| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
offscreen backend
Add new GLJPanel method 'setSkipGLOrientationVerticalFlip(..)':
/**
* Set skipping {@link #isGLOriented()} based vertical flip,
* which usually is required by the offscreen backend,
* see details about <a href="#verticalFlip">vertical flip</a>
* and <a href="#fboGLSLVerticalFlip">FBO / GLSL vertical flip</a>.
* <p>
* If set to <code>true</code>, user needs to flip the OpenGL rendered scene
* <i>if {@link #isGLOriented()} == true</i>, e.g. via the PMV matrix.<br/>
* See constraints of {@link #isGLOriented()}.
* </p>
*/
public final void setSkipGLOrientationVerticalFlip(boolean v) {
GearsES2: Handles 'flipVerticalInGLOrientation'
Unit test 'TestPerf001GLJPanelInit02AWT' validates and measures performance.
|
| |
|
|
|
|
| |
GLSharedContextSetter (remove <> in links, more details on Immersion.16.
|
|
|
|
|
|
|
|
|
|
|
|
| |
simple 'synchronized' on Animator for field-get, which is already used in most methods
Utilizing a 2nd synchronization object 'stateSync' besides the main sync object, Animator itself,
is hard to maintain. It's performance advantages for querying states ae questionable and may even
introduce bugs.
Use synchronization on Animator instance for all field read/write access.
Fix unsynchronized write access of 'animThread' in Animator.MainLoop.run().
|
|
|
|
|
|
|
|
|
| |
'GL_OES_EGL_image_external' not supported
+ // Bug on Nexus 10, ES3 - Android 4.3, where
+ // GL_OES_EGL_image_external extension directive leads to a failure _with_ '#version 300 es' !
+ // P0003: Extension 'GL_OES_EGL_image_external' not supported
+ preludeGLSLVersion = false;
|
|
|
|
| |
passed, this allows native GLCaps/config query to follow fast-path
|
|
|
|
|
|
|
|
| |
(!skipCapsChooser mode)
Regression of commit cf1163fc88976e7087d3a17524a49139e35a4708:
Commit dropped seeking recommendedIndex of chosenPFDID within cleaned-up availableCaps
when in !skipCapsChooser mode.
|
| |
|
|
|
|
| |
initializeBackendImpl(), already tested before function entry!
|
|
|
|
| |
in !blocking mode, i.e. notifyAll() if waitCondition holds and test again
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Validate isGLES*() usage and definition ; Add and use ShaderCode.createExtensionDirective(..)
- Fix GLES3 Profile Mapping, i.e. GL2ES2 queries and mappings
- GLProfile: Add GL2ES2 -> ES3 mapping
- EGLContext: Reuqest major '3' for ES3
- EGLGLCapabilities/EGLGraphicsConfiguration: Consider EGLExt.EGL_OPENGL_ES3_BIT_KHR
- Validate isGLES*() usage and definition
- Fix BuildComposablePipeline's isGLES() code
- For GLSL related queries use isGLES() instead of isGLES2(),
which would exclude ES3
- Add and use ShaderCode.createExtensionDirective(..)
- Supporting creating GLSL extension directives while reusing strings from GLExtensions
- Minor cleanup of GLContextImpl.setGLFuncAvail(..)
|
|
|
|
|
|
| |
user to trigger backend initialization eagerly and offthread (optional, !WINDOWS)
TestPerf001GLJPanelInit02AWT compares all variations: no-gl, glcanvas, gljpanel and gljpanel-initMT (offthread)
|
|
|
|
| |
device does _not_ require locking on global shared device.
|
| |
|
|
|
|
|
|
|
|
|
| |
Even though the test case itself cannot show the proper initialization time,
it can be used w/ an attached profiler i.e.
Test w/ 50 X11GLXGraphicsConfigurationFactory.chooseGraphicsConfigurationFBConfig() invocations:
- pre change: 1.708 ms
- post change: 650 ms (613e33ee8ffc1f2b9c5db1e1b5bb5253a159ed6d)
|
|
|
|
|
|
|
|
| |
unified gl- and glx-extension strings.
- HashMap is more efficient than HashSet
- No need to use sub HashSet's .. we can use global HashMap.
'Loosing' duplicate GLX entries due to GL duplicates is acceptable.
|
|
|
|
|
|
|
| |
613e33ee8ffc1f2b9c5db1e1b5bb5253a159ed6d to EGL and WGL.
Note: WGL config query is already performed as a bulk operation.
Note: OSX does not perform such queries.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
X11GLXGraphicsConfiguration.GLXFBConfig2GLCapabilities(..)
X11GLXGraphicsConfiguration.GLXFBConfig2GLCapabilities(..) ran over all FB configs and for each it grabbed
native config values separately. Fetching them in bulk mode saves around 7% of this function's cost.
Also reuse XRenderPictFormat instance for 'XRenderDirectFormat XRenderFindVisualFormat(..)' call,
saving a few NIO creation cycles w/ StructAccessor.
Biggest savior is X11GLXGraphicsConfigurationFactory.chooseGraphicsConfigurationFBConfig()'s
fast path w/o chooser and usable 1st FBConfig. Here we only issue 'GLXFBConfig2GLCapabilities(..)'
on the first valid entry.
Test w/ 50 X11GLXGraphicsConfigurationFactory.chooseGraphicsConfigurationFBConfig() invocations:
- pre change: 1.708 ms
- post change: 650 ms
Time is no spent almost solely on native glXChooseFBConfig (546ms).
|
|
|
|
| |
and let it fail there instead of within impl. class, only pass the handle - simplifies and removes redundancy.
|
|
|
|
|
|
|
| |
'Hisilicon Immersion.16')
Note: Even though Mesa Intel driver crashes w/ heavy multithreading (Bug 873),
it works well w/ our multithreaded GLMediaPlayer.
|
|
|
|
|
|
|
|
|
|
|
|
| |
multiple media textures (Android) or shared GL context are not usable.
- GLMediaPlayer:
- TEXTURE_COUNT_MIN is the new minimum: '1' - i.e. no multithreading, single threaded player
- TEXTURE_COUNT_DEFAULT is '4' - multithreaded
- GLMediaPlayerImpl:
- Add Single threaded mode, but perform initStreamImpl(..) off-thread.
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(spec doesn't mention it, and it does not show results w/ CPU sourced rendering) ; Clean up GLBuffer*Tracker
+ * Note that VAO initialization does unbind the VBO .. since otherwise they are still bound
+ * and the CPU_SRC test will fail!<br/>
+ * The OpenGL spec does not mention that unbinding a VAO will also unbind the bound VBOs
+ * during their setup.<br/>
+ * Local tests here on NV and AMD proprietary driver resulted in <i>no ourput image</i>
+ * when not unbinding said VBOs before the CPU_SRC tests.<br/>
+ * Hence Bug 692 Comment 5 is invalid, i.e. <https://jogamp.org/bugzilla/show_bug.cgi?id=692#c5>,
+ * and we should throw an exception to give users a hint!
Leaving uncommented code in GLBufferStateTracker ..
+++
- Clean up GLBuffer*Tracker
- Use final
- Use static final keyNotFound value.
|
|
|
|
|
|
|
| |
use EGLDummyUpstreamSurfaceHook .. a pbuffer offscreen
fixOffscreenBitOnly(..) may yield FBO queries / chosing .. usually doesn't matter (on devices here),
but not accurate.
|
|
|
|
| |
GLCapabilitiesChooser instead of GLProfile, allowing using same or similar caps - important for sharing ctx
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
reduce footprint for the masses.
Remove the ubuntu fonts from atomic/jogl-util-graph.jar and hence all derivated 'all' JAR files.
The Android jar files still contain the fonts as assets!
atomic/jogl-util-graph-fonts-p0.jar contains the fonts and is either referenced by:
- UbuntuFontLoader: Using class based Jar URI derivation using TempJarCache to [down]load
and extract the jar file (similar to native lib-loading).
- Explicitly via traditional classpath, see jnlp-files/jogl-applet-runner-newt-GraphTextDemo01b-napplet.html
The pack200 jogl-all.jar file is now below 1MB
|
|
|
|
| |
point values. Derived from http://mvn.io7m.com/ieee754b16, of which I am the original author.
|
|
|
|
|
|
|
|
|
|
|
|
| |
strictMatch minor) and reuse isES; EGLContext: Use strictMatch for setGLFunctionAvailability() and handle failure; EGLDrawableFactory: Either detect ES3 or ES2.
- Cleanup branch using VersionNumber (fix 'Int' check strictMatch minor) and reuse isES
- EGLContext: Use strictMatch for setGLFunctionAvailability() and handle failure
On ES, we require strictMatch, cleanup if failing.
- EGLDrawableFactory: Either detect ES3 or ES2.
Both only available with proper EGL context creation for ES profiles (TODO)
|
|
|
|
| |
operation, throw InternalError if failing
|
|\
| |
| | |
Fix #875 - ES version should be strictly validated
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When initializing the context in GLContextImpl.setGLFuncAvailability
ES devices must be validated by strictly matching the major version,
otherwise on ES3 devices we were mixing ES1 implementation with ES3
contexts, ultimately crashing in a safeguard.
Signed-off-by: Brice Figureau <[email protected]>
|
| |
| |
| |
| | |
to lifecycle considerations ; GearsES2: Add glFinish() after init().
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
stability constraints' ; Fixing Test cases: Enable all, GearsObject*: Check VBO
- GLSharedContextSetter API Doc: No 'Driver stability constraints'
- No driver issues ..
- Use 'Lifecycle Considerations' to describe usage issues ..
- Fixing Test cases: Enable all, GearsObject*: Check VBO
- GearsObject* needs to check whether VBO is 'still alive'
if sharing is enabled.
- Enable all unit tests.
|
| |
| |
| |
| |
| |
| | |
instance of GLArrayHandler of same type; Simplify GLArrayHandler inheritance.
Refines commit 9f2a9df0a4b7093925c8854b37fba053469a4b35
|
| |
| |
| |
| | |
field usage; GLDrawableImpl: Make read-only fields final.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Refine GearsObject* GLArrayDataServer copying; GearsES*: Init VBO eagerly
Add copy-ctor to GLArrayData* w/ sliced Buffer to allow general sharing of VBO via these high-level types.
Refine GearsObject* GLArrayDataServer copying (commit bcfaa149b9803ce33c5a356cbcb45f7dfd3e4361):
Utilize new GLArrayData* copy-ctor.
GearsES*: Init VBO eagerly, allowing VBO usage after init(..).
|
| |
| |
| |
| |
| |
| |
| | |
convenience
Added:
GLAutoDrawable createDummyAutoDrawable(AbstractGraphicsDevice deviceReq, boolean createNewDevice, GLProfile glp)
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
downstream.getGL*() [TraceGL* / DebugGL*] ; Simplify GLContextImpl's set Debug/Trace Pipeline
Regression of commit 0002fccdcd6383874b2813dc6bbe3e33f5f00924:
"Trace/Debug shall utilize downstream identification for isGL*() and getGL*() methods."
Using the downstream identification commit is right for the isGL*() case,
however, getGL*() returned the downstream object which makes the caller loosing the pipeline!
Instead, we shall produce !GEN_GL_IDENTITY_BY_ASSIGNABLE_CLASS:
"if( isGL<type>() ) { return this; }
throw new GLException("Not a <type> implementation");"
or for GEN_GL_IDENTITY_BY_ASSIGNABLE_CLASS:
"return this;"
|
| |
| |
| |
| | |
glViewport(..) and dump it (Add stack trace if DEBUG)
|
|/ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
order) ; Fix unit tests (Shared Gears, wait for created context and destruction order)
- Add note about driver stability (destruction order)
- See GLSharedContextSetter: Don't destroy master context before slaves!
- Fix spec-overview.html#SHARED links, add link to GLSharedContextSetter in SHARED subsection.
- Fix unit tests (Shared Gears, wait for created context and destruction order)
- The GearsObject sharing was completly bogus!
It simply used the _same_ GLArrayDataServer instance for sharing,
but it should use a _copy_ of the shared GLArrayDataServer while
only preserving the VBO object!
Fixed, while adding required methods to GLArrayDataServer.
- Waiting for the created GLContext of a GLAutoDrawable required us
to pass the latter _and_ check whether it's GLContext exists and is natively created.
- Accomodated the destruction order - see above!
|
| |
|
|
|
|
|
|
| |
local shared ctx fields.
Regression of 7f7a23dd0ddf106e6f0c69fc2a05ff92ac56200e
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
; Fix GLContext memory contract (volatile)
(Unit test remarks see below)
- Add shared GLContext queries
- Refined GLContextShareSet:
- Use IdentityHashMap since GLContext's can only be identical w/ same reference (footprint, performance)
- Add API doc for clarification
- Add methods:
- ArrayList<GLContext> getCreatedShares(final GLContext context)
- ArrayList<GLContext> getDestroyedShares(final GLContext context)
- Use 'final' where possible
- Add GLContext methods:
- boolean isShared()
- List<GLContext> getCreatedShares()
- List<GLContext> getDestroyedShares()
- Add GLSharedContextSetter interface defining setting a shared GLContext
directly (GLContext) or via a GLAutoDrawable:
- setSharedContext(GLContext)
- setSharedAutoDrawable(GLAutoDrawable)
Both cause initialization/creation of GLAutoDrawable's drawable/context to be postponed,
if the shared GLContext is not yet created natively or
the shared GLAutoDrawable's GLContext does not yet exist.
Most of impl. resides in GLDrawableHelper
Implemented in:
- GLAutoDrawableBase, GLOffscreenAutoDrawable
- GLWindow
- AWT GLCanvas
TODO:
- GLJPanel
- SWT GLCanvas
- GLDrawableFactory:
- Add 'GLOffscreenAutoDrawable createOffscreenAutoDrawable(..)' variant w/o passing the
optional shared GLContext _and_ specifying lazy GLContext
creation. This allows to benefit from GLSharedContextSetter contract.
Lazy GLContext creation is performed at 2st display() call at the latest.
All JOGL code and unit tests use this new method now.
- Mark 'createOffscreenAutoDrawable(..)' w/ shared GLContext argument
and immediate GLContext creation deprecated - shall be removed in 2.2.0
- Make reference to GLContext and it's native handle volatile
Since we rely on the query 'GLContext.isCreated()' to properly allow GLAutoDrawable's to query whether
a shared GLContext is natively created (already), the handle must be volatile
since such query and the actual creation may operate on different threads.
+++++
- Add/Refine shared GLContext unit tests demonstrating diff. sharing methods.
All variants of using shared GLContext:
com.jogamp.opengl.test.junit.jogl.acore.TestSharedContextVBO*
Most convenient way to share via setSharedAutoDrawable(GLAutoDrawable):
com.jogamp.opengl.test.junit.jogl.acore.TestSharedContextVBOES2[NEWT|AWT]3
AWT use w/ JTabbedPane using setSharedAutoDrawable(GLAutoDrawable):
com.jogamp.opengl.test.junit.jogl.acore.TestSharedContextWithJTabbedPaneAWT
|
|
|
|
| |
GLDrawable (Shared Context)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
GLContext:
- Proper API doc for Version* fields
- getStaticGLSLVersionNumber(): ES3 -> Version300
- hasGLSL(): Add ES3
ShaderCode:
- addDefaultShaderPrecision():
- ES2 default precision: Don't 'tune up' default precision for fragment shader, use 'mediump'
- Add ES3 default precision (equal to ES2 default precision)
- requiresDefaultPrecision(): Shall returns 'true' for ES3 as well!
|
|
|
|
| |
GLContextImpl.createContextARBVersions(..) erroneous upper bounds check.
|
|\ |
|
| |
| |
| |
| | |
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| | |
Signed-off-by: Harvey Harrison <[email protected]>
|