dslg0849@uxa.cso.uiuc.edu (Daniel S. Lewart) (11/30/90)
Can anyone point me to or provide me with some code which will determine the clock speed? Thanks in advance, Daniel Lewart d-lewart@uiuc.edu
felton@eng3.UUCP (Ed Felton) (12/06/90)
Daniel S. Lewart writes: Can anyone point me to or provide me with some code which will determine the clock speed? Thanks in advance, Daniel Lewart d-lewart@uiuc.edu I mailed this out, Then realized that others might like to see this also: To: dslg0849@uxa.cso.uiuc.edu Subject: Re: How do I determine the clock speed? Newsgroups: comp.lang.pascal In-Reply-To: <1990Nov30.121028.98@ux1.cso.uiuc.edu> Organization: SCI Technology, Inc., Huntsville, Al. In article <1990Nov30.121028.98@ux1.cso.uiuc.edu> you write: } Can anyone point me to or provide me with some code which will determine the } clock speed? } } Thanks in advance, } } Daniel Lewart } d-lewart@uiuc.edu Boy! are you lucky. Just two weeks ago, I had to figure this out. Included below is some code I wrote (intermediate form for testing purposes, Final code is many K, and depends on many other things) for MASM and MSC. First is the MASM routine, which executes a fixed number of machine cycles, and returns the number of clock ticks it took. ------8<------------------------ CUT HERE ----8<----------------------------- ;****************************************************************************; ;* *; ;* Description: This file contains the low-level functions *; ;* required for general usage within the diagnostics *; ;* *; ;* Medium Model Code *; ;* *; ;* Functions: *; ;* *; ;* _GetClkTicks() *; ;* *; ;****************************************************************************; .286 .MODEL MEDIUM, C .DATA control dw ? status dw ? .CODE ;****************************************************************************; GetClkTicks PROC FAR USES BX CX DX, pushf ; save flag register cli ; clear interrupts ; ; Setup a countdown timer. ; mov al, 0bch ; rw lsb then msb, mode2 out 43h, al ; command to counter mov al, 0FFh out 42h, al ; load lsb of counter out 42h, al ; load msb of counter mov cx, 00080h ; set up cpu loop mov bx, 0FFh in al, 61h ; read status byte of 82C301 or al, 1 ; set lsb to 1 out 61h, al ; enables the counter to start decrementing ; ; Timing Loop ; mov ax, 0FFFFh ALIGN 4 ; ; block MACRO idiv bl ; 17 cycles imul bl ; 13 cycles mov ax, 0FFFFh ; 2 cycles ENDM ; ; l1: REPT (2048/32) ; repeat this many times block ; 32 machine cycles per "BLOCK" ENDM dec cx ; 1 machine cycle jcxz done ; 5 machine cycles jmp l1 ; 3 machine cycles ; 1 loop is 2057 CPU clocks done: ; ; End Timing Loop ; mov al, 0d8h out 43h, al ; halts the counter push ax pop ax in al, 42h ; read counter lsb mov ah, al in al, 42h ; read counter msb mov bl, al mov al, ah mov ah, bl popf ret GetClkTicks ENDP ;**************************************************************************** END ------8<------------------------ CUT HERE ----8<----------------------------- And now comes the MS C code, which basically performs interpretation of the data returned from the assembly. Not clean, and not guaranteed to be 100% accurate, but by screwing with the fudge factor, you should be able to get it close enough.... Hint, Right now, on the 20 machines I tried it on, only one 286 didn't read true, and it only read high by a 1 Mhz. ------8<------------------------ CUT HERE ----8<----------------------------- #define FUDGE_286 1.3178 #define FUDGE_386 1.3424 #define FUDGE_486 1.52 #define FUDGE_486C 1.19 #define FUDGE 1.315 main() { unsigned long cycles, ticks; float mhz; unsigned x; x = GetClkTicks(); cycles = (long) 2057 * 0x0080; ticks = 0xFFFF - x; mhz = (FUDGE * cycles) / ticks; printf( "Clock = %u, [%4x]\n",x,x ); printf( "cycles= %lu\n",cycles ); printf( "ticks = %lu\n",ticks); printf( "MHz =%8.5f\n",mhz ); } ------8<------------------------ CUT HERE ----8<----------------------------- And now, for the rest of the story: The fudge factor is determined by taking the known clock rate, and running, making changes until the numbers ring true. This we did on about 20 machines, so the fudge factor should be 95% good to go. Anyhow, if you need to know more, or are p*ssed because I posted C code, email to me. (Haven't spoken pascal in about 4 years, and figured that if you need it bad enough you'd convert it to pascal). So, Good Luck, and let me know how it goes. _______________________________________________________________________________ Ed Felton | When you wish to produce a result by means of an felton@eng3.sci.com | instrument, do not allow yourself to complicate it | -- Leonardo da Vinci Disclaimer: Even if I had opinions, who would want them???
phys169@csc.canterbury.ac.nz (12/07/90)
In article <699@eng3.UUCP>, felton@eng3.UUCP (Ed Felton) writes: > Daniel S. Lewart writes: > Can anyone point me to or provide me with some code which will determine the > clock speed? > And now comes the MS C code... If anyone does want a Pascal+inline asm program to do it (and a bit more), let me know. Mine works on old 8088's, V20's, 286's, 8086's, and 80386(s/d)X. Mark Aitchison, Physics, University of Canterbury, New Zealand.