|
|
|
@ -236,13 +236,13 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && defined(__sparc64__) && !defined(NO_ASM) |
|
|
|
#define mulu32(x,y,hi_zuweisung,lo_zuweisung) \ |
|
|
|
({ var register uint64 _hi; \ |
|
|
|
var register uint64 _lo; \ |
|
|
|
__asm__("umul %2,%3,%1\n\trd %y,%0" \ |
|
|
|
: "=r" (_hi), "=r" (_lo) \ |
|
|
|
({ var register uint64 _prod; \ |
|
|
|
__asm__("umul %1,%2,%0" \ |
|
|
|
: "=r" (_prod) \ |
|
|
|
: "r" ((uint32)(x)), "r" ((uint32)(y)) \ |
|
|
|
); \ |
|
|
|
hi_zuweisung (uint32)_hi; lo_zuweisung (uint32)_lo; \ |
|
|
|
hi_zuweisung (uint32)(_prod>>32); \ |
|
|
|
lo_zuweisung (uint32)(_prod); \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && defined(__sparc__) && !defined(NO_ASM) |
|
|
|
#define mulu32(x,y,hi_zuweisung,lo_zuweisung) \ |
|
|
|
@ -314,7 +314,17 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
// mulu32_w(arg1,arg2) |
|
|
|
// > arg1, arg2 : zwei 32-Bit-Zahlen |
|
|
|
// < result : eine 64-Bit-Zahl |
|
|
|
#if defined(__GNUC__) |
|
|
|
#if defined(__GNUC__) && defined(__sparc64__) && !defined(NO_ASM) |
|
|
|
// Prefer the umul instruction over the mulx instruction (overkill). |
|
|
|
#define mulu32_w(x,y) \ |
|
|
|
({ var register uint64 _prod; \ |
|
|
|
__asm__("umul %1,%2,%0" \ |
|
|
|
: "=r" (_prod) \ |
|
|
|
: "r" ((uint32)(x)), "r" ((uint32)(y)) \ |
|
|
|
); \ |
|
|
|
_prod; \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) |
|
|
|
#define mulu32_w(x,y) ((uint64)(uint32)(x) * (uint64)(uint32)(y)) |
|
|
|
#else |
|
|
|
extern "C" uint64 mulu32_w (uint32 arg1, uint32 arg2); |
|
|
|
@ -356,7 +366,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
var register uint64 _lo; \ |
|
|
|
__asm__("mulq %2" \ |
|
|
|
: "=d" /* %rdx */ (_hi), "=a" /* %rax */ (_lo) \ |
|
|
|
: "g" ((uint64)(x)), "1" /* %rax */ ((uint64)(y)) \ |
|
|
|
: "rm" ((uint64)(x)), "1" /* %rax */ ((uint64)(y)) \ |
|
|
|
); \ |
|
|
|
hi_zuweisung _hi; lo_zuweisung _lo; \ |
|
|
|
}) |
|
|
|
@ -468,9 +478,9 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && defined(__arm__) && 0 // see comment cl_asm_arm.cc |
|
|
|
#define divu_3216_1616(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{ var uint32 _q = divu_3216_1616_(x,y); /* extern in Assembler */ \ |
|
|
|
var register uint32 _r __asm__("%r1"/*"%a2"*/); \ |
|
|
|
q_zuweisung _q; r_zuweisung _r; \ |
|
|
|
{ var uint32 __q = divu_3216_1616_(x,y); /* extern in Assembler */ \ |
|
|
|
var register uint32 __r __asm__("%r1"/*"%a2"*/); \ |
|
|
|
q_zuweisung __q; r_zuweisung __r; \ |
|
|
|
} |
|
|
|
#elif defined(__GNUC__) && !defined(__arm__) |
|
|
|
#define divu_3216_1616(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
@ -768,13 +778,54 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
// < uint32 q: floor(x/y) |
|
|
|
// < uint32 r: x mod y |
|
|
|
// < x = q*y+r |
|
|
|
#if defined(__GNUC__) |
|
|
|
#if defined(__GNUC__) && defined(__sparc64__) && !defined(NO_ASM) |
|
|
|
// Prefer the udiv and umul instructions over the udivx and mulx instructions |
|
|
|
// (overkill). |
|
|
|
#define divu_6432_3232_w(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 __x = (x); \ |
|
|
|
var uint32 __xhi = high32(__x); \ |
|
|
|
var uint32 __xlo = low32(__x); \ |
|
|
|
var uint32 __y = (y); \ |
|
|
|
var uint64 __q; \ |
|
|
|
var uint64 __r; \ |
|
|
|
__asm__ __volatile__ ( \ |
|
|
|
"wr %2,%%g0,%%y\n\t" \ |
|
|
|
"udiv %3,%4,%0\n\t" \ |
|
|
|
"umul %0,%4,%1" \ |
|
|
|
"sub %3,%1,%1" \ |
|
|
|
: "=&r" (__q), "=&r" (__r) \ |
|
|
|
: "r" (__xhi), "r" (__xlo), "r" (__y)); \ |
|
|
|
q_zuweisung (uint32)__q; \ |
|
|
|
r_zuweisung (uint32)__r; \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && (defined(__alpha__) || defined(__ia64__) || defined(__mips64__) || defined(__sparc64__)) |
|
|
|
// On __alpha__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __remqu (libc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __ia64__, computing the remainder by multiplication is just four |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __mips64__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __sparc64__, computing the remainder by multiplication uses a 32-bit |
|
|
|
// multiplication instruction, compared to a 64-bit multiplication when the % |
|
|
|
// operator is used. |
|
|
|
#define divu_6432_3232_w(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 __x = (x); \ |
|
|
|
var uint32 __y = (y); \ |
|
|
|
var uint32 __q = floor(__x,(uint64)__y); \ |
|
|
|
q_zuweisung __q; r_zuweisung (uint32)__x - __q * __y; \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && defined(__x86_64__) |
|
|
|
// On __x86_64__, gcc 4.0 performs both quotient and remainder computation |
|
|
|
// in a single instruction. |
|
|
|
#define divu_6432_3232_w(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 __x = (x); \ |
|
|
|
var uint32 __y = (y); \ |
|
|
|
var uint32 __q = floor(__x,(uint64)__y); \ |
|
|
|
q_zuweisung __q; r_zuweisung __x % (uint64)__y; \ |
|
|
|
}) |
|
|
|
#else |
|
|
|
#define divu_6432_3232_w(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{ var uint64 __x = (x); \ |
|
|
|
@ -782,22 +833,125 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
} |
|
|
|
#endif |
|
|
|
|
|
|
|
// Dividiert eine 64-Bit-Zahl durch eine 32-Bit-Zahl und |
|
|
|
// liefert einen 64-Bit-Quotienten und einen 32-Bit-Rest. |
|
|
|
// divu_6432_6432(x,y,q=,r=); |
|
|
|
// > uint64 x: Zähler |
|
|
|
// > uint32 y: Nenner |
|
|
|
// > Es sei bekannt, daß y>0. |
|
|
|
// < uint64 q: floor(x/y) |
|
|
|
// < uint32 r: x mod y |
|
|
|
// < x = q*y+r |
|
|
|
#if defined(__GNUC__) && (defined(__alpha__) || defined(__ia64__) || defined(__mips64__) || defined(__sparc64__)) |
|
|
|
// On __alpha__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __remqu (libc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __ia64__, computing the remainder by multiplication is just four |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __mips64__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __sparc64__, computing the remainder by multiplication uses a 32-bit |
|
|
|
// multiplication instruction, compared to a 64-bit multiplication when the % |
|
|
|
// operator is used. |
|
|
|
#define divu_6432_6432(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 _x = (x); \ |
|
|
|
var uint32 _y = (y); \ |
|
|
|
var uint64 _q; \ |
|
|
|
q_zuweisung _q = floor(_x,(uint64)_y); \ |
|
|
|
r_zuweisung low32(_x) - low32(_q) * _y; \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && defined(__x86_64__) |
|
|
|
// On __x86_64__, gcc 4.0 performs both quotient and remainder computation |
|
|
|
// in a single instruction. |
|
|
|
#define divu_6432_6432(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 _x = (x); \ |
|
|
|
var uint32 _y = (y); \ |
|
|
|
q_zuweisung floor(_x,(uint64)_y); \ |
|
|
|
r_zuweisung _x % (uint64)_y; \ |
|
|
|
}) |
|
|
|
#else |
|
|
|
// Methode: (beta = 2^32) |
|
|
|
// x = x1*beta+x0 schreiben. |
|
|
|
// Division mit Rest: x1 = q1*y + r1, wobei 0 <= x1 < beta <= beta*y. |
|
|
|
// Also 0 <= q1 < beta, 0 <= r1 < y. |
|
|
|
// Division mit Rest: (r1*beta+x0) = q0*y + r0, wobei 0 <= r1*beta+x0 < beta*y. |
|
|
|
// Also 0 <= q0 < beta, 0 <= r0 < y |
|
|
|
// und x = x1*beta+x0 = (q1*beta+q0)*y + r0. |
|
|
|
// Setze q := q1*beta+q0 und r := r0. |
|
|
|
#if defined(__GNUC__) |
|
|
|
#define divu_6432_6432(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 _x = (x); \ |
|
|
|
var uint32 _y = (y); \ |
|
|
|
var uint32 _q1; \ |
|
|
|
var uint32 _q0; \ |
|
|
|
var uint32 _r1; \ |
|
|
|
divu_6432_3232(0,high32(_x),_y, _q1 = , _r1 = ); \ |
|
|
|
divu_6432_3232(_r1,low32(_x),_y, _q0 = , r_zuweisung); \ |
|
|
|
q_zuweisung highlow64(_q1,_q0); \ |
|
|
|
}) |
|
|
|
#else |
|
|
|
#define divu_6432_6432(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{var uint64 _x = (x); \ |
|
|
|
var uint32 _y = (y); \ |
|
|
|
var uint32 _q1; \ |
|
|
|
var uint32 _q0; \ |
|
|
|
var uint32 _r1; \ |
|
|
|
divu_6432_3232(0,high32(_x),_y, _q1 = , _r1 = ); \ |
|
|
|
divu_6432_3232(_r1,low32(_x),_y, _q0 = , r_zuweisung); \ |
|
|
|
q_zuweisung highlow64(_q1,_q0); \ |
|
|
|
} |
|
|
|
#endif |
|
|
|
#endif |
|
|
|
|
|
|
|
// Dividiert eine 64-Bit-Zahl durch eine 64-Bit-Zahl und |
|
|
|
// liefert einen 64-Bit-Quotienten und einen 64-Bit-Rest. |
|
|
|
// divu_6464_6464(x,y,q=,r=); |
|
|
|
// > uint64 x: Zähler |
|
|
|
// > uint64 y: Nenner |
|
|
|
// Es sei bekannt, daß y>0. |
|
|
|
// > Es sei bekannt, daß y>0. |
|
|
|
// < uint64 q: floor(x/y) |
|
|
|
// < uint64 r: x mod y |
|
|
|
// < x = q*y+r |
|
|
|
#if defined(__alpha__) || 1 |
|
|
|
#if defined(__GNUC__) && (defined(__alpha__) || defined(__ia64__) || defined(__mips64__) || defined(__sparc64__)) |
|
|
|
// On __alpha__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __remqu (libc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __ia64__, computing the remainder by multiplication is just four |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __mips64__, computing the remainder by multiplication is just two |
|
|
|
// instructions, compared to the __umoddi3 (libgcc) function call for the % |
|
|
|
// operator. |
|
|
|
// On __sparc64__, it doesn't matter. |
|
|
|
#define divu_6464_6464(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{ var uint64 __x = (x); \ |
|
|
|
var uint64 __y = (y); \ |
|
|
|
q_zuweisung (__x / __y); \ |
|
|
|
r_zuweisung (__x % __y); \ |
|
|
|
} |
|
|
|
({var uint64 _x = (x); \ |
|
|
|
var uint64 _y = (y); \ |
|
|
|
var uint64 _q; \ |
|
|
|
q_zuweisung _q = floor(_x,_y); \ |
|
|
|
r_zuweisung _x - _q * _y; \ |
|
|
|
}) |
|
|
|
#elif defined(__GNUC__) && (defined(__sparc64__) || defined(__x86_64__)) |
|
|
|
// On __sparc64__, it doesn't matter. |
|
|
|
// On __x86_64__, gcc 4.0 performs both quotient and remainder computation |
|
|
|
// in a single instruction. |
|
|
|
#define divu_6464_6464(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
({var uint64 _x = (x); \ |
|
|
|
var uint64 _y = (y); \ |
|
|
|
q_zuweisung floor(_x,_y); \ |
|
|
|
r_zuweisung _x % _y; \ |
|
|
|
}) |
|
|
|
#else |
|
|
|
// For unknown CPUs, we don't know whether gcc's __udivdi3 function plus a |
|
|
|
// multiplication is slower or faster than our own divu_6464_6464_ routine. |
|
|
|
// Anyway, call our own routine. |
|
|
|
extern "C" uint64 divu_6464_6464_ (uint64 x, uint64 y); // -> Quotient q |
|
|
|
extern "C" uint64 divu_64_rest; // -> Rest r |
|
|
|
#define divu_6464_6464(x,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{ q_zuweisung divu_6464_6464_(x,y); r_zuweisung divu_64_rest; } |
|
|
|
#define NEED_VAR_divu_64_rest |
|
|
|
#define NEED_FUNCTION_divu_6464_6464_ |
|
|
|
#endif |
|
|
|
|
|
|
|
// Dividiert eine 128-Bit-Zahl durch eine 64-Bit-Zahl und |
|
|
|
@ -831,6 +985,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
#else |
|
|
|
#define divu_12864_6464(xhi,xlo,y,q_zuweisung,r_zuweisung) \ |
|
|
|
{ q_zuweisung divu_12864_6464_(xhi,xlo,y); r_zuweisung divu_64_rest; } |
|
|
|
#define NEED_VAR_divu_64_rest |
|
|
|
#define NEED_FUNCTION_divu_12864_6464_ |
|
|
|
#endif |
|
|
|
|
|
|
|
@ -1234,6 +1389,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
/* _x32 hat höchstens 1 Bit. Dieses Bit muß gesetzt sein. */\ |
|
|
|
size_zuweisung _bitsize; \ |
|
|
|
} |
|
|
|
#define GENERIC_INTEGERLENGTH32 |
|
|
|
#endif |
|
|
|
|
|
|
|
// Bits einer 64-Bit-Zahl zählen: |
|
|
|
@ -1241,6 +1397,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
// setzt size auf die höchste in digit vorkommende Bitnummer. |
|
|
|
// > digit: ein uint64 >0 |
|
|
|
// < size: >0, <=64, mit 2^(size-1) <= digit < 2^size |
|
|
|
#ifdef GENERIC_INTEGERLENGTH32 |
|
|
|
#define integerlength64(digit,size_zuweisung) \ |
|
|
|
{ var uintC _bitsize = 1; \ |
|
|
|
var uint64 _x64 = (uint64)(digit); \ |
|
|
|
@ -1259,6 +1416,19 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
/* _x64 hat höchstens 1 Bit. Dieses Bit muß gesetzt sein. */\ |
|
|
|
size_zuweisung _bitsize; \ |
|
|
|
} |
|
|
|
#else |
|
|
|
#define integerlength64(digit,size_zuweisung) \ |
|
|
|
{ var uint64 _x64 = (digit); \ |
|
|
|
var uintC _bitsize64 = 0; \ |
|
|
|
var uint32 _x32_from_integerlength64; \ |
|
|
|
if (_x64 >= (1ULL << 32)) { \ |
|
|
|
_x32_from_integerlength64 = _x64>>32; _bitsize64 += 32; \ |
|
|
|
} else { \ |
|
|
|
_x32_from_integerlength64 = _x64; \ |
|
|
|
} \ |
|
|
|
integerlength32(_x32_from_integerlength64, size_zuweisung _bitsize64 + ); \ |
|
|
|
} |
|
|
|
#endif |
|
|
|
|
|
|
|
// Hintere Nullbits eines 32-Bit-Wortes zählen: |
|
|
|
// ord2_32(digit,count=); |
|
|
|
@ -1288,11 +1458,23 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 arg2) |
|
|
|
// Sei n = ord2(x). Dann ist logxor(x,x-1) = 2^n + (2^n-1) = 2^(n+1)-1. |
|
|
|
// Also (ord2 x) = (1- (integer-length (logxor x (1- x)))) . |
|
|
|
#define ord2_32(digit,count_zuweisung) \ |
|
|
|
{ var uint32 _digit = digit ^ (digit - 1); \ |
|
|
|
{ var uint32 _digit = (digit) ^ ((digit) - 1); \ |
|
|
|
integerlength32(_digit,count_zuweisung -1 + ) \ |
|
|
|
} |
|
|
|
#endif |
|
|
|
|
|
|
|
// Hintere Nullbits eines 64-Bit-Wortes zählen: |
|
|
|
// ord2_64(digit,count=); |
|
|
|
// setzt size auf die kleinste in digit vorkommende Bitnummer. |
|
|
|
// > digit: ein uint64 >0 |
|
|
|
// < count: >=0, <64, mit 2^count | digit, digit/2^count ungerade |
|
|
|
// Sei n = ord2(x). Dann ist logxor(x,x-1) = 2^n + (2^n-1) = 2^(n+1)-1. |
|
|
|
// Also (ord2 x) = (1- (integer-length (logxor x (1- x)))) . |
|
|
|
#define ord2_64(digit,count_zuweisung) \ |
|
|
|
{ var uint64 _digit = (digit) ^ ((digit) - 1); \ |
|
|
|
integerlength64(_digit,count_zuweisung -1 + ) \ |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// Bits eines Wortes zählen. |
|
|
|
// logcount_NN(); |
|
|
|
|
Richard Kreckel
20 years ago