On 4 July 2018 at 12:44, wrote:
> From: Roland Scheidegger
>
> fold_assoc() called from fold_alu_op3() can lower the number of src to 2,
> which then leads to an invalid access to n.src[2]->gvalue().
> This didn't seem to have caused much harm in the past, but on Fedora 28
> it will crash (presu
On 4 July 2018 at 16:02, Marek Olšák wrote:
> From: Marek Olšák
>
> This is basically LLVMTargetMachineEmitToMemoryBuffer inlined and reworked.
>
> struct ac_compiler_passes (opaque type) contains the main pass manager.
>
> ac_create_llvm_passes -- the result can go to thread local storage
> ac_d
This is a bit hacky, (esp not sure about the nir bits).
When we enable variable pointers on radv we don't want to do the
offset lowering stuff, so this fixes up all the fallout from enabling
things.
The big issue is a select followed by a load, since the select
can't have a variable pointer in it
From: Dave Airlie
---
src/amd/common/ac_llvm_build.c | 3 +++
src/amd/common/ac_nir_to_llvm.c | 33 +++--
src/amd/vulkan/radv_device.c| 2 +-
3 files changed, 27 insertions(+), 11 deletions(-)
diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llv
From: Dave Airlie
---
src/compiler/spirv/spirv_to_nir.c | 15 ---
src/compiler/spirv/vtn_variables.c | 3 ++-
2 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/src/compiler/spirv/spirv_to_nir.c
b/src/compiler/spirv/spirv_to_nir.c
index fb4211193fb..f8c11aaf929 100644
From: Dave Airlie
---
src/compiler/nir/nir_gather_info.c | 2 ++
src/compiler/nir/nir_lower_indirect_derefs.c | 4
src/compiler/nir/nir_lower_vars_to_ssa.c | 4
3 files changed, 10 insertions(+)
diff --git a/src/compiler/nir/nir_gather_info.c
b/src/compiler/nir/nir_gath
According to RenderDoc, this shaves about 15% off of one of the render
passes in the Skyrim SE frame I've been analyzing on my KBL laptop and
about 8% off the whole frame. It's a bit noisy and hard to tell with
just the DXVK HUD but there appears to be a small FPS bump in-game as
well.
---
src/in
Many fragment shaders do a discard using relatively little information
but still put the discard fairly far down in the shader for no good
reason. If the discard is moved higher up, we can possibly avoid doing
some or almost all of the work in the shader. When this lets us skip
texturing operatio
This makes the arguments match the (thing, container) pattern used in
other nir_foreach macros and also renames it to make that a bit more
clear.
---
src/compiler/nir/nir_opt_dce.c | 3 +--
src/compiler/nir/nir_worklist.h | 4 ++--
2 files changed, 3 insertions(+), 4 deletions(-)
diff --git a/sr
From: Marek Olšák
Compile times of simple shaders are reduced by ~20%.
Compile times of prologs and epilogs are reduced by up to 40%.
---
src/gallium/drivers/radeonsi/si_pipe.c| 2 ++
.../drivers/radeonsi/si_shader_tgsi_setup.c | 33 ++-
2 files changed, 4 insertions(+
From: Marek Olšák
---
.../drivers/radeonsi/si_shader_tgsi_setup.c | 26 +--
1 file changed, 13 insertions(+), 13 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
index 6c1e18ff812..b8cfd15a67f
From: Marek Olšák
---
src/amd/common/ac_llvm_build.c | 6 +++---
src/amd/common/ac_llvm_build.h | 2 +-
src/amd/vulkan/radv_nir_to_llvm.c | 10 --
src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c | 5 +
4 files changed
From: Marek Olšák
This is basically LLVMTargetMachineEmitToMemoryBuffer inlined and reworked.
struct ac_compiler_passes (opaque type) contains the main pass manager.
ac_create_llvm_passes -- the result can go to thread local storage
ac_destroy_llvm_passes -- can be called by a destructor in TLS
Hi,
Reviewed-by: Mathias Fröhlich
best
Mathias
On Sunday, 1 July 2018 10:05:35 CEST Gert Wollny wrote:
> From: Gert Wollny
>
> restart_index is later always used in a comparison, so it should be
> initialized properly.
>
> Fixes valgrind warning:
> Conditional jump or move depends on uninit
Reviewed-by: Timothy Arceri
On 04/07/18 11:51, srol...@vmware.com wrote:
From: Roland Scheidegger
Empty initializer braces aren't valid c (it's a gnu extension, and
it's valid in c++).
Hopefully fixes appveyor / msvc build...
Fixes 6677e131b806b10754adcb7cf3f427a7fcc2aa09
---
src/compiler/
From: Roland Scheidegger
fold_assoc() called from fold_alu_op3() can lower the number of src to 2,
which then leads to an invalid access to n.src[2]->gvalue().
This didn't seem to have caused much harm in the past, but on Fedora 28
it will crash (presumably because -D_GLIBCXX_ASSERTIONS is used,
https://bugs.freedesktop.org/show_bug.cgi?id=107092
--- Comment #5 from Roland Scheidegger ---
(In reply to Cory Quammen from comment #4)
> Thank you for your quick reply and insight, Roland. Creating and destroying
> glX contexts is a bit of a peculiarity in VTK that we may be able to do away
>
From: Roland Scheidegger
Empty initializer braces aren't valid c (it's a gnu extension, and
it's valid in c++).
Hopefully fixes appveyor / msvc build...
Fixes 6677e131b806b10754adcb7cf3f427a7fcc2aa09
---
src/compiler/glsl/gl_nir_link_atomics.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(
Hi,
> + /* Look for any follow-up instructions that sources from the boolean
> +* result of the producer instruction and rewrite them to use the correct
> +* bit-size.
> +*/
> + foreach_inst_in_block_starting_from(fs_inst, fixup_inst, inst) {
> + if (!inst_supports_boolean(fi
great work, some comments below though.
On Wed, Jul 4, 2018 at 12:37 AM, Rhys Perry wrote:
> Signed-off-by: Rhys Perry
> ---
> .../drivers/nouveau/codegen/nv50_ir_lowering_gm107.cpp | 16
>
> .../drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp| 3 +++
> .../drivers/no
> +/**
> + * Check that the register region given by r [r.offset, r.offset + dr[
> + * is exactly the same as the the register region given by s
> + * [s.offset, s.offset + ds[
> + */
> +static inline bool
> +region_match(const fs_reg &r, unsigned dr, const fs_reg &s, unsigned ds)
> +{
> + return
Its unlikely anyone will add proper ARB_direct_state_access compat
support before we branch 18.2. Enabling the extension in 4.5 at
least allows users to make use of MESA_GL_VERSION_OVERRIDE=4.5COMPAT
for games like No Mans Sky.
---
src/mapi/glapi/gen/apiexec.py| 194 +++
Reviewed-by: Bruce Cherniak
> On Jun 28, 2018, at 2:17 PM, Alok Hota wrote:
>
> Sorry for the churn on these patches. They had to be split in three due
> to some code change in between a mass formatting change. Anyway this
> patch contains only formatting changes for the files that were update
The game forgets to enable multiple extensions in its shaders, one
of those extesions is EXT_texture_array. But enabling this config
entry fixes at least one other rendering issue that enabling
EXT_texture_array on its own doesn't fix.
---
src/util/drirc | 4
1 file changed, 4 insertions(+)
Series:
Reviewed-by: Timothy Arceri
On 04/07/18 09:16, Marek Olšák wrote:
From: Marek Olšák
There is a 15-character limit for thread names shared by the queue name
and process name. Shorten the thread name to make space for the process
name.
---
src/gallium/auxiliary/util/u_threaded_contex
From: Marek Olšák
---
src/gallium/auxiliary/os/os_process.c | 16 ++--
src/util/process.c| 2 ++
2 files changed, 4 insertions(+), 14 deletions(-)
diff --git a/src/gallium/auxiliary/os/os_process.c
b/src/gallium/auxiliary/os/os_process.c
index 035bd228e76..8707
From: Marek Olšák
---
src/util/Makefile.sources | 2 +
src/util/process.c| 116 ++
src/util/process.h| 34 +++
src/util/xmlconfig.c | 86 +---
4 files changed, 154 insertions(+), 84 deletions(-)
create
From: Marek Olšák
There is a 15-character limit for thread names shared by the queue name
and process name. Shorten the thread name to make space for the process
name.
---
src/gallium/auxiliary/util/u_threaded_context.c | 2 +-
src/gallium/drivers/radeonsi/si_pipe.c| 4 ++--
src/ga
From: Marek Olšák
---
src/util/u_queue.c | 35 +--
src/util/u_queue.h | 2 +-
2 files changed, 34 insertions(+), 3 deletions(-)
diff --git a/src/util/u_queue.c b/src/util/u_queue.c
index da513fd9cc5..6c92e8140a1 100644
--- a/src/util/u_queue.c
+++ b/src/util/u_q
For patches 9-11:
Reviewed-by: Marek Olšák
Marek
On Mon, Jul 2, 2018 at 8:42 PM, Dave Airlie wrote:
> From: Dave Airlie
>
> Some of the compiler functions are no longer called outside
> the util file.
> ---
> src/amd/common/ac_llvm_util.c | 12 ++--
> src/amd/common/ac_llvm_util.h |
Signed-off-by: Rhys Perry
---
.../drivers/nouveau/codegen/nv50_ir_lowering_gm107.cpp | 16
.../drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp| 3 +++
.../drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h | 2 +-
src/gallium/drivers/nouveau/nv50/nv50_resource.h
On Mon, Jul 2, 2018 at 9:30 AM, Brian Paul wrote:
> On 07/01/2018 02:05 AM, Gert Wollny wrote:
>>
>> From: Gert Wollny
>>
>> restart_index is later always used in a comparison, so it should be
>> initialized properly.
>>
>> Fixes valgrind warning:
>> Conditional jump or move depends on uninitia
Reviewed-by: Caio Marcelo de Oliveira Filho
Also consider including a fixes tag.
Fixes: 6f5abf31466 "i965: Fix output register sizes when multiple variables
share a slot."
On Fri, May 18, 2018 at 03:31:00PM +0200, Neil Roberts wrote:
> In 6f5abf31466aed this code was fixed to calculate the max
On Mon, Jun 18, 2018 at 03:42:13PM -0700, Nanley Chery wrote:
> On Thu, Jun 14, 2018 at 07:50:54PM +0100, Chris Wilson wrote:
> > Quoting Nanley Chery (2018-06-14 19:46:09)
> > > On Thu, Jun 14, 2018 at 10:01:18AM -0700, Nanley Chery wrote:
> > > > On Thu, Jun 14, 2018 at 04:18:30PM +0300, Martin P
Both are R-b me
On July 3, 2018 15:25:33 Anuj Phogat wrote:
Bump
On Fri, Jun 1, 2018 at 2:40 PM Anuj Phogat wrote:
CACHE_MODE_SS is not listed in gfxspecs table for user mode
non-privileged registers. So, making any changes from Mesa
will do nothing. Kernel is already setting this bit in
CA
Rb
On July 3, 2018 15:31:38 "Ian Romanick" wrote:
From: Ian Romanick
Some of the lowering passes, nir_lower_locals_to_regs for example, can
cause some previously live code to be dead. This pass in particular
leaves a bunch of nir_instr_type_deref instructions floating around.
This causes sh
From: Ian Romanick
Some of the lowering passes, nir_lower_locals_to_regs for example, can
cause some previously live code to be dead. This pass in particular
leaves a bunch of nir_instr_type_deref instructions floating around.
This causes shader-db runs on Gen5 through Haswell to spew tons of
me
Bump.
On Fri, Jun 1, 2018 at 2:40 PM Anuj Phogat wrote:
>
> CACHE_MODE_SS is not listed in gfxspecs table for user mode
> non-privileged registers. So, making any changes from Mesa
> will do nothing. Kernel is already setting this bit in
> CACHE_MODE_SS register which is saved/restored to/from
> t
Bump
On Fri, Jun 1, 2018 at 2:40 PM Anuj Phogat wrote:
>
> CACHE_MODE_SS is not listed in gfxspecs table for user mode
> non-privileged registers. So, making any changes from Mesa
> will do nothing. Kernel is already setting this bit in
> CACHE_MODE_SS register which is saved/restored to/from
> th
From: Dave Airlie
I'd like to encourage people to test this to see if it helps (like
does it make app startup better or less hitching in dxvk).
The basic idea is to store a bunch of LLVM related data structs
in thread local storage so we can avoid reiniting them every time
we compile a shader. S
https://bugs.freedesktop.org/show_bug.cgi?id=107090
Caio Marcelo de Oliveira Filho changed:
What|Removed |Added
Status|NEEDINFO|RESOLVED
Resol
Build mesa 8054 failed
Commit 584ad1eda9 by Dave Airlie on 6/26/2018 10:52 PM:
ac/radeonsi: refactor out pass manager init to common code.\n\nReviewed-by: Marek Olšák
Configure your notification preferences
__
On 4 July 2018 at 03:11, Marek Olšák wrote:
> If radeonsi adds target library info, I highly recommend that radv does the
> same thing.
>
We can't due to the leak situation, there is no case where radv can just
leak stuff like radeonsi does here, until we get to the thread allocs.
Dave.
Build mesa 8053 completed
Commit 81af1a0ae2 by Timothy Arceri on 6/4/2018 6:26 AM:
nir: fix selection of loop terminator when two or more have the same limit\n\nWe need to add loop terminators to the list in the order we come\nacross them otherwise if two or mo
Hi, I'm having an error using DXVK in some games: `VK_ERROR_DEVICE_LOST`
Also, it was checked with the new Mesa version 18.1.3.
You can will get acquainted with a problem according to the reference:
https://github.com/doitsujin/dxvk/issues/469
___
mesa-d
On Tue, Jul 3, 2018 at 10:24 AM, Ian Romanick wrote:
> On 07/03/2018 07:09 AM, Jason Ekstrand wrote:
> > Do we do this in fs? I'm very confused as to how any deref instructions
> > can survive this late. They can for fs but only for images (not
> > textures). I would expect them to be killed of
On 07/03/2018 07:09 AM, Jason Ekstrand wrote:
> Do we do this in fs? I'm very confused as to how any deref instructions
> can survive this late. They can for fs but only for images (not
> textures). I would expect them to be killed off by DCE long before we
> get here.
I didn't debug this to gre
For patches 4-8:
Reviewed-by: Marek Olšák
Marek
On Mon, Jul 2, 2018, 8:48 PM Dave Airlie wrote:
> From: Dave Airlie
>
> This doesn't do much yet, but it makes it easier to move the code
> to a common shared code base.
> ---
> src/amd/common/ac_llvm_util.h | 1 +
> src/amd/vulkan/rad
If radeonsi adds target library info, I highly recommend that radv does the
same thing.
Marek
On Mon, Jul 2, 2018, 8:48 PM Dave Airlie wrote:
> From: Dave Airlie
>
> This adds a inline always pass, but otherwise should work the
> same.
> ---
> src/amd/common/ac_llvm_util.c | 5 +++--
> s
s/init/create/ ?
Anyway, patch 1-3:
Reviewed-by: Marek Olšák
On Mon, Jul 2, 2018, 8:48 PM Dave Airlie wrote:
> From: Dave Airlie
>
> ---
> src/amd/common/ac_llvm_util.c | 30 ++
> src/amd/common/ac_llvm_util.h | 2 ++
> src/gallium/drivers/radeonsi/
https://bugs.freedesktop.org/show_bug.cgi?id=107092
--- Comment #4 from Cory Quammen ---
Thank you for your quick reply and insight, Roland. Creating and destroying glX
contexts is a bit of a peculiarity in VTK that we may be able to do away with -
I'm not sure why VTK does that, so maybe we can
https://bugs.freedesktop.org/show_bug.cgi?id=107092
Cory Quammen changed:
What|Removed |Added
CC||chuck.atk...@kitware.com
--
You are rec
Do we do this in fs? I'm very confused as to how any deref instructions
can survive this late. They can for fs but only for images (not textures).
I would expect them to be killed off by DCE long before we get here.
On July 3, 2018 00:26:32 "Ian Romanick" wrote:
From: Ian Romanick
Previo
Ugh... Why was that not unreachable before?... R-b.
On July 3, 2018 00:26:36 "Ian Romanick" wrote:
From: Ian Romanick
The bug fixed by the previous commit went undetected because extra
stderr messages are not flagged by the CI. Copy the solution from
fs_visitor::nir_emit_instr and mark the
Hi mesa devs,
Can anyone please review this ?
This patch fixes following bugs:
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104388
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106158
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106667
2018-06-07 18:27 GMT+03:00 Vad
On 03/07/18 02:29, Timothy Arceri wrote:
> Series looks good to me.
Just pushed to master. Thanks!
> How much to go before we can turn this extension on?
TL; DR: In summary there is a lot already done, but there is still work
to do. In general, we tried to cover as much features as possible, an
Build mesa 8052 failed
Commit 590cc7c8f6 by Neil Roberts on 11/29/2017 9:14 AM:
i965: Use the new nir atomic counter linker for SPIR-V shaders\n\nReviewed-by: Timothy Arceri
Configure your notification preferences
___
Signed-off-by: Gert Wollny
---
Note that I used (1 << 3) instead of (1 << 2) because Dave's experimental patch
already uses the latter for computer shader caps.
The host side patch starting to implement this can be found here:
https://patchwork.freedesktop.org/series/45064/
Thanks for reviewi
That might reduce CPU overhead a little bit when using
RADV_TRACE_FILE.
Signed-off-by: Samuel Pitoiset
---
src/amd/vulkan/radv_cmd_buffer.c | 11 +++
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c
index 26
The number of enabled descriptors for a given pipeline stage
can be computed at compile time.
Signed-off-by: Samuel Pitoiset
---
src/amd/vulkan/radv_cmd_buffer.c | 11 ++-
src/amd/vulkan/radv_nir_to_llvm.c | 7 +--
src/amd/vulkan/radv_shader.h | 1 +
3 files changed, 8 insert
Hi Frank,
On 28 June 2018 at 11:12, Frank Binns wrote:
> Hi Qiang,
>
> Qiang Yu writes:
>
>> Hi Harish,
>>
>> I want to implement EGL_KHR_partial_update for lima mesa driver and find you
>> worked on Android/Wayland support for it:
>> https://patchwork.freedesktop.org/patch/160944/
>> https://pa
On Tuesday, 2018-07-03 11:26:44 +1000, Dave Airlie wrote:
> From: Dave Airlie
>
> I'd like to encourage people to test this to see if it helps (like
> does it make app startup better or less hitching in dxvk).
>
> The basic idea is to store a bunch of LLVM related data structs
> in thread local
On 03/07/18 05:51, Eric Anholt wrote:
Eric Anholt writes:
[ Unknown signature status ]
Timothy Arceri writes:
nir_compact_varyings() is meant to skip over varyings used by xfb:
/* We can't repack xfb varyings. */
if (var->data.always_active_io)
continue
https://bugs.freedesktop.org/show_bug.cgi?id=107098
Bug ID: 107098
Summary: Segfault after munmap(kms_sw_dt->ro_mapped)
Product: Mesa
Version: 18.0
Hardware: x86-64 (AMD64)
OS: Linux (All)
Status: NEW
Seve
https://bugs.freedesktop.org/show_bug.cgi?id=107096
Michel Dänzer changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
On Mon, 2018-07-02 at 08:23 -0500, Jason Ekstrand wrote:
> On July 2, 2018 01:09:38 Iago Toral wrote:
>
> > On Sun, 2018-07-01 at 18:30 -0500, Jason Ekstrand wrote:
> > > On June 29, 2018 03:11:00 Iago Toral Quiroga
> > > wrote:
> > >
> > > > ---
> > > > src/intel/vulkan/anv_private.h | 5
66 matches
Mail list logo
