src/glsl/ir_constant_express
,
bool uniform_locations_assigned,
diff --git a/src/glsl/opt_from_ssa.cpp b/src/glsl/opt_from_ssa.cpp
new file mode 100644
index 000..6071c45
--- /dev/null
+++ b/src/glsl/opt_from_ssa.cpp
@@ -0,0 +1,198 @@
+/*
+ * Copyright © 2013 Connor Abbott (con...@abbott.cx)
+ *
+ * Permission is hereby
This patch introduces all the changes to the IR that are necessary for
representing programs in the SSA form. This consists of a new variable
mode, the SSA temporary, which is guarenteed to be written to exactly
once, and classes to represent phi nodes in the IR.
In the current code, variables are
edu/courses/cse501/04wi/papers/click-pldi95.pdf
Connor Abbott (11):
glsl: fix handling of quadop_vector constant expression
glsl: add as_loop_jump() method to ir_instruction
glsl: add a foreach_list_reverse macro
glsl: add dead branch analysis
glsl: add loop jump visitor
glsl: add swi
This will let us dynamically downcast to ir_loop_jump, which will be
needed later.
---
src/glsl/ir.h | 6 ++
1 file changed, 6 insertions(+)
diff --git a/src/glsl/ir.h b/src/glsl/ir.h
index 19e8383..d1e790d 100644
--- a/src/glsl/ir.h
+++ b/src/glsl/ir.h
@@ -138,6 +138,7 @@ public:
virtual
ssa_assign() creates an SSA variable and assignment at the same
time. With this, simple sequences of SSA statements can be easily
created.
---
src/glsl/ir_builder.cpp | 14 ++
src/glsl/ir_builder.h | 1 +
2 files changed, 15 insertions(+)
diff --git a/src/glsl/ir_builder.cpp b/src/
This patch is mainly for allowing me to test these changes with piglit.
In the future, a do_ssa_optimizations() function will need to be
created and used by this code, as well as all other users of
do_common_optimizations().
---
src/glsl/glsl_parser_extras.cpp | 4
1 file changed, 4 insertion
---
src/glsl/list.h | 5 +
1 file changed, 5 insertions(+)
diff --git a/src/glsl/list.h b/src/glsl/list.h
index 694b686..2de20c3 100644
--- a/src/glsl/list.h
+++ b/src/glsl/list.h
@@ -383,6 +383,11 @@ inline void exec_node::insert_before(exec_list *before)
; (__node)->next != NULL
This way, when someone modifies create_test_cases.py and forgets to
commit their changes again, people will notice.
Signed-off-by: Connor Abbott
---
src/glsl/tests/optimization-test | 7 +++
1 file changed, 7 insertions(+)
diff --git a/src/glsl/tests/optimization-test b/src/glsl/tests
will
bring create_test_cases.py in line with them.
Signed-off-by: Connor Abbott
---
src/glsl/ir_print_visitor.cpp | 2 +-
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test | 3 +--
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test.expected| 3 +--
src/
he used to create the tests.
Fix that, so that now create_test_cases.py is synced with the generated
tests.
Signed-off-by: Connor Abbott
---
src/glsl/tests/lower_jumps/create_test_cases.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/glsl/tests/lower_jumps
They were made unneccesary by the last commit.
Signed-off-by: Connor Abbott
---
src/glsl/tests/lower_jumps/.gitignore | 2 ++
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test | 12 -
.../lower_jumps/lower_breaks_1.opt_test.expected | 4 ---
src/glsl/tests/lower_jumps
ect in the input files.
I fixed both problems, and then I removed the generated tests so that
stuff like this won't happen again.
Connor Abbott (4):
glsl: be more consistent about printing constants
glsl/tests/lower_jumps: fix generated sexpr's for loops
glsl/tests: call create_test_
he used to create the tests.
Fix that, so that now create_test_cases.py is synced with the generated
tests.
Signed-off-by: Connor Abbott
---
src/glsl/tests/lower_jumps/create_test_cases.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/glsl/tests/lower_jumps
in the input files.
This series fixes both problems, and then removes the generated tests so
that stuff like this won't happen again.
v2: actually generate the test files
Connor Abbott (4):
glsl: be more consistent about printing constants
glsl/tests/lower_jumps: fix generated sexpr
They were made unneccesary by the last commit.
Signed-off-by: Connor Abbott
---
src/glsl/tests/lower_jumps/.gitignore | 2 ++
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test | 12 -
.../lower_jumps/lower_breaks_1.opt_test.expected | 4 ---
src/glsl/tests/lower_jumps
This way, when someone modifies create_test_cases.py and forgets to
commit their changes again, people will notice.
v2: make sure we parse the right directories and check for existance the
right way.
Signed-off-by: Connor Abbott
---
src/glsl/tests/optimization-test | 8
1 file changed
will
bring create_test_cases.py in line with them.
Signed-off-by: Connor Abbott
---
src/glsl/ir_print_visitor.cpp | 2 +-
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test | 3 +--
src/glsl/tests/lower_jumps/lower_breaks_1.opt_test.expected| 3 +--
src/
On Wed, May 28, 2014 at 2:37 PM, Eric Anholt wrote:
> Here's a series I started back in January as a little experiment.
> Basically, I feel guilty for pushing GLSL IR into the driver, and wish I'd
> just fixed up Mesa IR back in the day. But, given that we're still
> feeding Mesa IR into drivers
On Tue, May 27, 2014 at 9:47 PM, Matt Turner wrote:
> Here's a respin of my load_payload series from mid-April with some
> feedback from Ken addressed and some bugs fixed.
>
> This series is available in my tree (with a few unrelated patches
> before it)
>
>git://people.freedesktop.org/~mattst
On Wed, May 28, 2014 at 6:18 PM, Kenneth Graunke wrote:
> On 05/28/2014 01:17 PM, Matt Turner wrote:
>> On Wed, May 28, 2014 at 1:09 PM, Kenneth Graunke
>> wrote:
>>> On 05/27/2014 06:23 PM, Connor Abbott wrote:
>>>> This way, when someone modifies
On Wed, May 28, 2014 at 8:50 PM, Eric Anholt wrote:
> Connor Abbott writes:
>
>> On Wed, May 28, 2014 at 2:37 PM, Eric Anholt wrote:
>>> Here's a series I started back in January as a little experiment.
>>> Basically, I feel guilty for pushing GLSL IR int
On Fri, Jan 31, 2014 at 3:34 PM, Paul Berry wrote:
> On 22 January 2014 09:16, Connor Abbott wrote:
>
>> This series enables GLSL IR support for SSA, including passes to convert
>> to and from SSA form. SSA is a form of the intermediate representation
>> of a compiler i
On Tue, Jan 28, 2014 at 2:45 PM, Paul Berry wrote:
> On 22 January 2014 09:16, Connor Abbott wrote:
>
>> diff --git a/src/glsl/ir_clone.cpp b/src/glsl/ir_clone.cpp
>> index cb732a5..7075579 100644
>> --- a/src/glsl/ir_clone.cpp
>> +++ b/src/glsl/ir_clone.cpp
&
On Tue, Jan 28, 2014 at 1:50 PM, Paul Berry wrote:
> On 22 January 2014 09:16, Connor Abbott wrote:
>
>> This patch introduces all the changes to the IR that are necessary for
>> representing programs in the SSA form. This consists of a new variable
>> mode, the
On Fri, Jan 31, 2014 at 11:55 AM, Paul Berry wrote:
> On 22 January 2014 09:16, Connor Abbott wrote:
>
>> Right now we are being basically as naive as possible, and inserting
>> more copies than necessary. It is possible to implement a more
>> sophisticated algorithm l
Hi,
So I believe that we can all agree that the tree-based representation
that GLSL IR currently uses for shaders needs to go. For the benefit
of those that didn't watch Ian Romanick's talk at FOSDEM, I'll
reiterate some of the problems with it as of now:
- All the ir_dereference chains blow up t
On Fri, Feb 7, 2014 at 3:13 PM, Ian Romanick wrote:
> On 02/06/2014 09:34 PM, Connor Abbott wrote:
>> Hi,
>>
>> So I believe that we can all agree that the tree-based representation
>> that GLSL IR currently uses for shaders needs to go. For the benefit
>&g
On Tue, Jan 28, 2014 at 1:50 PM, Paul Berry wrote:
> On 22 January 2014 09:16, Connor Abbott wrote:
>>
>> This patch introduces all the changes to the IR that are necessary for
>> representing programs in the SSA form. This consists of a new variable
>> mode,
On Wed, Mar 12, 2014 at 6:49 PM, Ian Romanick wrote:
> From: Ian Romanick
>
> Signed-off-by: Ian Romanick
> ---
> src/glsl/ir_clone.cpp | 10 +-
> src/glsl/ir_constant_expression.cpp | 195
> +++-
> 2 files changed, 109 insertions(+), 96 deletions
On Wed, Mar 12, 2014 at 6:49 PM, Ian Romanick wrote:
> From: Ian Romanick
>
> Signed-off-by: Ian Romanick
> ---
> src/glsl/ir_constant_expression.cpp | 52
> +
> 1 file changed, 36 insertions(+), 16 deletions(-)
>
> diff --git a/src/glsl/ir_constant_expressi
On Tue, Dec 9, 2014 at 12:44 AM, Roland Scheidegger wrote:
> Am 09.12.2014 um 02:31 schrieb Dave Airlie:
>> From: Dave Airlie
>>
>> Since d8da6deceadf5e48201d848b7061dad17a5b7cac where the
>> state tracker started using UCMP on cayman a number of tests
>> regressed.
>>
>> this seems to be r600g i
On Tue, Dec 16, 2014 at 1:04 AM, Jason Ekstrand wrote:
> ---
> src/glsl/nir/nir_opcodes.h| 6 +++---
> src/glsl/nir/nir_opt_copy_propagate.c | 2 +-
> 2 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/src/glsl/nir/nir_opcodes.h b/src/glsl/nir/nir_opcodes.h
> index 36a5
When Ken rebased my texture patches, he changed the titles to say
"make thing() not take an ir_variable", which I like. Can you change
the title of this one too?
On Tue, Dec 16, 2014 at 1:04 AM, Jason Ekstrand wrote:
> From: Connor Abbott
>
> ---
> src/mesa/drive
On Tue, Dec 16, 2014 at 1:04 AM, Jason Ekstrand wrote:
> From: Connor Abbott
>
> This is similar to the GLSL IR frontend, except consuming NIR. This lets
> us test NIR as part of an actual compiler.
>
> v2: Jason Ekstrand :
>Make brw_fs_nir build again
>Only use
Can we squash this into wherever we introduced these print statements,
and instead port over Matt's i965 fs debugging infrastructure so we
can do this with an envvar enable, plus all the other goodies?
On Tue, Dec 16, 2014 at 1:11 AM, Jason Ekstrand wrote:
> This is killing piglit. I'll leave th
> diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
> b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
> index 3e447b8..fd7dd43 100644
> --- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
> @@ -97,7 +97,11 @@ fs_visitor::visit(ir_var
On Tue, Dec 16, 2014 at 5:22 PM, Jason Ekstrand wrote:
>
>
> On Tue, Dec 16, 2014 at 1:28 PM, Connor Abbott wrote:
>>
>> On Tue, Dec 16, 2014 at 1:04 AM, Jason Ekstrand
>> wrote:
>> > From: Connor Abbott
>> >
>> > This is similar to the GLS
_field.prev, __field))
> +
> #endif /* LIST_CONTAINER_H */
> --
> 2.2.0
>
> ___
> mesa-dev mailing list
> mesa-dev@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev
Reviewed-by: Connor Abbott
__
On Tue, Dec 16, 2014 at 1:05 AM, Jason Ekstrand wrote:
> Some time while refactoring things to make it look nicer before pushing to
> master, I completely broke the function. This fixes it to be correct.
> Just goes to show you why you souldn't push code that has no users yet...
>
> Signed-off-by
asier to look at the end result rather than the
> patches. The branch can be found via freedesktop cgit here:
>
> http://cgit.freedesktop.org/~jekstrand/mesa/log/?h=review/nir-v1
>
> Last week, I did a presentation for some of the other Intel people to try
> and help bring them up t
On Tue, Dec 16, 2014 at 5:48 PM, Eric Anholt wrote:
> Jason Ekstrand writes:
>
>> From: Connor Abbott
>>
>> This includes all the instructions, ifs, loops, functions, etc. This is
>> similar to the information in ir.h.
>>
>> v2: Jason Ekstrand :
>>
ype_unsigned, nir_type_unsigned, nir_type_unsigned,
> nir_type_unsigned), \
>0)
> #define QUADOP_HORIZ(name, output_size, src1_size, src2_size, src3_size, \
> src4_size) \
> - OPCODE(name, 4, false, output_size, nir_type_unsigned, \
> +
Signed-off-by: Connor Abbott
---
src/glsl/ir_function_can_inline.cpp | 9 -
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/src/glsl/ir_function_can_inline.cpp
b/src/glsl/ir_function_can_inline.cpp
index 7b15d5d..3b1d15f 100644
--- a/src/glsl/ir_function_can_inline.cpp
Oh, I forgot... patch 20 should probably get cherry-picked as well.
On Tue, Dec 16, 2014 at 6:01 PM, Jason Ekstrand wrote:
>
>
> On Tue, Dec 16, 2014 at 2:52 PM, Connor Abbott wrote:
>>
>> Hi,
>>
>> On Tue, Dec 16, 2014 at 1:04 AM, Jason Ekstrand
>> wrot
301 - 400 of 1264 matches
Mail list logo
