QCOM_binning_control

GL_QCOM_binning_control

Maurice Ribble

Maurice Ribble (mribble 'at' qualcomm.com)

Copyright Qualcomm 2010

Qualcomm Proprietary.

Draft

Date: May 2, 2012

OpenGL ES Extension #119

OpenGL ES 1.0 is required.

This extension is written against the OpenGL ES 2.0 specification.

This extension adds some new hints to give more control to application
developers over the driver's binning algorithm.

Only change this state right before changing rendertargets or right after
a swap or there will be a large performance penalty.

None

None

Accepted by the <target> parameter of Hint:

BINNING_CONTROL_HINT_QCOM           0x8FB0

Accepted by the <hint> parameter of Hint:
CPU_OPTIMIZED_QCOM                  0x8FB1
GPU_OPTIMIZED_QCOM                  0x8FB2
RENDER_DIRECT_TO_FRAMEBUFFER_QCOM   0x8FB3
DONT_CARE                           0x1100

Replace "target is a symoblic...." with:

target is a symbolic constant indicating the behavior to be controlled, and
hint is a symbolic constant indicating what type of behavior is desired. 
target can be GENERATE_MIPMAP_HINT or BINNING_CONTROL_HINT_QCOM.
GENERATE_MIPMAP_HINT  indicates the desired quality and performance of
mipmap level generation with GenerateMipmap. When target is
GENERATE_MIPMAP_HINT hint must be one of FASTEST, indicating that the most
efficient option should be chosen; NICEST, indicating that the highest
quality option should be chosen; and DONT_CARE, indicating no preference in
the matter.  A target of BINNING_CONTROL_HINT_QCOM gives hints at what 
binning algorithm is to be used.  When the target is BINNING_CONTROL_QCOM 
the hint must be one of the values below:

CPU_OPTIMIZED_QCOM                - binning algorithm focuses on lower CPU
                                    utilization (this path increases vertex
                                    processing)
GPU_OPTIMIZED_QCOM                - binning algorithm focuses on lower GPU
                                    utilization (this path increases CPU
                                    usage)
RENDER_DIRECT_TO_FRAMEBUFFER_QCOM - render directly to the final 
                                    framebuffer and bypass tile memory 
                                    (this path has a low CPU usage, but
                                    in some cases uses more memory 
                                    bandwidth)
DONT_CARE                         - the driver picks which binning 
                                    algorithm to use (default)

The Qualcomm Adreno 200 family does not support 
RENDER_DIRECT_TO_FRAMEBUFFER_QCOM option and this hint will be ignored on that
hardware.

When BINNING_CONTROL_QCOM do so right before changing rendertargets or right after
swap or there will be a large performance penalty.

6/15/2010  Created.
6/22/2010  Name changes, cleanup, add token numbers.
10/11/2010 Simplified extension to remove some of the extra modes.
1/27/2012  Added in RENDER_DIRECT_TO_FRAMEBUFFER_QCOM
5/2/2012   Added IP Status