[Dev] 2D/3D drivers

[criminal]

Hi,

I've succesfully loaded intercept drivers into GB:

I/SurfaceFlinger( 1996): EGL informations:
I/SurfaceFlinger( 1996): # of configs : 23
I/SurfaceFlinger( 1996): vendor    : Android
I/SurfaceFlinger( 1996): version   : 1.4 Android META-EGL
I/SurfaceFlinger( 1996): extensions: EGL_KHR_image EGL_KHR_image_base EGL_KHR_image_pixmap EGL_KHR_gl_texture_2D_image EGL_KHR_gl_texture_cubemap_image EGL_KHR_gl_renderbuffer_image EGL_ANDROID_image_native_buffer EGL_ANDROID_swap_rectangle 
I/SurfaceFlinger( 1996): Client API: OpenGL ES
I/SurfaceFlinger( 1996): EGLSurface: 5-6-5-0, config=0x0
I/SurfaceFlinger( 1996): OpenGL informations:
I/SurfaceFlinger( 1996): vendor    : Samsung Electronics
I/SurfaceFlinger( 1996): renderer  : FIMG-3DSE v1.5
I/SurfaceFlinger( 1996): version   : OpenGL ES-CM 1.1
I/SurfaceFlinger( 1996): extensions: GL_OES_compressed_paletted_texture GL_OES_point_size_array GL_OES_point_sprite GL_OES_blend_equation_separate GL_OES_blend_func_separate GL_OES_blend_subtract GL_OES_depth24 GL_OES_draw_texture GL_OES_framebuffer_object	GL_OES_mapbuffer GL_OES_matrix_get GL_OES_matrix_palette GL_OES_query_matrix GL_OES_rgb8_rgba8 GL_OES_stencil8 GL_OES_stencil_wrap GL_OES_texture_mirrored_repeat GL_OES_texture_npot GL_OES_read_format GL_EXT_discard_framebuffer GL_EXT_texture_format_BGRA8888 GL_OES_byte_coordinates GL_OES_element_index_uint GL_APPLE_texture_2D_limited_npot GL_EXT_texture_compression_s3tc
I/SurfaceFlinger( 1996): GL_MAX_TEXTURE_SIZE = 4096
I/SurfaceFlinger( 1996): GL_MAX_VIEWPORT_DIMS = 4096
I/SurfaceFlinger( 1996): flags = 00090000

It could work, BUT
Intercept (and spica & acclaim) drivers has no GL_OES_EGL_image extension, so blank screen effect
So, We can wait for tom3q or, we need a hack to GB without GL_OES_EGL_image...

Ideas?

[LeshaK]

I have already written, the driver of the i5700 and Intercept are loaded successfully but not working (black screen)

[tom3q]

in progress

Yep, it's just as little as I could do to see anything on the screen.
Proper implementation needs extending memory and shader subsystems.
Unfortunately I've got the end of semester on the Uni now, which means pretty lot of work with tests and projects and then exams, so it will take a while until I find some time to do it.

[tom3q]

I've done some changes to allow Gingerbread to work a bit better, so no more red screen, colors are ok and the performance is pretty reasonable.

You can find my development tree (with the changes) here:
https://github.com/tom3q/gles6410/tree/devel

To make everything work correctly, make sure that /dev/s3c-g3d and /dev/pmem_gpu1 have proper permissions (0666).

[LeshaK]

BIG THANKS!

[LeshaK]

I've done some changes to allow Gingerbread to work a bit better, so no more red screen, colors are ok and the performance is pretty reasonable.

You can find my development tree (with the changes) here:
https://github.com/tom3q/gles6410/tree/devel

To make everything work correctly, make sure that /dev/s3c-g3d and /dev/pmem_gpu1 have proper permissions (0666).

I built your driver (+module +kernel patches), but there is a sleep bug

[tom3q]

I built your driver (+module +kernel patches), but there is a sleep bug

Unfortunately, it is independent of the graphic subsystem, so I've got no control of it... (If you don't have latest s3c_g3d module from devel branch, there may be an additional sleep bug introduced by domain gating, worked around in development version by disabling it.)
After the exam session I will try to debug both of them.

[LeshaK]

If you don't have latest s3c_g3d module from devel branch

I have the latest g3d from dev branch

[tom3q]

OK. Gingerbread works on my driver without display corruption. Performance varies, but it's on acceptable level all the time.

Code available in master branch of my GitHub repo.

[LeshaK]

Sleep bug?