This document deals only with SPARC based computers, in order to check, just type
uname -m command and you should read something like
sparc4x where x is blank,c,d,m,u if the system runs Solaris, or
sparc for 32 bits SPARC architectures and
sparc64 for 64 bits SPARC architectures if it runs Linux. 2.x.y
SPARC stands for Scalable Processor ARChitecture, it derives from research done between 1984-1988 on the RISC architecture at UC Berkeley. It exists 3 versions of this archiecture, SPARC-V7, SPARC-V8 (32 bits) and SPARC-V9 (64 bits). As you are likely to encounter a lot of implementations of the SPARC architecture, in the next section, the main features of theses processors are summarized.
This is quite obsolete, it is an implementation of the SPARC-V7 ; its main feature are an Integer unit (IU), an external Floating Point Unit (FPU), an unified data + instruction 64KB direct associative cache, and an Memory Managment Unit (MMU). There is a 4 stage pipeline for the integer instructions (fetch F, decode D, exec E, update WB) FPU and IU are synchronized.
This is Texas Instrument and Sun's brainchild, it is usualy found at around 50Mhz clok rates featuring up to 1MB of L2 cache, it is available both as single and dual processor modules (SparcStation 10 and SparcStation 20). The higher clock frequency I have encountered so far is 60Mhz.
On a technical point of view this is a SPARC-V8 implementation, it is a superscalar processor, having 2 caches, one for instruction the other one for data.
This is once again Texas Instrument and Sun's brainchild, it can be found in the SparcStation Classic, SparcStation LX, at frequency up to 50Mhz. Its derivative, the Micro SPARC II can be found in the SparcStation 4 and SparcStation 5 at frequencies up to 110Mhz.
On a technical point of view, its main features are a high level of integration, having 2 caches, one for instructions, the other one for data.
It is not possible to add an L2 cache. If you wish to learn more about the MicroSPARC processor you can browse Sun's Ultra SPARC ressources.
This processor was introduced by ROSS in 1993, it is usualy found in the SparcStation 10, and SparcStation 20, at frequencies up to 150Mhz (I have heard of 200Mhz dual processor modules, but Have not witnessed one yet). It can be found on single or dual processor modules.
On a technical point of view it is an implementation of the SPARC-V8, it is superscalar. It can be found with L2 cache up to 512KB
This is a radhard SPARC V7 microprocessor designed to be used on the space segment.
It comes as a single unit or as a three chip package. Main manufacturer is
ATMEL in Nantes, France. At least, one software vendor claims to have GNU/Linux running on this CPU, this is for the
http://dse.cyberclwn.com/sparc-rtems-erc32.htm. This project has not been updated since March 2001. As I have not had the opportunity to check this claim. I am more than doubtful.
This is also a radhard implementation of the SPARC V8 designed to be used on the space segment. It is the ESA's brainchild and the lead designer is jiri gaisler. More information can be found on LEON's website: http://www.gaisler.com/leonmain.html
The 2.4 and 2.5 kernel series are not yet supported, however the 2.0 kernel series is supported by the
uClinux MMU less GNU/Linux distribution. This distribution has been built at ESA/ESTEC December 26 2003 on a SuSE 8.0 GNU/Linux distribution with gcc version 2.95.3 20010315 and a 2.4.18 kernel. Hereafter is the boot sequence and a sample session inside the tsim-leon simulator.
piou@linux:~/uClinux-dist/images> ./tsim-leon -nfp image.elf TSIM/LEON SPARC simulator, version 1.1.4a (evaluation version) Copyright (C) 2001, Gaisler Research - all rights reserved. This software may only be used with a valid license. For latest updates, go to http://www.gaisler.com/ Comments or bug-reports to email@example.com FPU disabled serial port A on stdin/stdout allocated 4096 K RAM memory, in 1 bank(s) allocated 2048 K ROM memory icache: 1 * 4 kbytes, 16 bytes/line (4 kbytes total) dcache: 1 * 4 kbytes, 16 bytes/line (4 kbytes total) section: .text at 0x0, size 252944 bytes section: .data at 0x40000000, size 38452 bytes section: .romfs at 0x3dc10, size 67584 bytes tsim> g resuming at 0x00000000 aCDG512k RAM Found my key Moved .data Found my key uClinux/Sparc Flat model support (C) 1998-2000 Kenneth Albanowski, D. Jeff Dionne LEON-2.1 Sparc V8 support (C) 2000 D. Jeff Dionne, Lineo Inc. LEON-2.2/LEON-2.3 Sparc V8 support (C) 2001 The LEOX team <firstname.lastname@example.org>. Calibrating delay loop.. ok - 6.68 BogoMIPS Memory available: 3904k/4080k RAM, 0k/0k ROM (176k kernel data, 247k code) Swansea University Computer Society NET3.035 for Linux 2.0 NET3: Unix domain sockets 0.13 for Linux NET3.035. uClinux version 2.0.39.uc2 (root@linux) (gcc version 2.95.3 20010315 (release)) 6 Thu Dec 26 18:28:01 PST 2002 LEON serial driver version 0.9 ttyS0 (irq = 3) is a builtin LEON UART Blkmem copyright 1998,1999 D. Jeff Dionne Blkmem copyright 1998 Kenneth Albanowski Blkmem 1 disk images: 0: 3DC10-4E40F (RO) VFS: Mounted root (romfs filesystem) readonly. Sash command shell (version 1.1.1) /> pwd / /> cd bin /bin> pwd /bin /bin> ls sh /bin>
The Ultra SPARC processor is an extension of the SPARC-V9 architecture, it is a 64 bits processor, it features some video processing instructions. It is found in all the computer whose name start with Ultra.
The Ultra SPARC II is an improvement of the Ultra SPARC, the Ultra SPARCIII is actually the second generation of Ultra SPARC processors, it was first introduced in the SunBlade 1000 Workstation. If wish to learn more about the UltraSPARC processors you can browse Sun's Ultra SPARC ressources.
This processor is based on the SPARC V9 and is made by Fujitsu It is a 64bits CPU with some very interesting error handling features such as ECC memory for the L1 cache, hardware instruction retry, error classification.
There is a 64 bit virtual address space and 43 bit physical address space. It is used in the PRIMEPOWER high end servers to mainframe class of Fujitsu's offering.
The cache is organized as :
More information can be found on the http://www.fujitsu.com/downloads/PRMPWR/JPS1-R1.0-SPARC64V-pub.pdf whitepaper.
You may read the CPU-Design-HOWTO, this HOWTO has a lot of interesting links when it comes to studying the CPUs.
To summarize, the 32 bits workstations are the:
For more information on the SparcStation 5, 10, 20 you can read Sun's documentation online or download it available.
The following model have an 64 bits UltraSPARC architecture (sun4u). SunUltra 1, 2, 5, 10, 30, 60, 80 and SunBlade 1000, 1500, 2000. The SunUltra 2, 60, 80 and SunBlade 1000 are SMP capable, with the Ultra 80 and SunBlade 1000 and 2000 accepting up to 4 CPU modules, the SunUltra 2 and 60 accepting only 2 CPU modules.
The SunBlade 2000 is the latest one featuring Sun's latest marvel the Ultra III CPU, at a premium price of course. You can have a summary of the UltraWorkstation still in production at Sun's website.
At first, a reference like SM61 or RT-200-D-125/512 seems to be, to say the least, quite cryptic. Actually, understanding theses references is really easy.
Theses CPUs's naming scheme is
ais a digit:
6SPARC MP600 ( not exactly a workstation ).
bis a letter:
freqThe frequency expressed in Megahertz.
cacheThe amount of cache memory expressed in Kilobytes.
When these modules are in a workstation the naming convention is HSxy, for example
ywing is a SparcStation 20 HS22, thus it is easier to have a look inside the workstation.
This table is extracted from the FAQABOSS
Name Speed( MHz ) Cache( MB ) Number of SuperSparc Processors Series SM20 33 0 1 I SM30 36 0 1 I SM40 40 0 1 I SM41 40 1 1 I SM50 50 0 1 I SM51 50 1 1 I SM512 50 1 2 I SM51-2 50 2 1 I SM61 60 1 1 I SM61-2 60 2 1 I SM71 75 1 1 II SM71-2 75 2 1 II SM81 85 1 1 II SM81-5 85 2 1 II
SM100 is a
Cypress manufactured SPARC compliants processors; AFAIK their naming scheme is CYnnn.
As you can see, this is easy to understand.
This is a family of Network computers that used to be manufactured by Sun, there is a very good JavaStation-HOWTO about it.