; $$$$$$$$$$$$$$$$$$$ ABAKIS $$$$$$$$$$$$$$$$$$$$$
; *************** STAR^2 SOFTWARE ****************
; ????????????????? MEMORY.INC ???????????????????
; allocate n ; see SYSTEM.INC
; allocate.p p, n
; destroy p
; memory.set p, v, n ; 32BIT set/copy/zero
; memory.copy a, b, n
; memory.zero p, n
; get.bit v, n ; get/set/zero bit
; set.bit v, n
; zero.bit v, n
; enable.f v, n ; enable/disable flag
; disable.f v, n
; power.2 n ; is power of 2? which one?
; align.n n, p ; versatile align n to p
;;;;;;;;;;;;; MEMORY COPY, SET, ZERO ;;;;;;;;;;;;;
; portable. 32BIT
function memory.copy, a, b, n
alias p=r0, q=r1
. p=a, q=b, n>>>2
loop n, (u32) *p++=*q++, endl
endf
function memory.set, a, b, n
alias p=r0, v=r1, x=r2
. p=a, v=b, n>>>2
loop n, (u32) *p++=v, endl
endf
macro memory.zero p, n { memory.set p, 0, n }
; x86 specific. aligned
function memory.copy.x, a, b, n
push r6 r7
. r7=a,\
r6=b, r1=n
test r7, r6 ; address=0?
jz .e
cmp r1, 4 ; if n<4
jb @f
push r1
shr r1, 2 ; n/4
rep movsd ; copy dwords
pop r1
and r1, 3 ; modulo 4
jz .e ; remainder?
@@:
rep movsb ; copy bytes
.e:
pop r7 r6
endf
function memory.set.x, p, v, n
push r7
. r7=p, r0=v,\
r0*01010101h,\
r1=n
test r7, r7 ; address=0?
jz .e
cmp r1, 4 ; n<4?
jb @f
push r1
shr r1, 2
rep stosd ; copy dwords
pop r1
and r1, 3 ; modulo 4
jz .e ; remainder?
@@:
rep stosb ; copy bytes
.e:
pop r7
endf
;;;;;;;;;;;;;;;; GET/SET/ZERO BIT ;;;;;;;;;;;;;;;;
; 76543210. warning: r0/r1/r2 cannot be used
; as parameters. 'v' should be m, 'i' can be m/i
macro get.bit v, i { ; (v>>i)&1
. r0=v, r1=i, r0>>cl, r0&1
}
macro set.bit v, i { ; v|=(1<<i)
. r0=1, r1=i, r0<<cl, v|r0
}
macro zero.bit v, i { ; v&=~(1<<i)
. r0=1, r1=i, r0<<cl, not r0, v&r0
}
; 1111.0000
macro get.nibble v, i { ; (v>>(i*4))&1111b
. r0=v, r1=i, r1<<2, r0>>cl, r0&1111b
}
macro set.nibble v, i, n { ; v|=(n<<(i*4))
. r0=v, r1=i, r2=n, r1<<2, r2<<cl,\
r0|r2, v=r0
}
; 33.22.11.00
macro get.couple v, i { ; (v>>(i*2))&11b
. r0=v, r1=i, r1<<1, r0>>cl, r0&11b
}
macro set.couple v, i, n { ; v|=(n<<(i*2))
. r0=v, r1=i, r2=n, r1<<1, r2<<cl,\
r0|r2, v=r0
}
; enable/disable flag
macro enable.f v, n { . r0=n, v|r0 }
macro disable.f v, n
{ . r0=n, not r0, v&r0 }
macro toggle n { xor n, 1 } ; invert 1/0
; create AA.BBB.CCCb/AA.BB.CC.DDb BIT structures
function triplet, a, b, c
. r0=a, r0<<6, r1=b, r1<<3, r0|r1, r0|c
endf
function quadruplet, a, b, c, d
. r0=a, r0<<6, r1=b, r1<<4
. r2=c, r2<<2, r0|r1, r0|r2, r0|d
endf
; reverse byte order
macro reverse.32 n
{ . r0=n, bswap r0 }
macro reverse.24 n
{ . r0=n, bswap r0, r0>>>8 }
macro reverse.16 n
{ . r0=n, cl=al, al=ah, ah=cl }
;;;;;;;;;;;;;;;;;; POWERS OF 2 ;;;;;;;;;;;;;;;;;;;
; an unsigned number is a power of 2 if only
; 1 BIT is set: if !(n&n-1). subtracting 1
; inverts all BITs. if n=10000000b (80h/128),
; n&01111111b=0
; to find out which power of 2, search n
; for 1st 0 BIT from right to left
; is n power of 2? example: power.2 128
; returns 7
function power.2, n
locals i
. r0=n
if r0<2, go .r0, end
. r1=r0, r1-1, r0&r1
test r0, r0
jnz .r0
. n--, i=1
@@:
. r0=1, r1=i, r0<<cl, i++
test n, r0
jnz @b
. r0=i, r0--
jmp @f
.r0: . r0=0
@@:
endf
;;;;;;;;;;;;;;;;;;;;; ALIGN ;;;;;;;;;;;;;;;;;;;;;;
; versatile align n/umber by power of 2
; return n aligned to p in r0. in r1,
; return the quantity to add to make n
; divisible by p. algorithm:
; n+(((p-1)-(n+p-1))&(p-1))
function align.n, n, p
. r1=p, r1-1, r2=n, r2+r1, r0=r1
. r0-r2, r0&r1, r1=r0, r2=n, r0+r2
endf
;;;;;;;;;;;;;;;; SOURCE, DESTINY ;;;;;;;;;;;;;;;;;
align
void source, destiny
integer origin, omega
function create.source, size
destroy source
try source=allocate size
memory.zero source, size
. origin=0, omega=0
endf 1
function create.destiny, size
destroy destiny
try destiny=allocate size
memory.zero destiny, size
. origin=0, omega=0
endf 1
;;;;;;;;;;;;;;;;;;; TESTING... ;;;;;;;;;;;;;;;;;;;
; optimized 128-BIT copy/set. ".fast"=CPU
; specific, but they do not replace the
; portable algorithms
; address p must be aligned by 16 (movaps) and
; size n must be divisible by 16. v/alue must
; be 32BIT or use 1/2 macros to expand
function memory.set.fast, p, v, n
. r0=p, r2=v, r1=n,\ ; start at end
r0+r1, neg r1 ; negate index
movd xmm0, r2
pshufd xmm0, xmm0, 0 ; duplicate dwords
@@:
movaps [r0+r1], xmm0
add r1, 16
jnz @b
endf
macro memory.set.2.fast p, v, n {
. r0=v, r0*00010001h
memory.set.fast p, r0, n
}
macro memory.set.1.fast p, v, n {
. r0=v, r0*01010101h
memory.set.fast p, r0, n
}
; destiny a/ddress must be aligned by 16 and
; size n must be divisible by 16
function memory.copy.fast.a, a, b, n
. r0=a, r2=b, r1=n,\
r0+r1, r2+r1, neg r1
@@:
movaps xmm0, [r2+r1]
movaps [r0+r1], xmm0
add r1, 16
jnz @b
endf
; unaligned...
function memory.copy.fast, a, b, n
. r0=a, r2=b, r1=n,\
r0+r1, r2+r1, neg r1
@@:
movups xmm0, [r2+r1]
movups [r0+r1], xmm0
add r1, 16
jnz @b
endf