IC64 Libraries

syslib64

The system library provides arithmetic routines at 16-bit, 32-bit, and 64-bit widths — addition, subtraction, multiplication, division, modulo, random numbers, and 64-bit bitwise operations. Every routine is implemented in pure INTERCAL. There is no native fallback: when TIMES64 runs, it is INTERCAL all the way down.

Source vs. binary

The library ships two ways, and you choose based on how much you want to know.

Binary — link and forget

Every build produces syslib64.dll, a prebuilt .NET assembly. Reference it at compile time and call its routines directly:

churn -r:syslib64.dll fizzbuzz.i
./fizzbuzz.exe

Source — read, change, recompile

The library is not a black box. It is syslib64.ic64, a few hundred lines of pure INTERCAL you can read, edit, and rebuild into your own binary:

churn -t:library syslib64.ic64

It is a glass box full of COME FROM. If you disagree with how 64-bit multiplication is done, you are welcome to do it worse yourself.

Calling a routine

The 16-bit routines take operands in .1 and .2, return the result in .3, and set an overflow indicator in .4. The 32-bit routines use the two-spots :1–:4; the 64-bit routines use the double cateyes ::1–::3.

DO .1 <- #40
DO .2 <- #2
PLEASE DO (1000) NEXT
PLEASE READ OUT .3

Routines may be called by their numeric label or by their ASCII name label. Name labels are computed by interpreting the routine name as an 8-character big-endian 64-bit integer — the label for ADD16 is the integer whose bytes are A, D, D, 1, 6 = 4702958889031696384. The programmer who finds this inconvenient is reminded that convenience has never been a design goal.

16-bit arithmetic

Operands in .1, .2. Result in .3. Overflow in .4.

Label	Name	Description
(1000)	ADD16	.3 = .1 + .2 (no overflow check)
(1009)		.3 = .1 + .2 (with overflow check)
(1010)	MINUS16	.3 = .1 - .2
(1020)		.3 = .1 * .2 (low 16 bits)
(1030)	DIVIDE16	.3 = .1 / .2, .4 = .1 mod .2
(1040)	TIMES16	:3 = .1 * .2 (full 32-bit result)
(1050)	MODULO16	.3 = .1 mod .2

32-bit arithmetic

Operands in :1, :2. Result in :3. Overflow in :4.

Label	Name	Description
(1500)	ADD32	:3 = :1 + :2 (no overflow check)
(1509)		:3 = :1 + :2 (with overflow check)
(1510)	MINUS32	:3 = :1 - :2
(1540)	TIMES32	:3 = :1 * :2 (low 32 bits), :4 = high 32 bits

64-bit arithmetic

Operands in ::1, ::2. Result in ::3. Called by name label.

Label	Name	Description
4702958910472978432	ADD64	::3 = ::1 + ::2
5569068542595576832	MINUS64	::3 = ::1 - ::2
6073470532629967872	TIMES64	::3 = ::1 * ::2

Division, modulo, random, and bitwise

Label	Name	Result
(1030)	DIVIDE16	.3 = quotient, .4 = remainder
(1050)	MODULO16	.3 = remainder
4920558940556964658	DIVIDE32	:3 = quotient, :4 = remainder
5570746397223760690	MODULO32	:3 = remainder
(1900)	RANDOM16	.1 = random 16-bit value
5927104639891485490	RANDOM32	:1 = random 32-bit value
5927104639891486260	RANDOM64	::1 = random 64-bit value
4705773660240084992	AND64	::3 = ::1 AND ::2
5715690474052780032	OR64	::3 = ::1 OR ::2
6363395191251927040	XOR64	::3 = ::1 XOR ::2
5642821449895903232	NOT64	::3 = NOT ::1

The 64-bit bitwise routines work by splitting each value into 32-bit halves, mingling the corresponding halves, applying the unary operator, selecting the result bits, and recombining. The full name-label reference for every entry point is in the repository.

Overflow

Label (1999) is the overflow handler. It is abstained from by default when calling through (1000) or (1500), and reinstated on return. Programs that call (1009) or (1509) directly receive overflow errors when the result exceeds the operand width:

(1999) DOUBLE OR SINGLE PRECISION OVERFLOW

The programmer who encounters this error is encouraged to use wider variables.