Chia-I Wu [Mon, 2 Mar 2015 17:51:39 +0000 (10:51 -0700)]
intel: add intel_pipeline_init_default_sample_patterns()
Initialize sample patterns once, not every time a PSO is created.
Chia-I Wu [Thu, 5 Mar 2015 22:43:16 +0000 (15:43 -0700)]
intel: update genhw headers
Lots of renames and Gen8 support. Changes verified by "gcc -E" and diff to
have some confidence that I did not mess up.
Ian Elliott [Thu, 5 Mar 2015 22:45:47 +0000 (15:45 -0700)]
Win/glvdebug: Use a "\\" instead of "/" for Windows path separator.
Chia-I Wu [Thu, 5 Mar 2015 21:22:56 +0000 (14:22 -0700)]
nulldrv: build it on Linux
It compiles!
Chia-I Wu [Thu, 5 Mar 2015 21:26:54 +0000 (14:26 -0700)]
nulldrv: fix gcc errors/warnings
The only functional change should be that _aligned_malloc() is replaced by
malloc(). It does not seem to matter.
Chia-I Wu [Wed, 4 Mar 2015 18:19:59 +0000 (11:19 -0700)]
icd: remove stale function declarations
Remove declarations of
icdDbgRegisterMsgCallback()
icdDbgUnregisterMsgCallback()
icdDbgSetGlobalOption()
Chia-I Wu [Sun, 22 Feb 2015 06:41:03 +0000 (14:41 +0800)]
icd: remove unused icd-alloc.[ch]
Chia-I Wu [Sun, 22 Feb 2015 06:40:25 +0000 (14:40 +0800)]
nulldrv: remove icd-alloc.h dependency
Chia-I Wu [Sun, 22 Feb 2015 06:09:00 +0000 (14:09 +0800)]
intel: use intel_alloc()/intel_free() everywhere
Make intel_handle available in the functions that icd_alloc()/icd_free() is
still called, and convert them to use intel_alloc()/intel_free().
Chia-I Wu [Sun, 22 Feb 2015 05:49:15 +0000 (13:49 +0800)]
intel: use intel_alloc()/intel_free() when possible
There are still a few places that intel_handle is not readily available and we
are stuck with icd_alloc()/icd_free(). They will be fixed later.
Chia-I Wu [Mon, 19 Jan 2015 03:14:00 +0000 (11:14 +0800)]
intel: add intel_{alloc,free,logv,log}()
They are per-instance equivalents of icd_{alloc,free,logv,log}(). They also
require us to embed a pointer to icd_instance in intel_handle.
Chia-I Wu [Fri, 20 Feb 2015 20:41:57 +0000 (13:41 -0700)]
icd: add support for per-instance DRM enumeration
Add icd_instance to DRM enumeration functions, and use it for memory
allocation or logging.
Chia-I Wu [Sat, 21 Feb 2015 06:19:23 +0000 (14:19 +0800)]
intel: make intel_instance own an icd_instance
intel_instance is now correctly allocated by icd_instance.
Chia-I Wu [Sat, 21 Feb 2015 06:00:17 +0000 (14:00 +0800)]
icd: add generic icd_instance
It provides memory allocation and logging support.
Chia-I Wu [Sun, 22 Feb 2015 05:19:54 +0000 (13:19 +0800)]
intel: make intel_base_create() take a intel_handle
With this change, we now create intel_dev using
dev = intel_base_create(&gpu->handle, ...);
instead of
dev = intel_base_create(NULL, ...);
The calls to intel_dev_log() in base_dbg_copy_create_info() are removed. One
of them is replaced an assert() and the other should have been called in
intel_mem_alloc().
Ian Elliott [Thu, 5 Mar 2015 19:28:53 +0000 (12:28 -0700)]
Win: Fix compilation errors.
Ian Elliott [Thu, 5 Mar 2015 19:11:38 +0000 (12:11 -0700)]
Win/glvdebug: Need libGLESv2d.dll and libEGLd.dll to work on Window.
Ian Elliott [Wed, 4 Mar 2015 23:33:28 +0000 (16:33 -0700)]
Win/glvdebug: Eliminate dead code.
Ian Elliott [Wed, 4 Mar 2015 22:39:21 +0000 (15:39 -0700)]
Win/glvdebug: Changes needed to compile glvdebug on Windows.
Ian Elliott [Wed, 4 Mar 2015 20:48:59 +0000 (13:48 -0700)]
Win: Require CMake 3.0+, and look for Qt 5.3 in default location.
CMake 3.0+ has the ability to "find_package(Qt5". Thus, we will require at
least this version for Windows.
CMake still isn't smart enough to go looking for Qt 5.3, and another
modification is to look to see whether it's installed (on Windows) in the
default location. If so, set Qt5_DIR to point to it.
Tony Barbour [Wed, 25 Feb 2015 00:20:27 +0000 (17:20 -0700)]
tests: Remove Opengl / Glut dependencies
Tony Barbour [Wed, 25 Feb 2015 18:25:11 +0000 (11:25 -0700)]
tests: Use buffers format when writing PPM files
Also change tests to use BGRA by default
Also change WritePPM to use BGRA format
Tony Barbour [Wed, 25 Feb 2015 19:28:39 +0000 (12:28 -0700)]
tests: Use XGL to --show-images instead of OpenGL/Glut
Tony Barbour [Mon, 2 Mar 2015 23:38:52 +0000 (16:38 -0700)]
tests: Move environment object into sharable file
Courtney Goeltzenleuchter [Wed, 4 Mar 2015 22:47:34 +0000 (15:47 -0700)]
layers: Fix Windows build issue
Courtney Goeltzenleuchter [Wed, 4 Mar 2015 18:21:23 +0000 (11:21 -0700)]
object-tracker: Do not track GPU object
The GPU object is not something that is created by the app
and thus object tracker layer does not have to track it.
Mike Stroyan [Tue, 3 Mar 2015 23:54:24 +0000 (16:54 -0700)]
layers: allocate enough space for pointer sprintf
Jon Ashburn [Mon, 2 Mar 2015 23:51:38 +0000 (16:51 -0700)]
memTracker: Use local GetProcAddr rather than auto generated one
Clean up pass through functions. Generated GetProcAddr requires drawState
to implement all entrypoints so don't use this generated GPA.
Jon Ashburn [Mon, 2 Mar 2015 18:32:24 +0000 (11:32 -0700)]
drawState: Use local GetProcAddr rather than auto generated one
Clean up pass through functions. Generated GetProcAddr requires drawState
to implement all entrypoints so don't use this generated GPA.
Jon Ashburn [Tue, 3 Mar 2015 22:07:15 +0000 (15:07 -0700)]
layers: if no settings file or call overrides make callback override loging
Default debugAction in layers is to log based on variable init. Make the
overrides work as follows:
1) app calls xglRegisterMsgCallback()
2) settings file
3) Glave (or app) calls to setlayerOption()
Tobin Ehlis [Tue, 3 Mar 2015 20:58:42 +0000 (13:58 -0700)]
layers: Correctly set last bound PSO in DrawState layer
Tobin Ehlis [Tue, 3 Mar 2015 21:01:43 +0000 (14:01 -0700)]
glave: Fix from Peter to step by packet index
Jeremy Hayes [Mon, 2 Mar 2015 18:21:39 +0000 (11:21 -0700)]
glave: Fix svg resize behavior.
Ian Elliott [Tue, 3 Mar 2015 16:50:57 +0000 (09:50 -0700)]
Win: Cannot use the "dot" functionality yet on Windows.
Mark Lobodzinski [Tue, 3 Mar 2015 02:23:52 +0000 (20:23 -0600)]
layers: Made mem_tracker thread-safe
Brought up to par with DrawState and ObjectTracker.
Ian Elliott [Tue, 3 Mar 2015 00:49:02 +0000 (17:49 -0700)]
Fix "xcb_nvidia.c" for use of the nulldrv.
We're using "XGL_DRIVERS_PATH" for the env var.
Ian Elliott [Mon, 2 Mar 2015 21:05:18 +0000 (14:05 -0700)]
Turn on ability to trace on Windows with NVIDIA driver.
I can now trace the "tri" and "cube" demos. Two changes were needed:
1) Must link the programs against XGL.dll instead of XGL.lib (the "XGL" library
instead of "XGLstatic").
2) The glv_MessageStream_SetupHostSocket() function needed to "create" the
lock/critical section that's used in the Glave interconnect code.
Piers Daniell [Sun, 1 Mar 2015 16:32:56 +0000 (09:32 -0700)]
Add NOMINMAX and VC_EXTRALEAN to some windows files
This avoids these headers leaking conflicting Windows symbols to
non-Windows code.
Piers Daniell [Sun, 1 Mar 2015 14:50:14 +0000 (07:50 -0700)]
Improve compatibility of XCB_nvidia and Source2
Ian Elliott [Fri, 27 Feb 2015 18:14:56 +0000 (11:14 -0700)]
Win/xcb_nvidia: Remove compiler warnings with xcb_nvidia.
Ian Elliott [Fri, 27 Feb 2015 18:10:59 +0000 (11:10 -0700)]
Win/glave: Remove new compiler warnings on Windows.
Compiling the glvtrace_xgl64.dll library on Windows (see prior commit) resulted
in new warnings, some of which seemed serious (e.g. using "sizeof(void)", which
is 1 on Linux, with GCC, but is 0 on Windows).
Ian Elliott [Fri, 27 Feb 2015 00:01:30 +0000 (17:01 -0700)]
Win/glave: Compile/link the glvtrace_xgl64.dll library on Windows.
Ian Elliott [Thu, 26 Feb 2015 21:34:52 +0000 (14:34 -0700)]
Linux: Must use "__linux__" instead of "PLATFORM_LINUX".
There are some places (e.g. glave) where PLATFORM_LINUX is #define'd, but
normally we need to rely on "__linux__", which is #define'd by the compiler
(i.e. we can rely on it being there).
Ian Elliott [Thu, 26 Feb 2015 20:52:52 +0000 (13:52 -0700)]
xcb_nvidia: Hack in ability to use registry instead of getenv().
Trying to make it easy/quick to change between ICDs, which the registry allows.
Need to clean this up sometime.
Chia-I Wu [Mon, 23 Feb 2015 17:41:08 +0000 (10:41 -0700)]
demos/cube: use a single layout
Conflicts:
demos/cube.c
Piers Daniell [Tue, 24 Feb 2015 20:58:47 +0000 (13:58 -0700)]
Get the trace player to work with Windows
This change modifies many of the generators to use "#if
defined(PLATFORM_LINUX) || defined(XCB_NVIDIA)" to choose between using
XCB or WIN32 native.
Ian Elliott [Fri, 20 Feb 2015 21:12:18 +0000 (14:12 -0700)]
xcb: Deal with case of getenv() returning NULL.
At least on Windows, getenv() returns NULL when the environment variable is not
found. This causes a problem inside of "xcb_nvidia.cpp" when a C++ std::string
is assigned the return value of getenv().
Ian Elliott [Thu, 19 Feb 2015 21:26:19 +0000 (14:26 -0700)]
Modify xcb_nvidia/nulldrv to also support nulldrv.
This allows LunarG and others without the NVIDIA XGL driver to use the nulldrv
on Windows.
Piers Daniell [Mon, 23 Feb 2015 23:23:13 +0000 (16:23 -0700)]
Update tri and cube with stage
These changes were cherry-picked from nvidia-dev-
20150218 and
KhronosGroup/GDC-stable.
Jon Ashburn [Tue, 3 Mar 2015 00:12:31 +0000 (17:12 -0700)]
Revert "misc: Add xglRecordMemoryRange entrypoint and use it in Glave"
This reverts commit
fea3e5fc1ef2f3c9e4f06ab5a8d682f89b76377e.
Tobin Ehlis [Mon, 2 Mar 2015 23:09:15 +0000 (16:09 -0700)]
layers: Update last global DS in DrawState when UpdateDescriptors is called
Tobin Ehlis [Mon, 2 Mar 2015 17:16:40 +0000 (10:16 -0700)]
layers: DrawState Cmd Buffer tracking improvements
Add DbgMarker cmds to cmd buffer instruction linked list
Display last 10 touched CBs in dot file
Highlight last touched CB in dot file with green color
Jon Ashburn [Fri, 27 Feb 2015 19:53:41 +0000 (12:53 -0700)]
misc: Add xglRecordMemoryRange entrypoint and use it in Glave
Peter Lohrmann [Mon, 2 Mar 2015 17:41:29 +0000 (09:41 -0800)]
glvdebug: Disable the timeline when replay is in progress.
* Currently the timeline doesn't appear any different when it is disabled, but the user cannot interact with it, which is the intent.
Peter Lohrmann [Mon, 2 Mar 2015 17:36:28 +0000 (09:36 -0800)]
glvdebug_xgl: Draw State Diagram and Command Buffer Diagram are now only added or enabled if the associated SVG diagram has been generated.
* This should help prevent empty windows from appearing in the UI.
* Also fix a regression that was causing most of the UI to be enabled during replay, when it should have been disabled so that the user could not navigate between API Calls.
Peter Lohrmann [Mon, 2 Mar 2015 17:04:28 +0000 (09:04 -0800)]
glvdebug: Add method for qcontroller to enable/disable state viewer tabs.
* glvdebug_xgl now uses this to properly disable the Draw State Diagram and Cmd Buffer Diagram during replay, and
to re-enable them if an updated svg diagram is available after pausing.
Peter Lohrmann [Mon, 2 Mar 2015 16:29:23 +0000 (08:29 -0800)]
glvdebug_xgl: Improve performance of appling settings while a ProxyModel is being used.
* If the ProxyModel (currently either GroupByThreads or GroupByFrames) is not changed when the settings are applied,
then don't reset the ProxyModel's source model. This was causing the tree to get recreated unexpectedly, which can
take significant time on a large trace file.
Peter Lohrmann [Fri, 27 Feb 2015 23:35:15 +0000 (15:35 -0800)]
glave: Set layer extension func pointers to NULL when the device is destroyed.
* This prevents invalid memory access if the pointers are referenced after the device is destroyed and before a new one is created.
Courtney Goeltzenleuchter [Thu, 26 Feb 2015 18:40:39 +0000 (11:40 -0700)]
layers: Issue endl at end of output
Courtney Goeltzenleuchter [Thu, 26 Feb 2015 00:53:18 +0000 (17:53 -0700)]
demos: Clean up renderpass and framebuffer
Chia-I Wu [Fri, 27 Feb 2015 16:51:16 +0000 (09:51 -0700)]
intel: enable L3 cache
Set GEN7_MOCS_L3_ON everywhere.
Tobin Ehlis [Fri, 27 Feb 2015 15:48:43 +0000 (08:48 -0700)]
layers: In DrawState track last global pipe state and dump that to dot diagram
Peter Lohrmann [Fri, 27 Feb 2015 17:30:35 +0000 (09:30 -0800)]
glvdebug: Minor clean up of the connection types between the ReplayWorker and the ReplayWidget.
Peter Lohrmann [Fri, 27 Feb 2015 16:34:47 +0000 (08:34 -0800)]
Merge ReplayThread change with Detach Replay Window.
Jeremy Hayes [Wed, 25 Feb 2015 23:56:06 +0000 (16:56 -0700)]
glave: Add "Detach" checkbox.
Peter Lohrmann [Thu, 26 Feb 2015 18:51:36 +0000 (10:51 -0800)]
glvdebug: Move QReplayWorker into a separate thread.
* In order to achieve this correctly, the QController could not inherit from the QReplayWorker; instead the controller now has a worker object. The QReplayWidget puts the worker on a separate thread.
* Any interaction between the QReplayWorker and the rest of the UI must be done through Qt's signals and slots, which knows how to pass data between threads.
* Performance of replaying within the UI is now on-par with replaying using glvreplay64 from the command line, based on the cube demo. Increasing the validation level will slow down the replay preformance (as expected), and printing of validation messages will also slow down replay performance (as expected).
Mark Lobodzinski [Thu, 26 Feb 2015 21:18:57 +0000 (15:18 -0600)]
layers: Fix null pointer reference in MemTracker
Copy-paste error caused segfault in Cube demo, among others.
Jeremy Hayes [Thu, 26 Feb 2015 22:59:19 +0000 (15:59 -0700)]
glave: Add command buffer diagram.
Mark Lobodzinski [Fri, 27 Feb 2015 00:42:09 +0000 (18:42 -0600)]
layers: Remove spurious assert.
This was unnecessarily crashing the valve tests.
Mark Lobodzinski [Fri, 27 Feb 2015 00:33:10 +0000 (18:33 -0600)]
layers: Fix MemTracker memory leak
Call to string helper function was not freeing returned memory.
Tobin Ehlis [Thu, 26 Feb 2015 19:57:08 +0000 (12:57 -0700)]
layers: Correctly represent Descriptor-related arrays in DrawState dot file
Jon Ashburn [Thu, 26 Feb 2015 18:45:28 +0000 (11:45 -0700)]
glave: cleanup mem_info list as it shrinks during tracing
Chia-I Wu [Mon, 19 Jan 2015 03:14:00 +0000 (11:14 +0800)]
intel: add intel_handle
intel_handle is the base struct of any API handle, with its own helper
functions. Move magic from intel_base to the new intel_handle.
Chia-I Wu [Fri, 20 Feb 2015 19:50:12 +0000 (12:50 -0700)]
intel: add support for per-instance GPU enumeration
Every instance should be indepedenent and have its own list of intel_gpus.
Chia-I Wu [Fri, 20 Feb 2015 19:26:08 +0000 (12:26 -0700)]
intel: an XGL_INSTANCE is not an XGL_OBJECT
It should not embed an intel_obj nor be created by intel_base_create(). Note
that there is an issue with both the old and new code: the allocation of
intel_instance uses the default allocator instead of the app-provided
allocator. It will be fixed later.
Chia-I Wu [Fri, 20 Feb 2015 20:10:25 +0000 (13:10 -0700)]
intel: allow GPU re-enumeration
When apps see XGL_ERROR_DEVICE_LOST, they release everything and enumerate
GPUs again. The call to intel_gpu_remove_all() was in
xglInitAndEnumerateGpus(), but it was missing in the new xglEnumerateGpus().
Chia-I Wu [Fri, 20 Feb 2015 22:09:11 +0000 (15:09 -0700)]
intel: rename intel.c to instance.c
There is now XGL_INSTANCE for GPU enumeration and things happen before it.
Chia-I Wu [Fri, 20 Feb 2015 22:08:31 +0000 (15:08 -0700)]
intel: rename intel_gpa.c to gpa.c
Chia-I Wu [Fri, 20 Feb 2015 22:07:57 +0000 (15:07 -0700)]
intel: move xglEnumerateLayers() to gpu.c
We are going to rename intel.c to instance.c, and xglEnumerateLayers()
definitely does not belong to instance.c.
Mark Lobodzinski [Wed, 25 Feb 2015 21:14:06 +0000 (15:14 -0600)]
layer: Fixed options settings for generated layers, fixed typos
Handling of non-zero default values now also fixed.
Chia-I Wu [Wed, 25 Feb 2015 22:06:15 +0000 (15:06 -0700)]
intel: fix waiting of present fences
For presentation fences, we should wait until the server has submitted the
presentation command to the kernel _and_ the GPU has executed the command.
We failed to wait for the latter until this fix.
Mark Lobodzinski [Thu, 26 Feb 2015 00:11:05 +0000 (18:11 -0600)]
layers: Fix Memtracker fence-related memory leak
MemTracker created internal fence objects to locally track command buffer
completion, but these objects were never destroyed.
Ian Elliott [Thu, 26 Feb 2015 00:40:38 +0000 (17:40 -0700)]
Win: Fix compilation error on Windows (xgl_helper.py).
Visual Studio doesn't like the following code construct:
if (<some test>)
return <something>;
<type> variable [= <intial-value>];
We have to use curly-braces, as in:
if (<some test>) {
return <something>;
}
<type> variable [= <intial-value>];
Ian Elliott [Thu, 26 Feb 2015 00:34:46 +0000 (17:34 -0700)]
Win: layers (draw_state.c): Fix compilation error on Windows.
Visual Studio doesn't like:
static const uint32_t NUM_COLORS = 7;
char* edgeColors[NUM_COLORS];
Instead, we need to do:
char* edgeColors[NUM_COLORS];
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 23:58:34 +0000 (16:58 -0700)]
layers: Track WsiCreatePresentableImage image object
The memory returned by WsiGetPresentableImage is unique
and cannot be freed or unbound by the app. Thus only object tracker
only needs to track the image object.
Courtney Goeltzenleuchter [Thu, 26 Feb 2015 00:13:27 +0000 (17:13 -0700)]
layers: Do not track PHYSICAL_GPUS
The application does not create PHYSICAL_GPU objects
so do not track them.
Courtney Goeltzenleuchter [Thu, 26 Feb 2015 00:12:26 +0000 (17:12 -0700)]
layers: Do not track Queue objects
Queue objects are not created by an application and
thus never destroyed. Thus do not need to track.
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 23:56:00 +0000 (16:56 -0700)]
layers: flush logged output so it's visible immediately
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 23:55:23 +0000 (16:55 -0700)]
demos: Clean up objects after cmd buf built
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 22:13:35 +0000 (15:13 -0700)]
demos: Update to single layout
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 20:54:22 +0000 (13:54 -0700)]
intel: use buffer type memory for linear formats
To behave in a similar fashion to other IHVs we need
to return XGL_MEMORY_TYPE_BUFFER for images that
use format_class of LINEAR and XGL_MEMORY_TYPE_IMAGE for
all other format classes.
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 19:49:04 +0000 (12:49 -0700)]
demos: Be sure to initialize pNext
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 19:48:19 +0000 (12:48 -0700)]
demos: Bug fix: Unbind image object
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 18:48:28 +0000 (11:48 -0700)]
demos: Be sure to request correct memory type
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 18:46:58 +0000 (11:46 -0700)]
demos: Add use_staging option
Specifying --use_staging as a command line argument to tri
will force tri to upload the texture via a linear mapped staging buffer.
Courtney Goeltzenleuchter [Tue, 17 Feb 2015 19:54:31 +0000 (12:54 -0700)]
cube: Use staging buffer for texture upload
Courtney Goeltzenleuchter [Sat, 14 Feb 2015 01:20:24 +0000 (18:20 -0700)]
demos: Integrate feedback from NVIDIA
Courtney Goeltzenleuchter [Sat, 14 Feb 2015 00:52:46 +0000 (17:52 -0700)]
demos: Refactor to simplify and optimize
Will now check if texturing is supported on linear, CPU accessible
textures. If so, just uses that. If not, will use a staging buffer
to copy to optimized layout texture image.
Courtney Goeltzenleuchter [Wed, 25 Feb 2015 18:01:35 +0000 (11:01 -0700)]
intel: linear images require buffer type memory
Jon Ashburn [Wed, 25 Feb 2015 19:45:23 +0000 (12:45 -0700)]
layers: Check for NULL pointer with dynamic_display()