Texas
Instruments OMAP4470 CGPU Information:
What is composition?
Composition is the eye candy a user sees onscreen. Whether it
isfancy UIs, graphics accelerated web browsing, heart-pounding game play,
multi-tasking while video chatting and looking through photos, or watching the
latest Hollywoodmovie streamed directly to you, all these use cases require composition.Fast
composition processing leads to a more responsive device without lag, ��butter smooth��
scrolling, and instantaneous content updates. The end result is an experience
to remember with a ��WOW�� factor.
In a nutshell, composition takes all the visual pieces of
what the operating system and application requires to be shown on the display, processes
them, and merges them together to produce the final image.The intricate details
hidden from the user are performed millions of times per second using the
OMAP4470 CGPU.
What is a CGPU?
CGPU stands for Composition Graphics Processing Unit and it
is fast becoming astandard for composition processing. This technological
innovation has seen rapidindustry adoption in many of the leading mobile and
consumer devices, including smartphones, tablets, and 1080p/4K2K connected
HDTVs. The CGPU is ideal for any screen size and it automatically scales to different
resolutions, aspect ratios, multiple screens, and formats.
CGPU is a task specific, multi-faceted processing core that
not only accelerates and benefits composition. It is used to lower power (up to
10X), reduce bandwidth (over 50%), increase pixel processing performance (up to
4X), filter, scale, rotate, and dynamically adjust screen content.
New advancements in composition built around Google Android
(HWC), Microsoft Windows (DirectX 2D), and other operating systems have
features built into the OMAP4470 CGPU to accelerate the platform experience.
How is the CGPU used?
Inside a leading applications processor like the OMAP4470,
there are many important functional blocks like the CPU, 3D GPU, IPU,
network/connectivity, and others. The CGPU is the new addition to this group as
it becomes afundamental building block in a SoC/ASIC. With this new paradigm
shift to include a CGPU, Texas Instruments has become a pioneer as the next
wave of exciting and novel user interfaces are built around the OMAP4470.
New experiences built around composition can now be realized
with: Lower Power, Higher Performance, and No Feature Compromise
CGPU in
Detail:
In previous versions of products that supported composition,
this function was under the umbrella of either a (custom) 2D graphics processoror
the CPU. Those implementations used simple tasks like BitBLTs, line drawings
and ROP functions, but nothing beyond basic functions that were limited in
scope. Custom cores also had many variations which were not easy to support
from an OS or platform point of view since they were not always matched to the
OS release cadence or feature sets/API.The CPU method was not optimal in terms
of performance, power, and capabilities since CPUs were not designed for these
tasks.
In other implementations, the (3D)GPU was used in place of
the CGPU, but this proved too costly from a power (battery life), bandwidth,
and performance perspective. Due to the large gate count and die areaof the GPU,
this large block consumed more dynamic and leakage power. Startup time to set
states in the GPUalso takes more cycles to configure, which slows down
performance especially when composition is dynamic and always changing. Most of
the time is spent rendering non-3D objects, like UIs, web browsers, or apps,so
the GPU is unnecessarily being powered up to perform non-3D tasks.
Enter the CGPU:Tiny, fast,
low power, easy to integrate, unified HW/SW stack,upstreamable open source
user/kernel mode drivers, and most importantly, supports industry standards
from a feature and API standpoint. No more custom hardware, no more special
programming required, and no more special builds or patches required by
developers and ISVs.
The CGPU can offload processing from the CPU, which lowers
power and frees the CPU for other tasks. The GPU can either be completely shut
off to conserve battery life or be free to focus on creating the best jaw
dropping 3D content without having to worry about composition functions or not
having enough processing power to simultaneously perform 3D and composition.
Block
diagram:
Key Advantages
and Innovations:
Note: This is a subset of features supported by the TI
OMAP4470. Please contact your local TI representative or FAE for more details.
��
2D Functions:
o
Bit BLT
o
Stretch BLT
o
Rectangle fill, clear, and clipping
o
Line drawing
o
Filter BLT
o
Mono expansion for text rendering
o
ROP2, ROP3, and ROP4
o
Alpha blending, including Java 2 Porter-Duff
compositing blending rules
o
32K x 32K coordinate system
o
90/180/270/X0Flip/Y-Flip/MirrorRotation/Source
Rotation, including filtering
o
Transparency by monochrome mask, chroma key, or
pattern mask
o
UI effects including blurring, sharpening, animation,
transitions, and more
��
Programmable filtering:
o
9-Phase, 32-Tap
o 1x1
up to 9x9 selectable kernels
o Used
for image previews, PiP, zoom/re-sizing, scaling
o Convolution
effects for image blurring, softening/sharpening, mood effects, and others
o Image
processing for special effects like mosaic, aging, silhouettes, and others
��
HW accelerated multi-surface composition
��
Multi-Format support: including RGB, YUV, HDR and
swizzle formats
o
RGB and swizzle formats, including ARGB formats
o
YUV and swizzle formats (U/V swap)
o
HDR and swizzle formats
o
YUV-to-RGB color space conversion (BT.601/BT.709)
o
Planar/Non-Planar YUV conversion
��
Full support for Multi source blending:
o
Single pass multi-source blending
o
Variable block sizes
o
Reduces bandwidth
o
Reduces software overhead and management
��
Stereoscopic 3D support:
o Fast,
scalable pixel processing engine to support high bandwidth stereo 3D
o Pixel
interleaving on an pixel and sub-pixel level
o Helps
SoC/ASIC support various formats of 3D video, games, and 3D desktop/windowing
composition
��
Full Function MMU:
o
Variable page size support
o
Secure MMU and special features for security and
virtualization
��
Bus Interface:
o
AXI
o
AHB
o
Cache Coherency with ACE-Lite (Optional)
��
Zero and Low CPU Overhead:
o Minimize
CPU wait states
o Asynchronous
operation through DMA command stream
o CPU
does not need to wait for 2D engine to complete execution before changing
states or issuing new commands (CPU can work on other system tasks instead of
servicing GPU core)
o Command
batching/grouping to minimize CPU utilization and increase throughput
o Improves
system performance while running applications
o Improves
system performance by increasing battery life
CGPU
Performance:
CGPU
Software Support:
��
BLTSVille: Open source kernel.org user and kernel mode
drivers
��
Android (��Butter Smooth�� HWC, Regionizer)
��
Microsoft Windows (DirectX 2D/GDI+)
��
Linux (DirectFB, Wayland)
��
QNX
��
GreenHills
��
Other RTOS/OS platforms
Putting the
pieces together:
Please click on the links below to access the latest OMAP4470
information:
OMAP 4 Platform: OMAP4470 (Main Page)
Add
Link(TI
Composition Whitepaper)
CGPU:
��
Focus On Composition: Tiny, Low Power, Task-Specific Core
��
Scales to High Resolution, Color Depth, and Aspect
Ratio WithAlmost Zero Incremental Power
��
Offload GPU and CPU
��
Reduce Power / Conserve Battery
Life
��
Reduce System Bandwidth (Improve System Performance)
��
4X Performance Boost (Pixel Processing, Responsiveness,
Smooth, Instant Content Update)
��
50% - 90% Power Savings
3D GPU:
��
FocusOnly On 3D Content Creation (Not Composition)
��
Power off when not used (conserve battery life)
��
GPU + CGPU = The Ideal Solution
Partners:
