15 #define c_offset	8
24 #define w	4096	/* 8 whitening keys (word) */
27 /* define a few register aliases to allow macro substitution */
60  * a input register containing a (rotated 16)
64  * operations on a and b are interleaved to increase performance
66 #define encrypt_round(a,b,c,d,round)\  argument
69 	movzx	a ## B,		%edi;\
74 	movzx	a ## H,		%edi;\
75 	ror	$16,		a ## D;\
79 	movzx	a ## B,		%edi;\
84 	movzx	a ## H,		%edi;\
95  * a input register containing a(rotated 16)
99  * operations on a and b are interleaved to increase performance
100  * during the round a and b are prepared for the output whitening
102 #define encrypt_last_round(a,b,c,d,round)\  argument
107 	movzx	a ## B,		%edi;\
112 	movzx	a ## H,		%edi;\
113 	ror	$16,		a ## D;\
117 	movzx	a ## B,		%edi;\
119 	xor	a,		%r10;\
122 	movzx	a ## H,		%edi;\
133  * a input register containing a
137  * operations on a and b are interleaved to increase performance
139 #define decrypt_round(a,b,c,d,round)\  argument
140 	movzx	a ## B,		%edi;\
144 	movzx	a ## H,		%edi;\
145 	ror	$16,		a ## D;\
150 	movzx	a ## B,		%edi;\
154 	movzx	a ## H,		%edi;\
155 	ror	$15,		a ## D;\
168  * a input register containing a
172  * operations on a and b are interleaved to increase performance
173  * during the round a and b are prepared for the output whitening
175 #define decrypt_last_round(a,b,c,d,round)\  argument
176 	movzx	a ## B,		%edi;\
183 	movzx	a ## H,		%edi;\
186 	xor	a,		%r10;\
187 	ror	$16,		a ## D;\
191 	movzx	a ## B,		%edi;\
195 	movzx	a ## H,		%edi;\
212 	as target for the 8bit high operations */
216 	movq	8(R3),	R3
227 	encrypt_round(R2,R3,R0,R1,8);
228 	encrypt_round(R0,R1,R2,R3,2*8);
229 	encrypt_round(R2,R3,R0,R1,3*8);
230 	encrypt_round(R0,R1,R2,R3,4*8);
231 	encrypt_round(R2,R3,R0,R1,5*8);
232 	encrypt_round(R0,R1,R2,R3,6*8);
233 	encrypt_round(R2,R3,R0,R1,7*8);
234 	encrypt_round(R0,R1,R2,R3,8*8);
235 	encrypt_round(R2,R3,R0,R1,9*8);
236 	encrypt_round(R0,R1,R2,R3,10*8);
237 	encrypt_round(R2,R3,R0,R1,11*8);
238 	encrypt_round(R0,R1,R2,R3,12*8);
239 	encrypt_round(R2,R3,R0,R1,13*8);
240 	encrypt_round(R0,R1,R2,R3,14*8);
241 	encrypt_last_round(R2,R3,R0,R1,15*8);
251 	movq	R1,	8(%rsi)
265 	as target for the 8bit high operations */
269 	movq	8(R3),	R3
279 	decrypt_round(R0,R1,R2,R3,15*8);
280 	decrypt_round(R2,R3,R0,R1,14*8);
281 	decrypt_round(R0,R1,R2,R3,13*8);
282 	decrypt_round(R2,R3,R0,R1,12*8);
283 	decrypt_round(R0,R1,R2,R3,11*8);
284 	decrypt_round(R2,R3,R0,R1,10*8);
285 	decrypt_round(R0,R1,R2,R3,9*8);
286 	decrypt_round(R2,R3,R0,R1,8*8);
287 	decrypt_round(R0,R1,R2,R3,7*8);
288 	decrypt_round(R2,R3,R0,R1,6*8);
289 	decrypt_round(R0,R1,R2,R3,5*8);
290 	decrypt_round(R2,R3,R0,R1,4*8);
291 	decrypt_round(R0,R1,R2,R3,3*8);
292 	decrypt_round(R2,R3,R0,R1,2*8);
293 	decrypt_round(R0,R1,R2,R3,1*8);
303 	movq	R1,	8(%rsi)