|
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Vol. II, No. 1 BDTI's DSP Insider January 2002
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
This month:
*** ARM9E Shows Surprising Speed
*** New BDTIsimMark2000(TM) Measures Pre-Silicon DSP Speed
*** Optimizing Streaming Media for Embedded Processors
*** "Impulse Response," a news analysis and opinion column written
by Jeff Bier, BDTI's General Manager, and featured in EE Times
*** BDTI Provides Software Expertise in Streaming Media
*** "Inside" Reports on 3DSP, ARM, and Hitachi/ST Processors
*** BDTImark2000(TM) for the Texas Instruments TMS320C55xx
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
*** ARM9E Shows Surprising Speed
Last month, BDTI published the first independent, in-depth report to
evaluate the ARM7, ARM9, and ARM9E processor cores. The ARM7 and ARM9
were not designed with the needs of DSP in mind, and these processors
have no DSP-oriented features. Not surprisingly, BDTI's analysis
shows that the ARM7 and ARM9 are not nearly as fast as mainstream DSP
processors; for example, the 160 MHz Texas Instruments TMS320C54xx is
about two times faster on DSP tasks than the 200 MHz ARM9.
The ARM9E—an ARM9 with DSP-oriented enhancements—offers
significantly better DSP performance than the ARM7 and ARM9.
According to BDTI's analysis, the 200 MHz ARM9E is about 70% faster
than the 200 MHz ARM9 on 16-bit DSP tasks and about 15% slower than
the 160 MHz TMS320C54xx. This level of performance may surprise
some—the ARM9E DSP enhancements are fairly modest compared to those
found in many other DSP-enhanced general-purpose processors.
Although the performance of the ARM9E is comparable to that of older
DSP-oriented processors, it is far slower than newer DSP-oriented
processors. According to BDTI's analysis, the successor to the
TMS320C54xx, the 200 MHz Texas Instruments TMS320C55xx, is over two
times faster than the 200 MHz ARM9E. This performance gap is largely
due to a difference in MAC throughput: nearly all recent DSP-oriented
architectures—including DSP-enhanced general-purpose processors—
support two (or more) MACs per cycle, but the ARM cores are limited to
only one multiply-accumulate per cycle. ARM seems to be aware of this
shortcoming: in October 2001 ARM revealed that its next-generation
core will include a dual-MAC SIMD operation.
Despite their modest speed, the ARM7 and ARM9 cores have already found
success in several DSP applications with modest computational
requirements. For example, many portable digital audio players rely
on ARM7 cores. ARM cores are also used in many applications that
currently require both a general-purpose processor and a DSP. With
its improved DSP performance, the ARM9E may eliminate the need for a
separate DSP in some of these applications.
Performance is only one of many factors that determine whether a core
will be successful. One of the key reasons for ARM's success is the
quality and quantity of the development tools and the software
available for ARM cores, which is in turn largely due to ARM's strong
track record of compatibility. In addition, ARM-based off-the-shelf
chips are available from a number of vendors. Thus, potential
licensees have the ability to develop initial system designs using
off-the-shelf chip-level products, then migrate to an SoC design when
volumes justify the additional costs. And unlike some licensable
cores, the ARM cores have been proven in silicon again and again,
making them a less risky choice than less established cores.
*** New BDTIsimMark2000(TM) Measures Pre-Silicon DSP Speed
Processor users often want a simplified way of comparing processors'
speeds in DSP applications. To this end, BDTI publishes the
BDTImark2000(TM), a single-number DSP speed metric based on BDTI's
suite of DSP benchmarks, the BDTI Benchmarks(TM). Because it is based
on realistic DSP algorithm kernel benchmarks, the BDTImark2000
characterizes a processor's signal processing speed far more
accurately than traditional simplified measures such as MIPS or
MFLOPS.
BDTI's policy is to verify its benchmarks on silicon before issuing a
BDTImark2000 score. This policy helps to ensure that the score
accurately reflects the performance that can be expected from actual
silicon rather than relying on simulator predictions. However,
performance information is sometimes needed prior to silicon
availability; for example, a system-on-a-chip (SoC) designer may need
to evaluate the performance of a licensable core before the core has
been fabricated. To meet such needs, BDTI now publishes the
BDTIsimMark2000(TM). This new metric is calculated in the same manner
as the BDTImark2000, but does not require hardware verification.
One key difference between the BDTImark2000 and the BDTIsimMark2000 is
the method used to select clock speeds. The BDTImark2000 is a score
based on silicon and as such is calculated using the clock speed of
the available silicon. For off-the-shelf chips, the BDTImark2000 is
calculated based on the clock speed of the fastest processor sampling
to customers; for cores, it is based on the highest clock speed that
has been demonstrated in silicon in either a development chip or an
SoC. In contrast, the BDTIsimMark2000 will use a projected clock
speed. For licensable cores, the BDTIsimMark2000 will use the
worst-case clock speed projected for the target fabrication process,
since this is the speed with which designers typically are most
concerned. Although BDTI will rely upon processor vendors to supply
the projected clock speed, it will evaluate the stated speed based on
a variety of factors (such as the target fabrication process) and will
adjust the projection as necessary to obtain what it believes to be a
credible clock speed.
Although BDTIsimMark2000 and BDTImark2000 scores are calculated in the
same manner, they should be compared with caution. In the time that
it takes for a pre-silicon processor to reach production, competing
vendors may achieve higher clock speeds or may introduce new
architectures with higher performance. In addition, a processor may
not perform as expected when it reaches production due to differences
between the projected and actual clock speed and/or differences
between the behavior of the silicon and that predicted by the
simulator.
The SP-5, a licensable core from 3DSP, is the first processor to
receive a BDTIsimMark2000 score. Its BDTIsimMark2000 score is 1720 at
225 MHz. (3DSP projects a typical clock speed of 320 MHz for a
0.13 um process; BDTI adjusted 3DSP's projection to obtain what it
believes to be a reasonable worst-case speed.) This score can be
compared with the BDTImark2000 score for the Carmel 10xx, which is
1390 at 188 MHz (188 MHz is the current top clock speed achieved in
silicon, according to Infineon). Based on their scores, the SP-5 at
225 MHz will be about 25% faster than the Carmel 10xx at 188 MHz;
however, the Carmel 10xx has had its BDTI Benchmarks verified on
silicon, while the SP-5 has not. Thus, the performance of the SP-5
indicated by its BDTIsimMark2000 score must be considered to be
somewhat less reliable than the result for the Carmel 10xx. (These
cores are analyzed in BDTI's reports, "Inside the 3DSP SP-5" and
"Inside the Infineon Carmel;" please visit
http://www.bdti.com/products/services_overview.htm#publications for
more information on these reports.)
BDTI will begin publishing BDTIsimMark2000 scores on its web site in
the coming months. BDTI's web site will also explain the assumptions
used to select the clock speeds for each of the BDTIsimMark2000
scores. For more information on the BDTImark2000, please visit
http://www.bdti.com/bdtimark/BDTImark2000.htm.
*** BDTI Case Study
This Month: Optimizing Streaming Media for Embedded Processors
Increasingly, embedded products include support for streaming media.
Examples of this trend include streaming audio support in palmtop
computers and wireless handsets and streaming audio and video support
in set-top boxes. Most of these embedded products have very limited
computational power and memory, and streaming media applications must
share these limited resources with other applications. Because
streaming media applications are resource-hungry, optimization is
usually required to meet performance goals, e.g., a MIPS usage target,
a memory usage target, or both.
Streaming media applications usually start as C/C++ code.
Optimization efforts typically focus on a few key functions that
consume the most processing time and/or memory; these functions are
re-written in hand-optimized assembly language, and the rest of the
code is left in C/C++. This approach works well because compilers
typically create code that is far slower and larger than the best
hand-assembled code. This is especially true for DSP-oriented code
such as the FFT and DCT functions found in most transform-based
multimedia compression algorithms.
Streaming media software can also be optimized by altering the
underlying algorithms. Algorithmic transformations, e.g., splitting
or combining processing steps, often lead to significantly faster
and/or smaller code. In some cases, it is possible to replace an
algorithm with a similar but more efficient algorithm. For example,
there are many different FFT algorithms; choosing one that better
matches a processor's capabilities can dramatically improve
performance.
Many optimization techniques involve tradeoffs between speed, code
size, and numerical precision; e.g., many speed-oriented optimizations
increase code size. Understanding these tradeoffs and striking an
appropriate balance requires expertise in both processor architectures
and DSP theory and a thorough understanding of the application.
Development tools often present additional challenges; for example,
compilers typically offer limited support for using processor features
like caches and DSP instruction-set enhancements.
Over the past several years, BDTI has created highly optimized
implementations of audio and video algorithms for a wide variety of
streaming media applications. BDTI's unique combination of expertise
in processor architectures, DSP theory, and software development gives
it a unique understanding of the tradeoffs involved in optimizing
streaming media applications. BDTI's specialized development
methodologies, which include proprietary floating-point to fixed-point
conversion tools and in-house profiling tools, help assure timely,
high-quality results.
BDTI has developed optimized streaming media software for a diverse
range of products, including an Internet radio, a portable digital
audio player, and a personal video recorder. For more information on
BDTI's past projects, please visit
http://www.bdti.com/products/services_software.htm. If you would like
to learn how BDTI can help you implement streaming media in your
embedded application, please contact Jeremy Giddings at BDTI
(giddings@BDTI.com).
*** Impulse Response, by Jeff Bier
This Month: Education Fuels Technological Competitiveness
In tough times, technology companies often curtail research and
development of new technologies and products. Too often, such
cutbacks are penny-wise and pound-foolish: while they reduce expenses
in the short term, they may also seriously damage a company's ability
to compete in the long term.
Sadly, this kind of shortsightedness can also be seen in the United
States' education funding policies. In response to shortfalls caused
by the current downturn, states are slashing education spending: in
December, the Florida Legislature passed a cutback of nearly $640
million, and California's governor has requested reductions of more
than $840 million. It is disturbing that, as a nation, one of our
first responses to a downturn is to cut education spending. This is
of particular concern because pre-college education in the U.S.
already falls short of that provided by many countries—U.S.
Department of Education statistics show that U.S. eighth graders
perform significantly worse in math and science than their
counterparts in countries like Korea and Belgium.
Have we already forgotten that the unprecedented U.S. economic
expansion of the 1990's was driven largely by American technological
innovations? Can we fail to make the connection between a
well-educated work force and the nation's economic prospects?
Consider the consumer electronics industry: at one time, the
U.S. played a major role in the design and manufacture of products
like televisions. Over time, other countries gained superior
technical expertise, and the U.S. role dwindled. In recent years,
though, digital signal processing has enabled something of a revival:
U.S. companies, based largely on their mastery of DSP, have regained a
significant role in designing—and in some cases manufacturing—
products like digital home audio equipment.
Make no mistake about it: U.S. engineers have been remarkably
successful in developing innovative technology not because we are any
smarter than our counterparts in other countries, but because we have
been provided with the appropriate tools—and education is primary
among these tools. If we are to maintain a vibrant national
technology industry, we must ensure that successive generations of
Americans are equipped to succeed in an increasingly competitive world
economy.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
*** Software Expertise in Streaming Media
BDTI provides its software development expertise to companies seeking
highly optimized, well documented, and efficient software. Our
particular skill is in the implementation of software for the support
of streaming media in embedded applications. We have expertise in:
- Audio and video codecs
- Player and viewer applications
- Pre- and post-processing issues (e.g., effects and conversion)
- Platform issues (e.g., getting the most out of your CPU)
- Implementation (e.g., float-to-fixed conversion, profiling, and
optimization)
- Integration (e.g., dealing with APIs, varying bitstream formats and
data types)
- Testing
- Documentation
- Packaging
Contact Jeremy Giddings (giddings@bdti.com) for more information.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
*** "Inside" Reports
BDTI's "Inside" reports combine performance data on the target
processors and key competitors with insightful analysis of
architectures, features, and tools. These reports are indispensible
to processor and systems designers, engineering and product marketing
managers, and others interested in evaluating technology.
Now shipping:
"Inside the 3DSP SP5"
"Inside the ARM ARM7, ARM9, and ARM9E"
"Inside the StarCore SC110"
"Inside the Hitachi SH-DSP and SH3-DSP"
"Inside the Analog Devices/Intel MSA"
Coming soon:
"Inside the Hitachi/STMicroelectronics SH-4/ST40 and SH-5/ST50"
For information on all of BDTI's technical reports, go to
www.BDTI.com/products/services_overview.htm#publications.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
*** BDTImark2000(TM) for the TI 'C55xx
BDTI has released a BDTImark2000 score for the Texas Instruments
TMS320C55xx.
For this score, as well as BDTImark2000 scores for other processors,
go to www.BDTI.com/bdtimark/BDTImark2000.htm
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
*** About BDTI
BDTI is an independent source for DSP technology analysis and
optimized DSP software. From rigorous technical analyses of
processors for DSP, such as the "Inside" series of processor
analyses, to highly regarded technology training classes, BDTI is the
trusted independent source for reliable information on DSP
technology. For more information, visit our Web site at www.BDTI.com.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
The next issue of BDTI's DSP Insider is coming in February. An
archive of previous issues of the BDTI DSP Insider will be available
in the future on BDTI's Web site. If you have comments, suggestions,
or other feedback about the DSP Insider, please send email to
dspinsider@bdti.com.
BDTI's DSP Insider is a free monthly electronic newsletter published
by Berkeley Design Technology, Inc. If our newsletter was forwarded
to you and you would like to receive it regularly, please register at
www.BDTI.com/dspinsider.htm.
If you no longer wish to receive the DSP Insider, send an email
message to dspinsider@bdti.com with the words "Remove me" in the
subject line.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
BDTI's DSP Insider (c) 2002 Berkeley Design Technology, Inc.
<< previous issue - next issue >>
|
