17 #define ctx       4  /* Twofish context structure */
20 #define b_offset	4
62  * b input register containing b
65  * operations on a and b are interleaved to increase performance
67 #define encrypt_round(a,b,c,d,round)\  argument
69 	movzx	b ## B,		%edi;\
70 	mov	s1(%ebp,%edi,4),d ## D;\
71 	movzx	a ## B,		%edi;\
72 	mov	s2(%ebp,%edi,4),%esi;\
73 	movzx	b ## H,		%edi;\
74 	ror	$16,		b ## D;\
75 	xor	s2(%ebp,%edi,4),d ## D;\
78 	xor	s3(%ebp,%edi,4),%esi;\
79 	movzx	b ## B,		%edi;\
80 	xor	s3(%ebp,%edi,4),d ## D;\
81 	movzx	a ## B,		%edi;\
82 	xor	(%ebp,%edi,4),	%esi;\
83 	movzx	b ## H,		%edi;\
84 	ror	$15,		b ## D;\
85 	xor	(%ebp,%edi,4),	d ## D;\
87 	xor	s1(%ebp,%edi,4),%esi;\
94 	add	k+4+round(%ebp),d ## D;\
99  * b input register containing b
102  * operations on a and b are interleaved to increase performance
105 #define encrypt_last_round(a,b,c,d,round)\  argument
107 	movzx	b ## B,		%edi;\
108 	mov	s1(%ebp,%edi,4),d ## D;\
109 	movzx	a ## B,		%edi;\
110 	mov	s2(%ebp,%edi,4),%esi;\
111 	movzx	b ## H,		%edi;\
112 	ror	$16,		b ## D;\
113 	xor	s2(%ebp,%edi,4),d ## D;\
116 	xor	s3(%ebp,%edi,4),%esi;\
117 	movzx	b ## B,		%edi;\
118 	xor	s3(%ebp,%edi,4),d ## D;\
119 	movzx	a ## B,		%edi;\
120 	xor	(%ebp,%edi,4),	%esi;\
121 	movzx	b ## H,		%edi;\
122 	ror	$16,		b ## D;\
123 	xor	(%ebp,%edi,4),	d ## D;\
125 	xor	s1(%ebp,%edi,4),%esi;\
132 	add	k+4+round(%ebp),d ## D;\
137  * b input register containing b (rotated 16)
140  * operations on a and b are interleaved to increase performance
142 #define decrypt_round(a,b,c,d,round)\  argument
144 	movzx	a ## B,		%edi;\
145 	mov	(%ebp,%edi,4),	c ## D;\
146 	movzx	b ## B,		%edi;\
147 	mov	s3(%ebp,%edi,4),%esi;\
150 	xor	s1(%ebp,%edi,4),c ## D;\
151 	movzx	b ## H,		%edi;\
152 	ror	$16,		b ## D;\
153 	xor	(%ebp,%edi,4),	%esi;\
154 	movzx	a ## B,		%edi;\
155 	xor	s2(%ebp,%edi,4),c ## D;\
156 	movzx	b ## B,		%edi;\
157 	xor	s1(%ebp,%edi,4),%esi;\
160 	xor	s3(%ebp,%edi,4),c ## D;\
161 	movzx	b ## H,		%edi;\
162 	xor	s2(%ebp,%edi,4),%esi;\
168 	add	k+4+round(%ebp),%esi;\
174  * b input register containing b (rotated 16)
177  * operations on a and b are interleaved to increase performance
180 #define decrypt_last_round(a,b,c,d,round)\  argument
182 	movzx	a ## B,		%edi;\
183 	mov	(%ebp,%edi,4),	c ## D;\
184 	movzx	b ## B,		%edi;\
185 	mov	s3(%ebp,%edi,4),%esi;\
188 	xor	s1(%ebp,%edi,4),c ## D;\
189 	movzx	b ## H,		%edi;\
190 	ror	$16,		b ## D;\
191 	xor	(%ebp,%edi,4),	%esi;\
192 	movzx	a ## B,		%edi;\
193 	xor	s2(%ebp,%edi,4),c ## D;\
194 	movzx	b ## B,		%edi;\
195 	xor	s1(%ebp,%edi,4),%esi;\
198 	xor	s3(%ebp,%edi,4),c ## D;\
199 	movzx	b ## H,		%edi;\
200 	xor	s2(%ebp,%edi,4),%esi;\
206 	add	k+4+round(%ebp),%esi;\
235 	encrypt_round(R0,R1,R2,R3,4*8);
299 	decrypt_round(R2,R3,R0,R1,4*8);