14 #define b_offset	4
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
68 	mov	s1(%r11,%rdi,4),%r8d;\
69 	movzx	a ## B,		%edi;\
70 	mov	s2(%r11,%rdi,4),%r9d;\
73 	xor	s2(%r11,%rdi,4),%r8d;\
74 	movzx	a ## H,		%edi;\
75 	ror	$16,		a ## D;\
76 	xor	s3(%r11,%rdi,4),%r9d;\
78 	xor	s3(%r11,%rdi,4),%r8d;\
79 	movzx	a ## B,		%edi;\
80 	xor	(%r11,%rdi,4),	%r9d;\
83 	xor	(%r11,%rdi,4),	%r8d;\
84 	movzx	a ## H,		%edi;\
85 	xor	s1(%r11,%rdi,4),%r9d;\
91 	add	k+4+round(%r11),%r8d;\
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
106 	mov	s1(%r11,%rdi,4),%r8d;\
107 	movzx	a ## B,		%edi;\
108 	mov	s2(%r11,%rdi,4),%r9d;\
111 	xor	s2(%r11,%rdi,4),%r8d;\
112 	movzx	a ## H,		%edi;\
113 	ror	$16,		a ## D;\
114 	xor	s3(%r11,%rdi,4),%r9d;\
116 	xor	s3(%r11,%rdi,4),%r8d;\
117 	movzx	a ## B,		%edi;\
118 	xor	(%r11,%rdi,4),	%r9d;\
119 	xor	a,		%r10;\
121 	xor	(%r11,%rdi,4),	%r8d;\
122 	movzx	a ## H,		%edi;\
123 	xor	s1(%r11,%rdi,4),%r9d;\
129 	add	k+4+round(%r11),%r8d;\
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;\
141 	mov	(%r11,%rdi,4),	%r9d;\
143 	mov	s3(%r11,%rdi,4),%r8d;\
144 	movzx	a ## H,		%edi;\
145 	ror	$16,		a ## D;\
146 	xor	s1(%r11,%rdi,4),%r9d;\
149 	xor	(%r11,%rdi,4),	%r8d;\
150 	movzx	a ## B,		%edi;\
151 	xor	s2(%r11,%rdi,4),%r9d;\
153 	xor	s1(%r11,%rdi,4),%r8d;\
154 	movzx	a ## H,		%edi;\
155 	ror	$15,		a ## D;\
156 	xor	s3(%r11,%rdi,4),%r9d;\
158 	xor	s2(%r11,%rdi,4),%r8d;\
163 	add	k+4+round(%r11),%r8d;\
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;\
177 	mov	(%r11,%rdi,4),	%r9d;\
179 	mov	s3(%r11,%rdi,4),%r8d;\
182 	xor	(%r11,%rdi,4),	%r8d;\
183 	movzx	a ## H,		%edi;\
186 	xor	a,		%r10;\
187 	ror	$16,		a ## D;\
188 	xor	s1(%r11,%rdi,4),%r9d;\
190 	xor	s1(%r11,%rdi,4),%r8d;\
191 	movzx	a ## B,		%edi;\
192 	xor	s2(%r11,%rdi,4),%r9d;\
194 	xor	s2(%r11,%rdi,4),%r8d;\
195 	movzx	a ## H,		%edi;\
196 	xor	s3(%r11,%rdi,4),%r9d;\
201 	add	k+4+round(%r11),%r8d;\
230 	encrypt_round(R0,R1,R2,R3,4*8);
290 	decrypt_round(R2,R3,R0,R1,4*8);