home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Archive Magazine 1996
/
ARCHIVE_96.iso
/
discs
/
mag_discs
/
volume_9
/
issue_07
/
heardnet
/
StrongSpec
< prev
next >
Wrap
Text File
|
1996-02-28
|
8KB
|
205 lines
<Digital Equipment Corporation>
<Digital Semiconductor>
<SA-110 microprocessor>
Description
The Digital Semiconductor SA-110 microprocessor (SA-110) is the first
member of the StrongARM family of high-performance, low-power
microprocessors. The SA-110 is the latest implementation of Advanced
RISC Machines Ltd. (ARM) Version 4 architecture and offers significant
advances in microprocessor design. The SA-110 has been designed to
further extend the ARM family as the world's leading source of
low-power, high-performance RISC processors for embedded consumer
markets such as smart hand-held devices and interactive digital video.
The SA-110 is a general-purpose, 32-bit microprocessor with a 16KB
instruction cache (Icache); a 16KB, write-back data cache (Dcache); a
write buffer; and a memory-management unit (MMU) combined in a single
chip. The five-stage pipeline distributes tasks evenly over time to
remove bottlenecks, ensuring high throughput for the core logic. The
SA-110 offers high-level RISC performance, yet it ensures minimal
power consumption, making it ideal for portable, low-cost systems.
The SA-110 onchip MMU supports a conventional 2-level page-table
structure, with a number of extensions, which makes it ideal for
embedded control systems and object-oriented systems. These features
result in a high instruction throughput and impressive real-time
response for a small and cost-effective chip.
Features
* Internal clock nominal frequencies
- 100 MHz @ 1.65 V
- 160 MHz @ 1.65 V
- 200 MHz @ 2.0 V
* High-performance benchmarks (est.)
- 115 V2.1 Dhrystone MIPS (100)
- 185 V2.1 Dhrystone MIPS (160)
- 230 V2.1 Dhrystone MIPS (200)
* Onchip phase-locked loop (PLL) referenced from a
3.68-MHz external oscillator
* Five-stage pipeline sequences tasks evenly for
high performance
* MMU support for virtual memory systems
* Write buffer to improve system performance
* Onchip, 32-way, set-associative cache
- 16KB instruction cache
- 16KB write-back data cache
* Big and little endian operating modes
* Static operation with low power use
* 3.3 V I/O interface with independent programmable
bus clock
* Power-down modes: idle and sleep
* Fast interrupt response less than 1 ms for real-time
applications
* Excellent support for high-level languages, including
exceptional code density
* Support for IEEE 1149.1 test interface port
* Packaged in a 144-pin thin quad flat pack (TQFP)
Applications
* Smart hand-held devices
- Personal digital assistant (PDA)
- Next-generation cellular telephone
- Organizer
- Pager
* Video conference equipment
* High-performance real-time systems
- Disk drives
- Data communications equipment
* Interactive digital video
- Video games
- Internet browsers
- Set-tops
Microarchitecture
The SA-110 microprocessor is a high-performance implementation of
Advanced RISC Machine's ARM Version 4 architecture specification. The
following sections provide an overview of the chip's
microarchitecture and major functional units.
Bus interface logic
The bus interface logic, consisting of the control logic and the
address register, controls the bus interface and unplanned events
such as interrupts, resets, and aborts. The bus interface logic can
also enable or disable wrapping of read transactions and merging of
write transactions.
The bus interface can be configured to run synchronously or
asynchronously to the core logic. In synchronous mode, the bus
interface clock speed is the core clock rate divided by a programmable
integer value from 2 to 9 (maximum of 66 MHz). In asynchronous mode,
the bus interface clock, from 1 MHz to 66 MHz, is supplied by an
external source.
Write buffer
The SA-110 has an 8-entry write buffer with each entry able to contain
1 byte to 16 bytes. The write buffer can be enabled or disabled by
software. The write buffer is further controlled by a bit in the MMU
page tables; so the MMU must be enabled before the write buffer can
be used. Software can cause the write buffer to be flushed.
Core logic
The core logic fetches and executes instructions by using a five-stage
pipeline. The five stages are: fetch, decode, arithmetic logic unit
(ALU), cache, and write-back. This pipeline arrangement, using the
onchip ALU, distributes tasks evenly in time and so contributes to
the high performance of the core logic. The ARM architecture
supports 30 general-purpose registers, 1 program counter, and 6 status
registers. There are 16 general-purpose registers (including the PC
register) and 1 or 2 status registers visible at any one time. The
processor operating mode determines which registers are visible. The
core logic executes the ARM instruction set, which supports
straightforward assembly language code programming. It does not
depend upon sophisticated compilers to manage complicated instruction
interdependence. The instruction set has eight instruction classes:
* Two instruction classes use the onchip ALU, barrel shifter, and
multiplier to perform high-speed operations on the data in a bank
of 16 logical (31 physical) 32-bit registers.
* Three instruction classes control data transfer between memory and
the registers. The classes are optimized for flexible addressing,
rapid context switching, and swapping data.
* Two instruction classes control execution flow and execution privilege level.
* One instruction class accesses the privileged state of the SA-110.
Memory-management units
The SA-110 has two memory-management units: instruction (IMMU) and
data (DMMU). Separate translation lookaside buffers (TLBs) are
implemented for the instruction and data streams. The TLBs each have
32 entries that can each map a segment, a large page, or a small page.
The TLB entry replacement algorithm is round-robin. The data TLB
supports both the flush-all and the flush-single-entry function,
while the instruction TLB supports only the flush-all function.
Memory-management exceptions preserve the base address registers,
eliminating the need for "fix-up" code.
Cache
The SA-110 has a 16KB, 32-way, set-associative Icache with 32-byte
blocks and a 16KB, 32-way, set-associative, write-back Dcache with
32-byte blocks.
Instruction cache
The Icache supports the flush-all-entry function, and the replacement
algorithm is round-robin within a set. The Icache can be enabled or
disabled independent of the memory-management function. When memory
management is disabled, the Icache control logic considers all memory
to be cacheable.
Data cache
The write-back Dcache supports the flush-all-entry, flush-entry, and
copyback-en try functions. The copyback-all function is not provided
in hardware but can be provided by software. The Dcache entries are
allocated with read transactions and the entry replacement logic uses
a round-robin algorithm.
Clocks
The SA-110 receives a 3.68-MHz clock from a crystal-based clock
generator. The SA-110 uses an internal phase-locked loop (PLL) to
multiply the frequency by a variable multiplier to produce a
high-speed clock. The high-speed clock is then divided internally by a
configurable ratio to provide a system clock for synchronous
operation. The 3.68-MHz oscillator and PLL run constantly in normal
and idle mode.
Boundary-scan test logic
The SA-110 boundary-scan interface provides for driving and sampling
all the external pins of the device except np wrslp, irrespective of
the core logic state. This ability permits testing of:
* SA-110 electrical connections to the circuit board
* Integrity of connections between devices having a similar interface on the circuit board
Ordering Information
Part Frequency Order #
SA-110 100MHz 21281-BA
SA-110 160MHz 21281-AA
SA-110 200MHz 21281-CA
To Order Now:
For U.S./Canada, contact your local Manufacturer's Representative
Outside U.S./Canada, contact your local Distributor
For Additional Information:
* SA-110 Microprocessor Product Brief (Postscript format)
* To Order Additional Literature