https://bugs.freedesktop.org/show_bug.cgi?id=84023
Fabio Pedretti changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
On 30/12/14 22:20, Laura Ekstrand wrote:
> To run this partial implementation,
>
> export MESA_EXTENSION_OVERRIDE=+GL_ARB_direct_state_access
>
Indeed that does the job. Yet it seems that I was slightly confused with
the usage/application of the boolean variable(s) - i.e. they seem to be
used whe
https://bugs.freedesktop.org/show_bug.cgi?id=87886
--- Comment #9 from almos ---
I just checked L4D2 with mesa 10.3.2 (AMD Barts), and I see no such fps drops.
Sure, in the beginning there are some hiccups, but once all shaders have been
used at least once, everything is smooth.
--
You are rece
On 02/01/15 04:14, Jeremy Huddleston Sequoia wrote:
> This is certainly not the best solution to the problem, so I'm just sending
> this patch to the list to get the discussion started on the best way to solve
> this problem. Currently, any platform that does not support _SC_PHYS_PAGES
> and _S
https://bugs.freedesktop.org/show_bug.cgi?id=87925
--- Comment #2 from Emil Velikov ---
Quick look at my old piglit results indicates:
- It dates back at least until 10.2
- Only the classic swrast is affected
--
You are receiving this mail because:
You are the assignee for the bug.
__
https://bugs.freedesktop.org/show_bug.cgi?id=87886
--- Comment #10 from Stéphane Travostino ---
(In reply to almos from comment #9)
> I just checked L4D2 with mesa 10.3.2 (AMD Barts), and I see no such fps
> drops. Sure, in the beginning there are some hiccups, but once all shaders
> have been us
We can use the intersection function to reduce the range
even further if the operand has bounds between 0.0 and 1.0.
v2: Rewrite as suggested by Ian.
Put in alphabetic order.
---
src/glsl/opt_minmax.cpp | 5 +
1 file changed, 5 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/gls
V2: Correct whitespace issues
Put in alphabetic order
Reviewed-by: Ian Romanick
---
src/glsl/opt_minmax.cpp | 12
1 file changed, 12 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index e4aa677..dd13800 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/sr
They are bound between -1 and 1, so exploit that.
v2: Put in alphabetical order
Just return directly instead of assigning to temp variable.
---
src/glsl/opt_minmax.cpp | 8
1 file changed, 8 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 2faa3c3..
V2: Fix whitespace issues
Put in alphabetic order
Split resolve of add into a separate function for reuse later
Reviewed-by: Ian Romanick
---
src/glsl/opt_minmax.cpp | 23 +++
1 file changed, 23 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax
V2: Put in alphabetic order
Add some air for readability
Reviewed-by: Ian Romanick
---
src/glsl/opt_minmax.cpp | 11 +++
1 file changed, 11 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 04ae0c0..43d6061 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/
This improves one shader from Brutal Legend.
It does a lrp that yields a result larger than 1,
that then gets saturated. This is replaced by a constant 1.
shaders/BrutalLegend/214.shader_test fs16:84 -> 73 (-13.10%)
shaders/BrutalLegend/214.shader_test fs8: 84 -> 73 (-13.10%)
---
This allows opt_algebraic to resolve open-coded
saturates into ir_unop_saturate before we potentially
mess it up by removing the min or max in min/max-pruning.
Since we are now emitting more free saturates on i965
this gives us some decrease in instruction count.
total instructions in shared prog
V2: Put in alphabetic order
Rework to use new IS_CONSTANT macro
Reduce code duplication by combining pos only and neg only variation
---
src/glsl/opt_minmax.cpp | 29 +
1 file changed, 29 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.
---
src/glsl/opt_minmax.cpp | 16
1 file changed, 16 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 8d63d96..f066b45 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/src/glsl/opt_minmax.cpp
@@ -457,6 +457,22 @@ get_range(ir_rvalue *rval)
//
---
src/glsl/opt_minmax.cpp | 9 +
1 file changed, 9 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index cca18e8..8d63d96 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/src/glsl/opt_minmax.cpp
@@ -448,6 +448,15 @@ get_range(ir_rvalue *rval)
return mi
V2: Add some air for readability
Use the new IS_CONSTANT macro
Combine if-blocks for reduced code-duplication
Split out into separate function for reuse later
---
src/glsl/opt_minmax.cpp | 33 +
1 file changed, 33 insertions(+)
diff --git a/src/glsl/opt
V2: Replace range-setting for undefined behavior with a TODO.
Use the new IS_CONSTANT macro
--- 3< ---
Ian mentioned he was not sure what the correct behavior should be.
Optimizing based on undefined behavior is not pretty,
so remove that and instead leave a note.
---
src/glsl/opt_minmax.cpp |
---
src/glsl/opt_minmax.cpp | 24
1 file changed, 24 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index f066b45..5515f51 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/src/glsl/opt_minmax.cpp
@@ -411,6 +411,30 @@ get_range(ir_rvalue *rval)
---
src/glsl/opt_minmax.cpp | 31 +++
1 file changed, 31 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 441ac69..07d3d53 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/src/glsl/opt_minmax.cpp
@@ -78,6 +78,8 @@ public:
ir_rvalue *op
Try to resolv compares by checking the possible variable range.
---
src/glsl/opt_minmax.cpp | 101
1 file changed, 101 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 07d3d53..8cf0fe5 100644
--- a/src/glsl/opt_min
The spec states that log / log2 of x <= 0 is undefined.
Just set the range to 0 if this is the case.
Also set the range according to how we are oriented
around one, as this causes us to shift from positive to
negative values.
V2: Utilize the new IS_CONSTANT() macro
Put in alphabetic order
---
v2: Set in alphabetic order
Change to use the new IS_CONSTANT macro
Add a function for creating a minmax_range from two constants
Use the function to improve the resolution of abs's range.
---
src/glsl/opt_minmax.cpp | 28
1 file changed, 28 insertions(+)
---
src/glsl/opt_minmax.cpp | 24
1 file changed, 24 insertions(+)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index 8cf0fe5..3d3cfff 100644
--- a/src/glsl/opt_minmax.cpp
+++ b/src/glsl/opt_minmax.cpp
@@ -80,6 +80,8 @@ public:
ir_rvalue *opt_compa
I couple months ago i posted a series for expanding the get_range
function in minmax_pruning with support for more operations.
So I've been hacking on this during my spare time in Christmas.
I've now gotten to a point where I think this is not getting us
anywhere, and I need some opinions from othe
This will make expansion easier and less cluttered.
v2: Correct missing space
---
src/glsl/opt_minmax.cpp | 20 +++-
1 file changed, 15 insertions(+), 5 deletions(-)
diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
index a3b0a65..56805c0 100644
--- a/src/glsl/opt_mi
This will allow for less code duplication.
I'll be using this in opt_minmax in the comming commits.
v2: Don't "use namespace"
Remove author tag
---
src/glsl/ir_constant_util.h | 121
src/glsl/opt_algebraic.cpp | 85 +--
Also handle undefined behaviour for sqrt(x) where x < 0
and rsq(x) where x <= 0.
This gives us some reduction in instruction count on three
Dungeon Defenders shaders as they are doing: max(exp(x), 0)
v2: Change to use new IS_CONSTANT() macro
Fix high unintenionally not being returned
Add
Change opt_algebraic to use the new macro.
Remove the old less than / greater than functions.
---
src/glsl/ir_constant_util.h | 49 ++---
src/glsl/opt_algebraic.cpp | 8
2 files changed, 23 insertions(+), 34 deletions(-)
diff --git a/src/glsl/ir_
We have had fast-paths in our driver for uploading 8-bit RGBA or BGRA
textures for some time now. The basic idea behind the fast-path is that we
can tile/detile in software faster than going through the GTT or waiting on
the blitter to do it for us. This series adds the capability of doing it
in
There's no reason why we should be doing this for 2D textures and not
rectangles. Just a matter of adding another hunk to the condition.
Signed-off-by: Jason Ekstrand
---
src/mesa/drivers/dri/i965/intel_tex_subimage.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/mes
From: Sisinty Sasmita Patra
Added intel_readpixels_tiled_mempcpy and intel_gettexsubimage_tiled_mempcpy
functions. These are the fast paths for glReadPixels and glGetTexImage.
v2: Jason Ekstrand
- Refactor to make the functions look more like the old
intel_tex_subimage_tiled_memcpy
-
From: Sisinty Sasmita Patra
This commit addes tiled copy functions for coping from tiled memory to
linear memory. These are very similar to the existing linear-to-tiled
paths.
v2: Jason Ekstrand
- New commit message
- Various whitespace fixes
- Added ptrdiff_t casts as done in commit
From: Sisinty Sasmita Patra
This commit refactors the tiled_memcpy code in intel_tex_subimage.c and
moves it into its own file intel_tiled_memcpy files. Also, xtile_copy and
ytile_copy are renamed to linear_to_xtiled and linear_to_ytiled
respectively. The *_faster functions are similarly rename
And I forgot to CC chad...
On Sat, Jan 3, 2015 at 11:54 AM, Jason Ekstrand
wrote:
> We have had fast-paths in our driver for uploading 8-bit RGBA or BGRA
> textures for some time now. The basic idea behind the fast-path is that we
> can tile/detile in software faster than going through the GTT
I was just looking at the git log and noticed the commit:
i965: Show opt_vector_float() and later passes in INTEL_DEBUG=optimizer.
It converts a macro to use the GNU Statement Expression extension.
This would greatly simplify the macro I've written in this patch,
so I'm considering rewriting it
On Sat, Jan 3, 2015 at 12:10 PM, Thomas Helland
wrote:
> I was just looking at the git log and noticed the commit:
>i965: Show opt_vector_float() and later passes in INTEL_DEBUG=optimizer.
> It converts a macro to use the GNU Statement Expression extension.
> This would greatly simplify the ma
On Sat, Jan 3, 2015 at 11:18 AM, Thomas Helland
wrote:
> ---
> src/glsl/opt_minmax.cpp | 16
> 1 file changed, 16 insertions(+)
>
> diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
> index 8d63d96..f066b45 100644
> --- a/src/glsl/opt_minmax.cpp
> +++ b/src/glsl/opt_
2015-01-03 21:55 GMT+01:00 Matt Turner :
> On Sat, Jan 3, 2015 at 11:18 AM, Thomas Helland
> wrote:
>> ---
>> src/glsl/opt_minmax.cpp | 16
>> 1 file changed, 16 insertions(+)
>>
>> diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp
>> index 8d63d96..f066b45 100644
>>
On Sat, Jan 3, 2015 at 11:18 AM, Thomas Helland
wrote:
> This allows opt_algebraic to resolve open-coded
> saturates into ir_unop_saturate before we potentially
> mess it up by removing the min or max in min/max-pruning.
>
> Since we are now emitting more free saturates on i965
> this gives us som
On Sat, Jan 3, 2015 at 11:18 AM, Thomas Helland
wrote:
> This improves one shader from Brutal Legend.
> It does a lrp that yields a result larger than 1,
> that then gets saturated. This is replaced by a constant 1.
Not a comment on your patch, but the code in question does this:
> const vec4 ps
V2: Fix the logic so we don't emit bogus results.
Fix the comments (spotted by Matt)
---
Now, when we do the correct thing, we no longer get a
reduction in instructions on that one Brutal Legend shader.
src/glsl/opt_minmax.cpp | 16
1 file changed, 16 insertions(+)
diff --g
https://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.69/html_node/Limitations-of-Usual-Tools.html#index-g_t_0040command_007btr_007d-1842
Without this fix, egl fails to build on Solaris, with the error:
:0:22: error: '_EGL_PLATFORM_x11' undeclared (first use in this
function)
egl
Reviewed-by: Matt Turner
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
https://bugs.freedesktop.org/show_bug.cgi?id=87886
--- Comment #11 from Stéphane Travostino ---
After spending half a day bisecting, I don't think there's any real difference
between Mesa 10.3.2 and master: the fps drops happen in both releases, although
they seem to be more infrequent in 10.3.2
Hi,
I noticed some GLSL related discussions talk "My shader-db is dominated by
TF2, DOTA2, Portal, Brutal Legend and Dungeon Defenders. Maybe
non-Source-engine
games show some benefit from this series?"
Now, shader-db that I can find is this:
http://cgit.freedesktop.org/mesa/shader-db/tree/shader
46 matches
Mail list logo
