| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
| |
All *Pointer methods used 'normalized:=false', but we cannot assume
the fixed function code does use normalized (0f..1f) values.
On the contrary, it usually uses the native format value range.
Hence we have to pass normalized:=true for all fixed point data types
and normalized:=false for floating point data types.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
GL_POINT_SOFT and dist/fade attenuation (Adding basic POINT unit tests)
gl_PointSize (and all other uniform array elems) was not propagated due to wrong usage of GLUniformData component param.
Due to efficiency, we use vec4[2] now and #defines in shader to easy readability.
GL_POINT_SOFT uses gl_PointCoord to determnine inside/outside circle position
while adding a seam of 10% in/out. This almost matches 'other' implementations and gives a nice smooth circle.
!GL_POINT_SOFT produces a proper square (billboard).
Point-Vertex shader takes dist/fade attentuation into account.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
allowing passing a GL action w/ custom argument and return value.
Adding simple POINTS shader not regarding POINTS parameters and not using a texture (commented out).
FIXME:
Event thought it works using a texture and gl_PointCoord in frag shader,
I don't see the point here (lol) if gl_PointSize must be 1.0 in vert shader ..
otherwise nothing is seen on ES2.0.
On Desktop POINTS are always shown as 1 pixel sized points!
|
|
|
|
|
|
| |
twice (duh!) almost halfed performance :)
TODO: Create GL_POINT texture and render w/ glDraw*()
|
|
|
|
|
|
|
|
|
|
|
| |
ScreenMode rotated resolution.
ScreenMode Change Failover
- In case a timeout appears (buggy XRandR),
double check current ScreenMode in case it has been set.
Window.setFullscreen() shall use current ScreenMode rotated resolution.
- The Screen's virtual size in not correct!
|
|
|
|
|
|
| |
ES1 impl. detection
'glBegin' is not ES1, duh!
|
| |
|
|
|
|
|
|
| |
found in ES1 library
This is the case in BCM-VC-IV blobs, tested on Raspeberry-Pi
|
|
|
|
| |
Tested manual w/ TestGearsES2NEWT on Raspberry Pi
|
|
|
|
| |
discarding pixels of culled faces.
|
|
|
|
| |
according it's usage (update Mvi/Mvit only if lighting is being used)
|
| |
|
|
|
|
| |
resize element count
|
|
|
|
|
|
|
|
| |
and size gross-net > PAGE_SIZE
Usually PAGE_SIZE is written within one DMA xfer command,
so if the gross buffer bulk transfer contains more unused data than PAGE_SIZE
we may win when transfering each single buffer at buffer update.
|
|
|
|
| |
imm. gl* functions; Default color padding is 1f; Make fields private.
|
|
|
|
| |
glColor4f() more efficient, use pre-alloc NIO buffer
|
| |
|
| |
|
|\ |
|
| |
| |
| |
| |
| |
| | |
Missing format specifier for the first argument would lead to this throwing IllegalFormatException.
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It is impossible to use this method as it will get into an infinite loop as it
just calls itself. Base the implementation on the contains method shortly before
this method. As this method is impossible to actually use, it could also just
be removed.
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| |
| | |
The readlong() method is attempting to build a 64bit value from two 32 bit reads.
The problem is that shifting an int only uses the lower 5 bits of the shift value,
so << 32 is the same as << 0. Cast to long and restore the original intention.
Signed-off-by: Harvey Harrison <[email protected]>
|
|/
|
|
| |
properties, drawIndices QUAD w/ proper range and add uint; FixedFunctionHook: drawIndices QUAD w/ proper range and add uint
|
| |
|
|
|
|
| |
code: Remove precision for default precision types.
|
| |
|
|
|
|
| |
ShaderSelectionMode.AUTO (good for mobile); Lazy shader instantiation.
|
|
|
|
| |
becomes public method to JoglVersion
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
923d9dd7f1d40db72d35ca76a761ca14babf147f
We are aware that Google's ANGLE (Windows EGL/ES2 impl. based on D3D)
crashes using eglInitialize(..) w/ EGL_DEFAULT_DISPLAY.
Commit 923d9dd7f1d40db72d35ca76a761ca14babf147f moved the EGL device initialization
into the EGLDrawableFactory ctor and hence slipped out ANGLE workaround to disable it per default.
- Moving property static flags from GLProfile -> GLDrawableFactory
- Moving ANGLE workaround right into EGLDrawableFactory (where it belongs)
- Moving optional EGL/ES disable code to GLDrawableFactory (where it belongs)
Tested on Windows w/ Java-32bit and latest Chrome ANGLE DLLs
|
| |
|
|
|
|
| |
GL_QUAD_STRIP/GL_POLYGON/GL_QUADS mapping, glTexImage2D internalformat/format match.
|
|
|
|
| |
required (similar sanity checks as enableBuffer())
|
|
|
|
|
|
|
|
|
| |
detail w/ data sync within GLArrayHandle,
which also removed redundant code (VBO data sync and binding).
Refines commit 8582ece7dc7f65271b3184261697a542766d9864
and f49f8e22953ed2426fd4264ee407e2dc3fc07cfc
|
|
|
|
| |
combinations, plus texture usage)
|
|
|
|
|
|
|
|
|
| |
(fix, incomplete still), ShaderSelectionMode, Fix default values
Besides the above mentioned additional features towards completness of the FFP emu,
the ShaderSelectionMode allows fixating a shader program configuration,
i.e. AUTO switch (default) or choosing a static shader program to avoid heavy program switches
incl. uniform/attribute updates.
|
|
|
|
|
|
|
| |
issues UseProgram is enable==true, but no program switch.
More versatile toString(StringBuilder sb, boolean alsoUnlocated), by default don't dump uniforms/attributes
w/o valid location.
|
|
|
|
|
|
|
|
|
| |
is not more bound after enableBuffer(); Fix unit test (test VBO bound).
Explicit bindBuffer(..) is required now, since enableBuffer() doesn't leave it bound.
See fixed VBORegion* patch for use case, i.e. using a VBO index buffer for glDrawElements().
Complets commit 8582ece7dc7f65271b3184261697a542766d9864.
|
|
|
|
|
|
|
|
|
|
| |
VBO: Always unbind VBO ASAP after data transfer (glBufferData())
and assignment (glVertexPointer(..), glVertexAttribPointer()).
It's a bug to leave it bound .. due to redundancy
and other calls which could have change the VBO binding.
Removed syncData(..), now it's only issued at enable
and hence migrated into the enable method.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
and add API docs. (API Change)
- Changed create*(..) factory methods (API Change)
- Drop passing GL instance, not needed
- allows creation of ImmModeSink as final field w/o GL context
- Use 'glBufferUsage' to determine whether to use VBO or not ( 0 == glBufferUsage )
- Use glBufferUsage in glBufferData(..) call (oops)
- Toggle vboUsage per object ( 0 == glBufferUsage ? nonVBO : VBO )
remove static VBO usage flag
- Fix render mode
- GL_POLYGON -> GL_TRIANGLE_FAN (not GL_LINES)
- GL_QUADS -> Looped GL_TRIANGLE_FAN (is !GL2) in draw(..) w/ and w/o indices
- Buffer usage
- documented
- allow creating sink w/ all components (vertices, color, normal and texCoords)
bit render and grow only used parts.
This allows proper usage of sink where it is not known
which types are being used.
- Added test case
- Manually tested w/ Jake2 ES1
Jake2 uses the FFP immediate mode rendering, where we utilize this sink
w/o rendering artifacts.
|
|
|
|
| |
byte-order for sliced result
|
| |
|
|
|
|
| |
toString(..) from PMVMatrix to FloatUtil and make it more generic; GLUniformData toString() also dumps it's matrices.
|
|
|
|
|
|
|
|
|
|
|
| |
WINDOWS if not started from console
Sadly it appears that w/ the ATI Catalyst driver,
the TestInitConcurrent02NEWT w/ 16 concurrent threads brings down the JVM sometimes.
This results in a locked singleton thread and distrubs the whole unit tests session.
We are aware of the ATI Catalyst threading bug on Windows and Linux.
|
| |
|
|
|
|
|
|
| |
of commit 455fed40391afe10ce5ffb9146ca325af63b0a49
Add drawable null check before using.
|