On 01/09/17 09:18 PM, Marek Olšák wrote:
> Well, this patch shouldn't change behavior. Are you sure it's not some
> random thing?
Pretty sure. I never saw this GPUVM fault before this commit, but with
it or later commits it's reproducible reliably.
> I know, for example, that SI isn't very stabl
https://bugs.freedesktop.org/show_bug.cgi?id=99591
--- Comment #16 from Michel Dänzer ---
The recommended LLVM build configuration is -DBUILD_SHARED_LIBS=OFF
-DLLVM_BUILD_LLVM_DYLIB=ON . BUILD_SHARED_LIBS is only intended for LLVM
developers.
--
You are receiving this mail because:
You are the
https://bugs.freedesktop.org/show_bug.cgi?id=102518
Timothy Arceri changed:
What|Removed |Added
Status|NEW |NEEDINFO
--
You are receiving this ma
https://bugs.freedesktop.org/show_bug.cgi?id=102518
--- Comment #2 from Timothy Arceri ---
It's hitting an error in the no error path. Did you force no error via the env
var?
--
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.___
For approx 5-10% performance improvement in dota2.
---
src/amd/vulkan/radv_debug.h | 2 +-
src/amd/vulkan/radv_device.c | 2 +-
src/amd/vulkan/winsys/amdgpu/radv_amdgpu_winsys.c | 2 +-
3 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/src/am
The separate flush cs gets in the way of batchchain.
---
src/amd/vulkan/radv_device.c | 94 +--
src/amd/vulkan/radv_private.h | 3 +-
2 files changed, 37 insertions(+), 60 deletions(-)
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device
On September 2, 2017 1:20:09 AM Chad Versace wrote:
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead
https://bugs.freedesktop.org/show_bug.cgi?id=101832
Bernhard Rosenkraenzer changed:
What|Removed |Added
Summary|[regression][bisect] Xorg |[PATCH][regression][bisect]
On Sat, Sep 2, 2017 at 2:55 AM, Jan Vesely wrote:
> Denotes native half precision float operations capability
>
> Signed-off-by: Jan Vesely
> ---
> I can change the spelling to HALVES, but simplified english sounded more
> appropriate.
>
> src/gallium/docs/source/screen.rst | 1 +
>
On Fri, Sep 1, 2017 at 7:55 PM, Jan Vesely wrote:
> Denotes native half precision float operations capability
>
> Signed-off-by: Jan Vesely
> ---
> I can change the spelling to HALVES, but simplified english sounded more
> appropriate.
>
> src/gallium/docs/source/screen.rst | 1 +
>
https://bugs.freedesktop.org/show_bug.cgi?id=77449
Kai changed:
What|Removed |Added
Depends on||95346
Referenced Bugs:
https://bugs.freedesktop.
https://bugs.freedesktop.org/show_bug.cgi?id=95346
Kai changed:
What|Removed |Added
Blocks||77449
Referenced Bugs:
https://bugs.freedesktop.
https://bugs.freedesktop.org/show_bug.cgi?id=102518
Kai changed:
What|Removed |Added
Summary|Crash in _mesa_is_bufferobj |[apitrace,backtrace] Crash
|dur
n Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> I'm bringing up Vulkan in the Android container of Chrome OS (ARC++).
>
> On Android, stdio goes to /dev/null. On Android, remote gdb is even more
> painful than the usual remote gdb. On Android, nothing works like you
> expect and debugging is
.
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> NOT FOR UPSTREAM.
>
> To get the driver's build-id, anv_physical_device_init_uuids() searches
> the current process for an ELF phdr for filename "libvulkan_intel.so".
> However, Android requires that the library be named
> "vulkan.${board}.s
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> If this flag is set, then the image and it's device memory have the same
> lifetime. vkDestroyImage will free its anv_device_memory.
>
> We need this for VK_ANDROID_native_buffer, because that extension
> creates the VkImage and imports its m
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> If ANV_BO_CACHE_IMPORT_IGNORE_SIZE_PARAM is set, then the function
> ignores the 'size' parameter. Instead, on a cache hit, it uses the cached
> bo's
> size; on a cache miss, it queries the fd's size.
>
Again, I don't really like the flag. I
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> This patch adds a flag param to anv_bo_cache_import() and defines
> exactly one flag, ANV_BO_CACHE_IMPORT_NO_CLOSE_FD. If set, the function
> will not close the fd.
>
The other option (which I think I like better) would be to move the call to
Reviewed-by: Jason Ekstrand
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> This helps when debugging a broken entrypoint table.
> ---
> src/intel/vulkan/anv_entrypoints_gen.py | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/src/intel/vulkan/anv_entrypoints_g
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> Android's Vulkan loader implements VK_KHR_surface and VK_KHR_swapchain,
> and applications cannot access the driver's implementation. Moreoever,
> if the driver exposes the those extension strings, then tests
> dEQP-VK.api.info.instance.extens
This is going to conflict badly with tapani's work to implement
VK_EXT_debug_report (which I need to finish reviewing).
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> In src/intel/vulkan/*, redirect all instances of printf, vk_error,
> anv_loge, anv_debug, anv_finishme, anv_perf_warn, anv
On Sat, Sep 2, 2017 at 1:17 AM, Chad Versace
wrote:
> This implementation is correct (afaict), but takes two shortcuts
> regarding the import/export of Android sync fds.
>
> Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
> fd into a VkSemaphore or VkFence, the driver in
https://bugs.freedesktop.org/show_bug.cgi?id=102518
Kai changed:
What|Removed |Added
Keywords||have-backtrace
--- Comment #1 from Kai ---
I've
https://bugs.freedesktop.org/show_bug.cgi?id=102502
--- Comment #1 from Alexandre Demers ---
It seems this change (commit 707d2e8b) is responsible for a segfault on launch.
However, even when reverting priror to 707d2e8b, Kodi will segfault when
launching a stream. I'm investigating this other bu
https://bugs.freedesktop.org/show_bug.cgi?id=102518
Bug ID: 102518
Summary: Crash in _mesa_is_bufferobj during load of "XCOM 2:
War of the Chosen"
Product: Mesa
Version: git
Hardware: Other
OS: All
https://bugs.freedesktop.org/show_bug.cgi?id=77449
Kai changed:
What|Removed |Added
Depends on||102518
Referenced Bugs:
https://bugs.freedesktop
Adding back people that got accidentally dropped from the CC list.
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
On Thu 31 Aug 2017, Tapani Pälli wrote:
> Hi;
Hi Tapani, I just sent my implementation to mesa-dev too. I see a few
major differences in our approaches:
- I'm using Autotools. You used Android.mk. I probably broke the
Android.mk build... sorry. I haven't found the time yet to test that
https://bugs.freedesktop.org/show_bug.cgi?id=102516
Bug ID: 102516
Summary: regression on glsl shaders on BE architecture
mesa-17.2.0-rc6
Product: Mesa
Version: git
Hardware: Other
OS: All
Status
NOT FOR UPSTREAM.
To get the driver's build-id, anv_physical_device_init_uuids() searches
the current process for an ELF phdr for filename "libvulkan_intel.so".
However, Android requires that the library be named
"vulkan.${board}.so".
Fix it with a quick #ifdef hack.
I have a plan to properly fi
This patch adds a flag param to anv_bo_cache_import() and defines
exactly one flag, ANV_BO_CACHE_IMPORT_NO_CLOSE_FD. If set, the function
will not close the fd.
This prepares for implementing VK_ANDROID_native_buffer, which must not
close the imported fd.
---
src/intel/vulkan/anv_allocator.c | 26
If ANV_BO_CACHE_IMPORT_IGNORE_SIZE_PARAM is set, then the function
ignores the 'size' parameter. Instead, on a cache hit, it uses the cached bo's
size; on a cache miss, it queries the fd's size.
This prepares for implementing VK_ANDROID_native_buffer, in which the
API provides no explicit size for
Will use in VK_ANDROID_native_buffer.
---
src/intel/vulkan/anv_gem.c | 16
src/intel/vulkan/anv_private.h | 1 +
2 files changed, 17 insertions(+)
diff --git a/src/intel/vulkan/anv_gem.c b/src/intel/vulkan/anv_gem.c
index 3994c6b66c2..34c09891086 100644
--- a/src/intel/vulka
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore
This helps when debugging a broken entrypoint table.
---
src/intel/vulkan/anv_entrypoints_gen.py | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/intel/vulkan/anv_entrypoints_gen.py
b/src/intel/vulkan/anv_entrypoints_gen.py
index 82221ef94bd..983be09a39b 100644
--- a/sr
I'm bringing up Vulkan in the Android container of Chrome OS (ARC++).
On Android, stdio goes to /dev/null. On Android, remote gdb is even more
painful than the usual remote gdb. On Android, nothing works like you
expect and debugging is hell. I need logging.
This patch introduces a small, simple
A first step to supporting Vulkan on ARC++. Mesa on ARC++ uses
Autotools, not Android.mk.
---
src/intel/Makefile.vulkan.am | 5 +
1 file changed, 5 insertions(+)
diff --git a/src/intel/Makefile.vulkan.am b/src/intel/Makefile.vulkan.am
index 8a19f96096a..d1b1132ed2e 100644
--- a/src/intel/Make
In src/intel/vulkan/*, redirect all instances of printf, vk_error,
anv_loge, anv_debug, anv_finishme, anv_perf_warn, anv_assert, and their
many variants to the new intel_log functions. I believe I caught them
all.
The other subdirs of src/intel are left for a future exercise.
---
src/intel/vulkan
This patch prevents compilation failures in upcoming Android Vulkan
patches, failures due to redefinition of LOG_TAG in Android system
headers.
This patch does not change the value of LOG_TAG. It remains
"INTEL-MESA". (I don't like it, though. The all-caps smells like
FORTRAN).
Only one Intel hea
This will reduce fragility in the implementation of
VK_ANDROID_native_buffer, which must do the equivalent of
vkAllocateMemory inside of vkCreateImage.
No behavioral change.
---
src/intel/vulkan/anv_blorp.c | 12 ++--
src/intel/vulkan/anv_image.c | 12 ++--
src/intel/v
Android's Vulkan loader implements VK_KHR_surface and VK_KHR_swapchain,
and applications cannot access the driver's implementation. Moreoever,
if the driver exposes the those extension strings, then tests
dEQP-VK.api.info.instance.extensions and dEQP-VK.api.info.device fail
due to the duplicated st
The code that restricts the VkImage's tiling flags, extract it into
a new function named choose_isl_tiling_flags(). This reduces the diff
in upcoming patches for VK_ANDROID_native_buffer.
---
src/intel/vulkan/anv_image.c | 31 ---
1 file changed, 20 insertions(+), 11 d
If this flag is set, then the image and it's device memory have the same
lifetime. vkDestroyImage will free its anv_device_memory.
We need this for VK_ANDROID_native_buffer, because that extension
creates the VkImage and imports its memory in the same call,
vkCreateImage.
---
src/intel/vulkan/an
Upcoming patches teach vkCreateImage to understand
VkNativeBufferANDROID. And much later patches will do the same for
whatever VkImageCreateInfo-chained struct imports dma-bufs. And then
again for the struct that will be introduced by the inevitable extension
VK_ANDROID_external_memory_whatever. So
Creation of hiz, ccs, and mcs surfaces was encoded in a large, deep 'if'
tree at the tail of make_surface(). This patch extracts that 'if' tree
into one function per aux type:
try_make_hiz_surface()
try_make_ccs_surface()
try_make_mcs_surface()
For clarity, also rename make_surface()
Feed the XML to anv_extensions.py and anv_entrypoints_gen.py.
Do it on all platforms, not just Android. Tested on Android and Fedora.
We always parse the Android XML, regardless of target platform, to
help reduce the chance that people working on non-Android break the
Android build.
---
src/intel
Tested on Android and Fedora.
---
src/vulkan/Android.mk | 9 +++--
src/vulkan/Makefile.am | 9 +++--
src/vulkan/util/gen_enum_to_str.py | 20 +++-
3 files changed, 33 insertions(+), 5 deletions(-)
diff --git a/src/vulkan/Android.mk b/src/vulkan/A
The taught scripts are anv_extensions.py and anv_entrypoints_gen.py. To
give a script multiple XML files, call it like so:
anv_extensions.py --xml a.xml --xml b.xml --xml c.xml ...
The scripts parse the XML files in the given order.
This will allow us to feed the scripts XML files for exten
The VK_ANDROID_native_buffer extension is missing from the official
vk.xml. This patch defines the extension in a separate, minimal XML
file: vk_android_native_buffer.xml.
I chose to add the extension to a new XML file instead of adding it to
the official vk.xml in order to avoid conflicts each ti
To give the script multiple XML files, call it like so:
gen_enum_to_str.py --xml a.xml --xml b.xml --xml c.xml ...
The script parses the XML files in the given order.
This will allow us to feed the script XML files for extensions that are
missing from the official vk.xml, such as VK_ANDROID_
This series adds Android support to Anvil. And Android requires
VK_ANDROID_native_buffer.
I tested the series on 64-bit ARC++ on a Skylake Chromebook with a 3.18
kernel. (ARC++ is the Android container in Chrome OS). (Yes, I said 3.18.
That's not a typo).
Here's my test results:
- A little,
Just as Mesa imports the Khronos Vulkan headers, it should import this
Android-private Vulkan header too. This guarantees that Mesa will
continue to build even when upstream Android breaks header
compatibility.
This header is only for *implementers* of Vulkan, not for consumers of
Vulkan.
Importe
This patch consolidates many potential `#ifdef ANDROID` messes
throughout src/vulkan and src/intel/vulkan into a simple, localized
hack. The hack is an `#ifdef ANDROID` in vk_android_native_buffer.h
that, on non-Android platorms, avoids including the Android platform
headers and typedefs any Androi
From: Eduardo Lima Mitev
Adds new INT16, UINT16 and FLOAT16 base types.
The corresponding GL types for half floats were reused from the
AMD_gpu_shader_half_float extension. The int16 and uint16 types come from
NV_gpu_shader_5 extension.
This adds the builtins and the lexer support.
To avoid a
54 matches
Mail list logo
