SWAGOLX.EXE (c) 1993 GDSOFT ALL RIGHTS RESERVED 00002 1 05-25-9408:03ALL DON PAULSEN Fast 16bit CRC SWAG9405 17 N0 π{πRE: SWAG submissionπ This 16-bit CRC function is compatible with those used in Chuckπ Forsberg's X-modem protocol. It's very fast, because I unrolledπ the "for (i = 0; i < 8; ++i)" loop. If a 32-bit CRC is notπ necessary, this is a great alternative because of its speed andπ small size.ππππ{==============================================================}πFUNCTION Crc16 (var buffer; size, seed: word): word; assembler;ππ{ Set size parameter to 0 to process 64K. If processing only one buffer, setπ seed parameter to 0 -- otherwise set to result from previous calculation.π C code translated by Don Paulsen. }ππ(* This routine is a translation of the following C code by Chuck Forsberg.π The added "seed" parameter allows for finding the CRC value of data spanningπ multiple buffers. The innermost loop has been unrolled at a cost of 32π bytes in code, but the speed increase is nearly two-fold.ππ int Crc16 (ptr, count)π char *ptr;π int count;ππ { int crc, i;ππ crc = 0;π while (--count >= 0) {π crc = crc ^ (int)*ptr++ << 8;π for (i = 0; i < 8; ++i)π if (crc & 0x8000)π crc = crc << 1 ^ 0x1021;π elseπ crc = crc << 1;π }π return (crc & 0xFFFF);π }π*)ππASMπ les di, bufferπ mov dx, sizeπ mov ax, seedπ mov si, 1021hπ@next:π xor bl, blπ mov bh, es:[di]π xor ax, bxππ shl ax, 1; jnc @noXor1; xor ax, siπ@noXor1:π shl ax, 1; jnc @noXor2; xor ax, siπ@noXor2:π shl ax, 1; jnc @noXor3; xor ax, siπ@noXor3:π shl ax, 1; jnc @noXor4; xor ax, siπ@noXor4:π shl ax, 1; jnc @noXor5; xor ax, siπ@noXor5:π shl ax, 1; jnc @noXor6; xor ax, siπ@noXor6:π shl ax, 1; jnc @noXor7; xor ax, siπ@noXor7:π shl ax, 1; jnc @noXor8; xor ax, siπ@noXor8:ππ inc diπ dec dxπ jnz @nextπEND;ππ 2 05-26-9406:16ALL DAVID DUNSON Normalize CRC CalculationSWAG9405 7 N0 π{πFrom what I gather from these two routines, in order to "Normalize" a crcπvalue, you must reverse the order of the four bytes in the value.ππExample: crc value = $01020304π normalized = $04030201ππAm I correct in assuming this?ππIf so, the two procedures above fail to perform that task, so here is a BASMπroutine that I have tested and works perfectly.π}ππProcedure Normalize(Var crc: LongInt); Assembler;πASMπ LES DI, crcπ MOV AX, WORD PTR ES:[DI]π MOV BX, WORD PTR ES:[DI + 2]π XCHG AH, ALπ XCHG BH, BLπ MOV WORD PTR ES:[DI + 2], AXπ MOV WORD PTR ES:[DI], BXπEnd;ππPlease forward a copy of your response to Serge Paquin who wrote the originalπrequest for CRC routines.π
