ref: 83246e296ea433b65b9d295b5e08fedd39ff1ab7
dir: /libmp/Plan9-amd64/mpvecdigmuladd.s/
/* * mpvecdigmul(mpdigit *b, int n, mpdigit m, mpdigit *p) * * p += b*m * * each step look like: * hi,lo = m*b[i] * lo += oldhi + carry * hi += carry * p[i] += lo * oldhi = hi * * the registers are: * hi = DX - constrained by hardware * lo = AX - constrained by hardware * b+n = SI - can't be BP * p+n = DI - can't be BP * i-n = BP * m = BX * oldhi = CX * */ TEXT mpvecdigmuladd(SB),$0 /* MOVQ b+0(FP),SI */ MOVQ RARG,SI MOVL n+8(FP),CX MOVL m+16(FP),BX MOVQ p+24(FP),DI MOVL CX,BP NEGQ BP /* BP = -n */ SHLL $2,CX ADDQ CX,SI /* SI = b + n */ ADDQ CX,DI /* DI = p + n */ XORL CX,CX _muladdloop: MOVL (SI)(BP*4),AX /* lo = b[i] */ MULL BX /* hi, lo = b[i] * m */ ADDL CX,AX /* lo += oldhi */ JCC _muladdnocarry1 INCL DX /* hi += carry */ _muladdnocarry1: ADDL AX,(DI)(BP*4) /* p[i] += lo */ JCC _muladdnocarry2 INCL DX /* hi += carry */ _muladdnocarry2: MOVL DX,CX /* oldhi = hi */ INCQ BP /* i++ */ JNZ _muladdloop XORL AX,AX ADDL CX,(DI)(BP*4) /* p[n] + oldhi */ ADCL AX,AX /* return carry out of p[n] */ RET