Power 6/32
The Power 6/32, code named Tahoe, from Computer Consoles Incorporated, was a supermini; it used the VERSAbus as its bus. It is most noted from the CSRG porting 4.3 BSD to it, in the 4.3 BSD Tahoe release.
Although it was not very successful commercially (in part because CCI exited the computer business shortly after it was released), it was very influential as the first non-VAX machine that BSD UNIX was ported to. The Tahoe did help separate out a lot of VAX-specific code from the base, allowing BSD to become more portable, but the Power 6/32 machine quickly disappeared off the market, and not much is known about them.
The CPU was re-sold (under other names) in systems from several vendors (viewed by CCI as OEMs): Unisys sold the 6/32 under the name 'Unisys 7000/40', using their own bastardized BSD+SYSV Unix. (Kuwait Petroleum in Denmark had one of them.) ICL had the 'Clan 7'. Harris had the HCX-5, -7, and -9 models (for which a small amount of information survives - below), running the SysV derivative HCX/UX.
Pixar used a Power 6/32 to render the stained-glass knight in the 1985 movie Young Sherlock Holmes.
Details
The Power 6/32 CPU was implemented in TTL; it had a clock rate of 100nS.
The performance is given in this table, from dhrystone.c (below):
*----------------DHRYSTONE VERSION 1.0 RESULTS BEGIN-------------------------- * * MACHINE MICROPROCESSOR OPERATING COMPILER DHRYSTONES/SEC. * TYPE SYSTEM NO REG REGS * -------------------------- ------------ ----------- --------------- * CCI POWER 6/32 COS(SV+4.2) cc 7500 7800 * CCI POWER 6/32 POWER 6 UNIX/V cc 8236 8498 * CCI POWER 6/32 4.2 Rel. 1.2b cc 8963 9544 * VAX 11/780 - UNIX 5.2 cc 1515 1562 * VAX 11/780 - UNIX 4.3bsd cc 1646 1662
Which may give some indication on the initial reasons why the Power 6/32 was chosen as the successor to the VAX by CSRG.
Instruction Set
Although apparently no documentation survives and is accessible, we can infer a lot from the instruction decoding/printing routines in the debuggers from 4.3 BSD Tahoe, and their input file 'instrs' (below).
- Instructions start with a 1-byte opcode, and, depending on the opcode and 'addressing mode' (below), are followed by a basic set of 0-3 operands. (Two instructions, 'ediv' and 'emul', take 4 operands; if any operand uses the 'indexed' addressing mode, an extra operand is required for each such use.)
- Operands start with 1 byte, containing 4 bits of 'addressing mode' and 4 bits of 'register':
Encoding Mode notation 0x0 short literals 0-15 0x1 short literals 16-31 0x2 short literals 32-47 0x3 short literals 48-63 0x4 indexed [r] 0x5 register immediate r 0x6 register deferred (r) 0x7 auto-decrement -(r) 0x8 auto-increment (r)+ 0x9 auto-increment deferred *(r)+ 0xA byte displacement BD(r) 0xB byte displacement deferred *BD(r) 0xC short-word displacement WD(r) 0xD short-word displacement deferred *WD(r) 0xE long-word displacement LD(r) 0xF long-word displacement deferred *LD(r)
- The "indexed" mode apparently adds the contents of the named register to the address generated by the following operand. (It likely cannot be used with the "register immediate" addressing mode.)
- For operands of the 6 'displacement' modes, the operand is followed by a value of the indicated length.
Opcode table
The opcode map and mnemonics for the instruction set in the 4.3 BSD assembler, along with the number of operands for each instruction (if any) are as follows:
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | |
|---|---|---|---|---|---|---|---|---|
| 0x00 | halt | sinf | ldf - 1 | ldd - 1 | ||||
| 0x10 | nop | brb - 1 | brw - 1 | cosf | lnf - 1 | lnd - 1 | ||
| 0x20 | rei | bneq/bnequ - 1 | cvtwl - 2 | atanf | stf - 1 | std - 1 | ||
| 0x30 | bpt | beql/beqlu - 1 | cvtwb - 2 | logf | cmpf - 1 | cmpd - 1 | ||
| 0x40 | ret | bgtr - 1 | sqrtf | cmpf2 - 2 | cmpd2 - 2 | |||
| 0x50 | bleq - 1 | expf | tstf | tstd | ||||
| 0x60 | ldpctx | pushd | ||||||
| 0x70 | svpctx | jmp - 1 | cvlf - 1 | cvld - 1 | ||||
| 0x80 | bgeq - 1 | movs2 | cvfl - 1 | cvdl - 1 | ||||
| 0x90 | blss - 1 | cmps2 | ldfd - 1 | |||||
| 0xA0 | bgtru - 1 | cvdf | ||||||
| 0xB0 | blequ - 1 | negf | negd | |||||
| 0xC0 | prober - 3 | bvc - 1 | movs3 | movzwl - 2 | addf - 1 | addd - 1 | ||
| 0xD0 | probew - 3 | bvs - 1 | cmps3 | subf - 1 | subd - 1 | |||
| 0xE0 | insque - 2 | bgequ/bcs - 1 | mulf - 1 | muld - 1 | ||||
| 0xF0 | remque - 1 | bcc/blssu - 1 | divf - 1 | divd - 1 |
| 8 | 9 | A | B | C | D | E | F | |
|---|---|---|---|---|---|---|---|---|
| 0x00 | addb2 - 2 | movb - 2 | addw2 - 2 | movw - 2 | addl2 - 2 | movl - 2 | bbs - 3 | |
| 0x10 | addb3 - 3 | cmpb - 2 | addw3 - 3 | cmpw - 2 | addl3 - 3 | cmpl - 2 | bbc - 3 | |
| 0x20 | subb2 - 2 | mcomb - 2 | subw2 - 2 | mcomw - 2 | subl2 - 2 | mcoml - 2 | emul - 4 | aoblss - 3 |
| 0x30 | subb3 - 3 | bitb - 2 | subw3 - 3 | bitw - 2 | subl3 - 3 | bitl - 2 | ediv - 4 | aobleq - 3 |
| 0x40 | shll - 3 | clrb - 1 | shlq - 3 | clrw - 1 | mull2 - 2 | clrl - 1 | shal - 3 | |
| 0x50 | shrl - 3 | tstb - 1 | shrq - 3 | tstw - 1 | mull3 - 3 | tstl - 1 | shar - 3 | bbssi - 3 |
| 0x60 | incb - 1 | incw - 1 | divl2 - 2 | incl - 1 | cvtlb - 2 | |||
| 0x70 | decb - 1 | decw - 1 | divl3 - 3 | decl - 1 | cvtlw - 2 | |||
| 0x80 | orb2 - 2 | cvtbl - 2 | orw2 - 2 | bispsw - 1 | orl2 - 2 | adwc - 2 | adda - 2 | |
| 0x90 | orb3 - 3 | cvtbw - 2 | orw3 - 3 | bicpsw - 1 | orl3 - 3 | sbwc - 2 | suba - 2 | |
| 0xA0 | andb2 - 2 | movzbl - 2 | andw2 - 2 | loadr - 2 | andl2 - 2 | mtpr - 2 | ffs - 2 | |
| 0xB0 | andb3 - 3 | movzbw - 2 | andw3 - 3 | storer - 2 | andl3 - 3 | mfpr - 2 | ffc - 2 | calls - 2 |
| 0xC0 | xorb2 - 2 | movob - 2 | xorw2 - 2 | movow - 2 | xorl2 - 2 | movpsl - 1 | btcs - 1 | kcall - 1 |
| 0xD0 | xorb3 - 3 | pushb - 1 | xorw3 - 3 | pushw - 1 | xorl3 - 3 | pushl - 1 | ||
| 0xE0 | mnegb - 2 | movab - 2 | mnegw - 2 | movaw - 2 | mnegl - 2 | moval - 2 | ||
| 0xF0 | movblk | pushab - 1 | pushaw - 1 | casel - 3 | pushal - 1 | callf - 2 |
A number of the more complex arithmetic instructions, many of them the floating point and the trigonometric ones (CVLF, CVLD, CVFL, CVDL, LDFD, CVDF, ADDF, ADDD, SUBF, SUBD, MULF, MULD, DIVF, DIVD, SINF, COSF, ATANF, LOGF, SQRTF and EXPF), do not seem to be implemented in the Power 6/32; instead, they were emulated in software, after an attempt to execute them resulted in a trap.
See also
External links
- Harris HCX-5 - ad
- Harris HCX-9 - ad
- The Harris HCX-9 aka TAHOE platform
- Installing and Operating 4.3BSD UNIX on the Tahoe
- adb - debugger source
- dbx
- instrs - basic instruction table
- instrs - alphabetical list
- dhrystone.c
- Cook, Carpenter, Catmull 1987 "The Reyes Image Rendering Architecture" in ACM Computer Graphics Vol. 21, No. 4 PDF indicating this machine was used at Pixar for a number of films
