On Sunday, December 21, 2014 09:19:53 PM Jason Ekstrand wrote:
> On Sun, Dec 21, 2014 at 9:11 PM, Kenneth Graunke
> wrote:
> >
> > On Friday, December 19, 2014 02:20:54 PM Ian Romanick wrote:
> > > From: Ian Romanick
> > >
> > > Instead of having an extra pointer indirection in one of the hottest
On Friday, December 19, 2014 02:20:52 PM Ian Romanick wrote:
> From: Ian Romanick
>
> ...so that it can be inlined in the two places that call it.
>
> On Bay Trail-D using Fedora 20 compile flags (-m64 -O2 -mtune=generic
> for 64-bit and -m32 -march=i686 -mtune=atom for 32-bit), affects
> Gl32Ba
On Friday, December 19, 2014 02:20:59 PM Ian Romanick wrote:
> From: Ian Romanick
>
> You would not believe the mess GCC 4.8.3 generated for the old
> switch-statement.
>
> On Bay Trail-D using Fedora 20 compile flags (-m64 -O2 -mtune=generic
> for 64-bit and -m32 -march=i686 -mtune=atom for 32-
On Friday, December 19, 2014 02:20:55 PM Ian Romanick wrote:
> From: Ian Romanick
>
> On Bay Trail-D using Fedora 20 compile flags (-m64 -O2 -mtune=generic
> for 64-bit and -m32 -march=i686 -mtune=atom for 32-bit), affects
> Gl32Batch7:
>
> 32-bit: Difference at 95.0% confidence 0.495267% +/- 0.
On Sun, Dec 21, 2014 at 9:11 PM, Kenneth Graunke
wrote:
>
> On Friday, December 19, 2014 02:20:54 PM Ian Romanick wrote:
> > From: Ian Romanick
> >
> > Instead of having an extra pointer indirection in one of the hottest
> > loops in the driver.
> >
> > On Bay Trail-D using Fedora 20 compile flag
On Friday, December 19, 2014 02:20:54 PM Ian Romanick wrote:
> From: Ian Romanick
>
> Instead of having an extra pointer indirection in one of the hottest
> loops in the driver.
>
> On Bay Trail-D using Fedora 20 compile flags (-m64 -O2 -mtune=generic
> for 64-bit and -m32 -march=i686 -mtune=ato
On Sun, Dec 21, 2014 at 7:50 PM, Kenneth Graunke wrote:
> On Sunday, December 21, 2014 03:37:25 PM Matt Turner wrote:
>> ---
>> No shader-db changes, unfortunately.
>
> Unsurprising :) This patch doesn't implement what you meant.
>
> 1. instructions_match() checks a->conditional_mod == b->conditio
On Friday, December 19, 2014 02:20:53 PM Ian Romanick wrote:
[snip]
> diff --git a/src/mesa/drivers/dri/i965/brw_context.h
> b/src/mesa/drivers/dri/i965/brw_context.h
> index a63c483..92eb022 100644
> --- a/src/mesa/drivers/dri/i965/brw_context.h
> +++ b/src/mesa/drivers/dri/i965/brw_context.h
> @
On Sunday, December 21, 2014 03:37:25 PM Matt Turner wrote:
> ---
> No shader-db changes, unfortunately.
Unsurprising :) This patch doesn't implement what you meant.
1. instructions_match() checks a->conditional_mod == b->conditional_mod.
So you won't ever see mixed conditional mods in operand
---
No shader-db changes, unfortunately.
src/mesa/drivers/dri/i965/brw_fs_cse.cpp | 14 ++
src/mesa/drivers/dri/i965/brw_vec4_cse.cpp | 18 --
2 files changed, 22 insertions(+), 10 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_fs_cse.cpp
b/src/mesa/driver
total instructions in shared programs: 5877012 -> 5876617 (-0.01%)
instructions in affected programs: 33140 -> 32745 (-1.19%)
From before the commit that allows VF constant propagation (which hurt
some programs) to here, the results are:
total instructions in shared programs: 5877951 -> 58766
I don't feel great about unreachable("unimplemented: ...") but these
cases do only seem possible under some currently impossible circumstances.
---
src/mesa/drivers/dri/i965/brw_fs.cpp | 16 +++
src/mesa/drivers/dri/i965/brw_shader.cpp | 47
src/mesa/dr
total instructions in shared programs: 5877951 -> 5877012 (-0.02%)
instructions in affected programs: 155923 -> 154984 (-0.60%)
Helps 1233, hurts 156 shaders. The hurt shaders are addressed in the
next commit.
---
.../drivers/dri/i965/brw_vec4_copy_propagation.cpp | 28 +-
After CSEing some MOV ..., VF instructions we have code like
mov tmp, [1F, 2F, 3F, 4F]VF
mov r10, tmp
mov r11, tmp
...
use r10
use r11
We want to copy propagate tmp into the uses of r10 and r11, but *not*
constant propagate the VF immediate into the uses of tmp.
---
src/mesa/dr
total instructions in shared programs: 5869005 -> 5868220 (-0.01%)
instructions in affected programs: 70208 -> 69423 (-1.12%)
---
src/mesa/drivers/dri/i965/brw_vec4.cpp | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp
b/src/mesa/d
Currently only handles consecutive instructions with the same
destination that collectively write all channels.
total instructions in shared programs: 5879798 -> 5869011 (-0.18%)
instructions in affected programs: 465236 -> 454449 (-2.32%)
---
src/mesa/drivers/dri/i965/brw_vec4.cpp | 61 +
Port of commit a28ad9d4 from the fs backend.
No shader-db changes since we don't emit MOV ..., VF instructions yet.
---
src/mesa/drivers/dri/i965/brw_vec4_cse.cpp | 16 +++-
1 file changed, 11 insertions(+), 5 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_cse.cpp
b/sr
Sometimes it's easier to generate 4x values into an array, and the
memcpy is 1 instruction, rather than 11 to piece 4 arguments together.
I'd forgotten to remove the prototype from fs_reg from a previous patch,
so it's already there for us here.
---
src/mesa/drivers/dri/i965/brw_fs.cpp | 9
This series adds support to i965's vec4 backend for using vector-float
immediates. The shader-db results are pretty nice:
total instructions in shared programs: 5889529 -> 5876617 (-0.22%)
instructions in affected programs: 465347 -> 452435 (-2.77%)
and there's still room for improvement. It
https://bugs.freedesktop.org/show_bug.cgi?id=87516
--- Comment #4 from Leith Bade ---
Looks good.
Are you able to send feedback upstream to the GL ARB about the vagueness around
having no support binaryFormats in the spec?
--
You are receiving this mail because:
You are the QA Contact for the
https://bugs.freedesktop.org/show_bug.cgi?id=87516
--- Comment #3 from Ian Romanick ---
I sent some patches to the mesa-dev mailing list:
http://lists.freedesktop.org/archives/mesa-dev/2014-December/073210.html
http://lists.freedesktop.org/archives/mesa-dev/2014-December/073212.html
http://lists
On Sun, Dec 21, 2014 at 11:40 AM, Ian Romanick wrote:
>>> diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c
>>> b/src/mesa/drivers/dri/i965/brw_draw_upload.c
>>> index 6e0cf3e..e46c54b 100644
>>> --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
>>> +++ b/src/mesa/drivers/dri/i965/brw_draw
From: Ian Romanick
Signed-off-by: Ian Romanick
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87516
---
src/mesa/main/shaderapi.c | 26 --
1 file changed, 20 insertions(+), 6 deletions(-)
diff --git a/src/mesa/main/shaderapi.c b/src/mesa/main/shaderapi.c
index 1
From: Ian Romanick
Signed-off-by: Ian Romanick
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87516
---
src/mesa/main/shaderapi.c | 27 +--
1 file changed, 25 insertions(+), 2 deletions(-)
diff --git a/src/mesa/main/shaderapi.c b/src/mesa/main/shaderapi.c
index
From: Ian Romanick
There are no binary formats supported, so what are you doing? At least
this gives the application developer some feedback about what's going
on. The spec gives no guidance about what to do in this scenario.
Signed-off-by: Ian Romanick
Bugzilla: https://bugs.freedesktop.org/
On 20 December 2014 at 14:21, Olivier Galibert wrote:
> Here is an implementation I've written myself, so no license issues.
>
Thanks OG,
Afaics the main issue is not the lack of implementation, but that
no-one wants to step up to "maintain" it.
Even adding code that is x2 the size is considered
On 12/19/2014 06:22 PM, Matt Turner wrote:
> On Fri, Dec 19, 2014 at 2:20 PM, Ian Romanick wrote:
>> From: Ian Romanick
>>
>> With the switch-statement, GCC 4.8.3 produces a small pile of code with
>> a branch.
>>
>> :
>> 00: 8b 54 24 04 mov0x4(%esp),%edx
>>
On 12/19/2014 06:10 PM, Matt Turner wrote:
> On Fri, Dec 19, 2014 at 2:20 PM, Ian Romanick wrote:
>> From: Ian Romanick
>>
>> ...so that it can be inlined in the two places that call it.
>
>if (brw->meta_in_progress != _mesa_meta_in_progress(ctx)) {
> brw->meta_in_progress = _mesa_meta
Jason asked on IRC what the overall results were for the series. Here
is that data. n=50 for all four.
Gl32Batch7 Gl32Multithread
32-bit 4.11846% +/- 0.216796% 1.97044% +/- 0.161024%
64-bit 5.48027% +/- 0.368804% 2.46053% +/- 0.19554%
On 12/19/2014 02:22 PM, Ian Romanic
Hello guys,
Can anyone backport the following patches for 10.3 and let me know if
we should drop them.
They have a bunch of non-trivial conflicts, thus i would rather avoid
butchering things up.
38ec1844193570486bf35e25a7a4339a00da127e r600: fix texture gradients
instruction emission (v2)
c883856
https://bugs.freedesktop.org/show_bug.cgi?id=85380
Niels Ole Salscheider changed:
What|Removed |Added
CC||niels_ole@salscheider-onlin
Rob Clark writes:
> Eric,
>
> we've managed to find the alpha-ref bitfield on a4xx, so suddenly I
> don't actually need this anymore..
>
> I've pushed a slightly updated version where I'd fixed a few more
> things (and was passing the fbo-alphatest-formats tests for me) to my
> github tree, in ca
32 matches
Mail list logo
