Merge pull request #4 from PSYCPU/asm

Asm
12 years ago · 9e039dda47
parent af073e29e4 d9137bb2d8
commit 9e039dda47
1 changed files with 114 additions and 100 deletions
--- a/lin64.asm
+++ b/lin64.asm
@ -1,41 +1,60 @@
-	.x64
+	;; Added by Diederik Huys, March 2013
+	;;
+	;; Provided public procedures:
+	;; 	ExSetMult
+	;; 	ExSetSquare
+	;;
+	;; Needed tools: JWASM (http://www.japheth.de/JWasm.html)
+	;;
+	;; !!! WARNING !!!   !!! WARNING !!!   !!! WARNING !!!
+	;;
+	;; Please note that recompiling this binary (jwasm) under a 64-bit OS
+	;; may yield unexpected results and create a corrupted ELF64 header.
+	;;
+	;; 
+
+	.x64			
 QTEST	EQU 1
 	.code
+	
 	;;  Register Layout:
 	;;  INPUT: 	rdi	= a.n
 	;; 	   	rsi  	= b.n
 	;; 	   	rdx  	= this.a
-	;;  OUTPUT:	[rbx]
+	;; 
 	;;  INTERNAL:	rdx:rax  = multiplication accumulator
-	;; 		rsi	 = b.n / t9
-	;; 		r8:r9    = c
-	;; 		r10-r15  = t0-t5
-	;; 		rbx	 = t6
-	;; 		rcx	 = t7
-	;; 		rbp	 = t8
+	;; 		r9:r8    = c
+	;; 		r10-r13  = t0-t3
+	;; 		r14	 = b.n[0] / t4
+	;; 		r15	 = b.n[1] / t5
+	;; 		rbx	 = b.n[2] / t6
+	;; 		rcx	 = b.n[3] / t7
+	;; 		rbp	 = Constant 0FFFFFFFFFFFFFh / t8
+	;; 		rsi	 = b.n / b.n[4] / t9
 ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	push rdx
-	mov r14,[rsi+8*0]
-	
+	mov r14,[rsi+8*0]	; preload b.n[0]. This will be the case until
+				; b.n[0] is no longer needed, then we reassign
+				; r14 to t4
 	;; c=a.n[0] * b.n[0]
-   	mov rax,[rdi+0*8]
+   	mov rax,[rdi+0*8]	; load a.n[0]
 	mov rbp,0FFFFFFFFFFFFFh
-	mul r14			; rsi=b.n[0]
+	mul r14			; rdx:rax=a.n[0]*b.n[0]
 	mov r15,[rsi+1*8]
-	mov r10,rbp
+	mov r10,rbp		; load modulus into target register for t0
 	mov r8,rax
-	and r10,rax		; only need lower qword
+	and r10,rax		; only need lower qword of c
 	shrd r8,rdx,52
-	xor r9,r9
+	xor r9,r9		; c < 2^64, so we ditch the HO part 

 	;; c+=a.n[0] * b.n[1] + a.n[1] * b.n[0]
 	mov rax,[rdi+0*8]
-	mul r15			; b.n[1]
+	mul r15			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+1*8]
-	mul r14			; b.n[0]
+	mul r14			
 	mov r11,rbp
 	mov rbx,[rsi+2*8]
 	add r8,rax
@ -46,44 +65,44 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	
 	;; c+=a.n[0 1 2] * b.n[2 1 0]
 	mov rax,[rdi+0*8]
-	mul rbx			; b.n[2]
+	mul rbx			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+1*8]
-	mul r15			; b.n[1]	
+	mul r15			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+2*8]
 	mul r14
-	mov r12,rbp		; modulus
+	mov r12,rbp		
 	mov rcx,[rsi+3*8]
 	add r8,rax
 	adc r9,rdx
-	and r12,r8		; only need lower dword
+	and r12,r8		
 	shrd r8,r9,52
 	xor r9,r9		

 	;; c+=a.n[0 1 2 3] * b.n[3 2 1 0]
 	mov rax,[rdi+0*8]
-	mul rcx			; b.n[3]
+	mul rcx			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+1*8]
-	mul rbx			; b.n[2]
+	mul rbx			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+2*8]
-	mul r15			; b.n[1]
+	mul r15			
 	add r8,rax
 	adc r9,rdx
 	
 	mov rax,[rdi+3*8]
-	mul r14			; b.n[0]
-	mov r13,rbp             ; modulus
+	mul r14			
+	mov r13,rbp             
 	mov rsi,[rsi+4*8]	; load b.n[4] and destroy pointer
 	add r8,rax
 	adc r9,rdx
@ -105,18 +124,18 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	adc r9,rdx

 	mov rax,[rdi+2*8]
-	mul rbx			; b.n[2] 
+	mul rbx			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+3*8]
-	mul r15			; b.n[1]
+	mul r15			
 	add r8,rax
 	adc r9,rdx

 	mov rax,[rdi+4*8]
-	mul r14			; b.n[0]
-	mov r14,rbp             ; modulus
+	mul r14			
+	mov r14,rbp             ; load modulus into t4 and destroy a.n[0]
 	add r8,rax
 	adc r9,rdx
 	and r14,r8
@ -141,7 +160,7 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15

 	mov rax,[rdi+4*8]
 	mul r15
-	mov r15,rbp		; modulus
+	mov r15,rbp		
 	add r8,rax
 	adc r9,rdx

@ -162,11 +181,11 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15

 	mov rax,[rdi+4*8]
 	mul rbx
-	mov rbx,rbp		; modulus
+	mov rbx,rbp		
 	add r8,rax
 	adc r9,rdx

-	and rbx,r8		; only need lower dword
+	and rbx,r8		
 	shrd r8,r9,52
 	xor r9,r9		

@ -178,10 +197,10 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15

 	mov rax,[rdi+4*8]
 	mul rcx
-	mov rcx,rbp		; modulus
+	mov rcx,rbp		
 	add r8,rax
 	adc r9,rdx
-	and rcx,r8		; only need lower dword
+	and rcx,r8		
 	shrd r8,r9,52
 	xor r9,r9		

@ -195,17 +214,17 @@ ExSetMult	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	shrd r8,r9,52
 	xor r9,r9		

-	mov rsi,r8
+	mov rsi,r8		; load c into t9 and destroy b.n[4]

 	;; *******************************************************
 common_exit_norm::
-	mov rdi,01000003D10h
+	mov rdi,01000003D10h	; load constant

 	mov rax,r15		; get t5
 	mul rdi
 	add rax,r10    		; +t0
 	adc rdx,0
-	mov r10,0FFFFFFFFFFFFFh ; modulus
+	mov r10,0FFFFFFFFFFFFFh ; modulus. Sadly, we ran out of registers!
 	mov r8,rax		; +c
 	and r10,rax
 	shrd r8,rdx,52
@ -226,12 +245,12 @@ common_exit_norm::
 	mul rdi
 	add rax,r12		; +t2
 	adc rdx,0
-	pop rbx			; retrieve pointer to this.a.n	
+	pop rbx			; retrieve pointer to this.n	
 	mov r12,0FFFFFFFFFFFFFh ; modulus
 	add r8,rax		; +c
 	adc r9,rdx
 	and r12,r8
-	mov [rbx+2*8],r12
+	mov [rbx+2*8],r12	; mov into this.n[2]
 	shrd r8,r9,52
 	xor r9,r9
 	
@ -243,7 +262,7 @@ common_exit_norm::
 	add r8,rax		; +c
 	adc r9,rdx
 	and r13,r8
-	mov [rbx+3*8],r13
+	mov [rbx+3*8],r13	; -> this.n[3]
 	shrd r8,r9,52
 	xor r9,r9
 	
@ -255,11 +274,11 @@ common_exit_norm::
 	add r8,rax		; +c
 	adc r9,rdx
 	and r14,r8
-	mov [rbx+4*8],r14
-	shrd r8,r9,48
+	mov [rbx+4*8],r14	; -> this.n[4]
+	shrd r8,r9,48		; !!!
 	xor r9,r9
 	
-	mov rax,01000003D1h	
+	mov rax,01000003D1h
 	mul r8		
 	add rax,r10
 	adc rdx,0
@ -267,50 +286,46 @@ common_exit_norm::
 	mov r8,rax
 	and rax,r10
 	shrd r8,rdx,52
-	mov [rbx+0*8],rax
+	mov [rbx+0*8],rax	; -> this.n[0]
 	add r8,r11
-	mov [rbx+1*8],r8
+	mov [rbx+1*8],r8	; -> this.n[1]
 	ret
 ExSetMult	ENDP


-
-
-
-
 	;;  Register Layout:
-	;;  INPUT: 	rdi	= a.n
-	;; 	   	rsi  	= this.a
-	;;  OUTPUT:	[rsi]
+	;;  INPUT: 	rdi	 = a.n
+	;; 	   	rsi  	 = this.a
 	;;  INTERNAL:	rdx:rax  = multiplication accumulator
-	;; 		r8:r9    = c
-	;; 		r10-r14  = t0-t4
-	;; 		r15	 = a.n[0]*2 / t5
-	;; 		rbx	 = a.n[1]*2 / t6
-	;; 		rcx	 = a.n[2]*2 / t7
-	;; 		rbp	 = a.n[3]*2 / t8
-	;; 		rsi	 = a.n[4] / t9
+	;; 		r9:r8    = c
+	;; 		r10-r13  = t0-t3
+	;; 		r14	 = a.n[0] / t4
+	;; 		r15	 = a.n[1] / t5
+	;; 		rbx	 = a.n[2] / t6
+	;; 		rcx	 = a.n[3] / t7
+	;; 		rbp	 = 0FFFFFFFFFFFFFh / t8
+	;; 		rsi	 = a.n[4] / a.n[4] /t9
 ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	push rsi
-	mov rsi,0FFFFFFFFFFFFFh
+	mov rbp,0FFFFFFFFFFFFFh
 	
 	;; c=a.n[0] * a.n[0]
-   	mov r15,[rdi+0*8]
-	mov r10,rsi		; modulus 
-	mov rax,r15
-	mul rax  		; rsi=b.n[0]
-	mov rbx,[rdi+1*8]	; a.n[1]
-	add r15,r15		; r15=2*a.n[0]
+   	mov r14,[rdi+0*8]	; r14=a.n[0]
+	mov r10,rbp		; modulus 
+	mov rax,r14
+	mul rax
+	mov r15,[rdi+1*8]	; a.n[1]
+	add r14,r14		; r14=2*a.n[0]
 	mov r8,rax
 	and r10,rax		; only need lower qword
 	shrd r8,rdx,52
 	xor r9,r9

 	;; c+=2*a.n[0] * a.n[1]
-	mov rax,r15
-	mul rbx 	
-	mov rcx,[rdi+2*8]	; rcx=a.n[2]
-	mov r11,rsi 		; modulus
+	mov rax,r14		; r14=2*a.n[0]
+	mul r15
+	mov rbx,[rdi+2*8]	; rbx=a.n[2]
+	mov r11,rbp 		; modulus
 	add r8,rax
 	adc r9,rdx
 	and r11,r8
@ -318,33 +333,32 @@ ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	xor r9,r9
 	
 	;; c+=2*a.n[0]*a.n[2]+a.n[1]*a.n[1]
-	mov rax,r15
-	mul rcx
+	mov rax,r14
+	mul rbx
 	add r8,rax
 	adc r9,rdx

-	mov rax,rbx
-	mov r12,rsi		; modulus
+	mov rax,r15
+	mov r12,rbp		; modulus
 	mul rax
-	mov rbp,[rdi+3*8]	; rbp=a.n[3]
-	add rbx,rbx		; rbx=a.n[1]*2
+	mov rcx,[rdi+3*8]	; rcx=a.n[3]
+	add r15,r15		; r15=a.n[1]*2
 	add r8,rax
 	adc r9,rdx
-
 	and r12,r8		; only need lower dword
 	shrd r8,r9,52
 	xor r9,r9		

 	;; c+=2*a.n[0]*a.n[3]+2*a.n[1]*a.n[2]
-	mov rax,r15
-	mul rbp
+	mov rax,r14
+	mul rcx
 	add r8,rax
 	adc r9,rdx

-	mov rax,rbx		; rax=2*a.n[1]
-	mov r13,rsi		; modulus
-	mul rcx
-	mov rsi,[rdi+4*8]	; rsi=a.n[4] / destroy constant
+	mov rax,r15		; rax=2*a.n[1]
+	mov r13,rbp		; modulus
+	mul rbx
+	mov rsi,[rdi+4*8]	; rsi=a.n[4]
 	add r8,rax
 	adc r9,rdx
 	and r13,r8
@ -352,20 +366,20 @@ ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	xor r9,r9		

 	;; c+=2*a.n[0]*a.n[4]+2*a.n[1]*a.n[3]+a.n[2]*a.n[2]
-	mov rax,r15		; last time we need 2*a.n[0]
+	mov rax,r14		; last time we need 2*a.n[0]
 	mul rsi
 	add r8,rax
 	adc r9,rdx

-	mov rax,rbx
-	mul rbp
-	mov r14,0FFFFFFFFFFFFFh ; modulus
+	mov rax,r15
+	mul rcx
+	mov r14,rbp		; modulus
 	add r8,rax
 	adc r9,rdx

-	mov rax,rcx
+	mov rax,rbx
 	mul rax
-	add rcx,rcx		; rcx=2*a.n[2]
+	add rbx,rbx		; rcx=2*a.n[2]
 	add r8,rax
 	adc r9,rdx
 	and r14,r8
@ -373,14 +387,14 @@ ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	xor r9,r9		

 	;; c+=2*a.n[1]*a.n[4]+2*a.n[2]*a.n[3]
-	mov rax,rbx
+	mov rax,r15		; last time we need 2*a.n[1]
 	mul rsi
 	add r8,rax
 	adc r9,rdx

-	mov rax,rcx
-	mul rbp
-	mov r15,0FFFFFFFFFFFFFh ; modulus
+	mov rax,rbx
+	mul rcx
+	mov r15,rbp		; modulus
 	add r8,rax
 	adc r9,rdx
 	and r15,r8
@ -388,24 +402,24 @@ ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	xor r9,r9		

 	;; c+=2*a.n[2]*a.n[4]+a.n[3]*a.n[3]
-	mov rax,rcx		; 2*a.n[2]
+	mov rax,rbx		; last time we need 2*a.n[2]
 	mul rsi
 	add r8,rax
 	adc r9,rdx

-	mov rax,rbp		; a.n[3]
+	mov rax,rcx		; a.n[3]
 	mul rax
-	mov rbx,0FFFFFFFFFFFFFh ; modulus
+	mov rbx,rbp		; modulus
 	add r8,rax
 	adc r9,rdx
 	and rbx,r8		; only need lower dword
-	lea rax,[2*rbp]
+	lea rax,[2*rcx]
 	shrd r8,r9,52
 	xor r9,r9		

 	;; c+=2*a.n[3]*a.n[4]
 	mul rsi
-	mov rcx,0FFFFFFFFFFFFFh ; modulus
+	mov rcx,rbp		; modulus
 	add r8,rax
 	adc r9,rdx
 	and rcx,r8		; only need lower dword
@ -415,7 +429,7 @@ ExSetSquare	PROC C PUBLIC USES rbx rbp r12 r13 r14 r15
 	;; c+=a.n[4]*a.n[4]
 	mov rax,rsi
 	mul rax
-	mov rbp,0FFFFFFFFFFFFFh ; modulus
+	;; mov rbp,rbp		; modulus is already there!
 	add r8,rax
 	adc r9,rdx
 	and rbp,r8