[fceu.git] / drivers / arm / neon_scale2x.S

@@\r
@@  Copyright (C) 2012 Roman Pauer\r
@@\r
@@  Permission is hereby granted, free of charge, to any person obtaining a copy of\r
@@  this software and associated documentation files (the "Software"), to deal in\r
@@  the Software without restriction, including without limitation the rights to\r
@@  use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies\r
@@  of the Software, and to permit persons to whom the Software is furnished to do\r
@@  so, subject to the following conditions:\r
@@\r
@@  The above copyright notice and this permission notice shall be included in all\r
@@  copies or substantial portions of the Software.\r
@@\r
@@  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\r
@@  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\r
@@  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\r
@@  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\r
@@  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\r
@@  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\r
@@  SOFTWARE.\r
@@\r
\r
.arm\r
\r
#include "neon_scale2x.Sinc"\r
#include "neon_normalxx.Sinc"\r
\r
.global neon_scale2x_8_8\r
.global neon_scale2x_16_16\r
.global neon_scale2x_8_16\r
\r
.align 4\r
neon_scale2x_8_8:\r
\r
@ r0     = const uint8_t *src\r
@ r1     = uint8_t *dst\r
@ r2     = unsigned int width (pixels)\r
@ r3     = unsigned int srcstride (bytes)\r
@ [sp]   = unsigned int dststride (bytes)\r
@ [sp+4] = unsigned int height\r
@ lr     = return address\r
\r
        ldr	ip, [sp]                    @ ip = dststride\r
        push {r4-r9}\r
        ldr	r9, [sp, #(7*4)]            @ r9 = height\r
        sub r4, r0, r3                  @ r4 = src - srcstride\r
        add r5, r0, r3                  @ r5 = src + srcstride\r
        add r6, r1, ip                  @ r6 = dst + dststride\r
        sub r3, r3, r2                  @ r3 = srcstride - width\r
        sub ip, ip, r2                  @ ip = dststride - width\r
        lsl ip, #1                      @ ip = 2 * dststride - 2 * width\r
        mov r7, r2                      @ r7 = width\r
        sub r9, r9, #2                  @ r9 = height - 2\r
\r
@ r0 = src\r
@ r1 = dst\r
@ r2 = width\r
@ r3 = srcdiff (srcstride - width)\r
@ r4 = src - srcstride\r
@ r5 = src + srcstride\r
@ r6 = dst + dststride\r
@ r7 = counter\r
@ r8 = tmpreg\r
@ r9 = height\r
@ ip = dstdiff (2 * dststride - 2 * width)\r
\r
    @ first line\r
        neon_scale2x_8_8_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        add r0, r0, r3\r
        add r4, r4, r3\r
        add r5, r5, r3\r
        add r1, r1, ip\r
        add r6, r6, ip\r
\r
    @ middle lines\r
    101:\r
        mov r7, r2\r
\r
        neon_scale2x_8_8_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        subS r9, r9, #1\r
        add r0, r0, r3\r
        add r4, r4, r3\r
        add r5, r5, r3\r
        add r1, r1, ip\r
        add r6, r6, ip\r
        bne 101b\r
\r
    @ last line\r
        mov r7, r2\r
\r
        neon_scale2x_8_8_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        pop {r4-r9}\r
        bx lr\r
\r
@ end procedure neon_scale2x_8_8\r
\r
\r
neon_scale2x_16_16:\r
\r
@ r0     = const uint16_t *src\r
@ r1     = uint16_t *dst\r
@ r2     = unsigned int width (pixels)\r
@ r3     = unsigned int srcstride (bytes)\r
@ [sp]   = unsigned int dststride (bytes)\r
@ [sp+4] = unsigned int height\r
@ lr     = return address\r
\r
        ldr	ip, [sp]                    @ ip = dststride\r
        push {r4-r9}\r
        ldr	r9, [sp, #(7*4)]            @ r9 = height\r
        sub r4, r0, r3                  @ r4 = src - srcstride\r
        add r5, r0, r3                  @ r5 = src + srcstride\r
        add r6, r1, ip                  @ r6 = dst + dststride\r
        sub r3, r3, r2, lsl #1          @ r3 = srcstride - 2 * width\r
        sub ip, ip, r2, lsl #1          @ ip = dststride - 2 * width\r
        lsl ip, #1                      @ ip = 2 * dststride - 4 * width\r
        mov r7, r2                      @ r7 = width\r
        sub r9, r9, #2                  @ r9 = height - 2\r
\r
@ r0 = src\r
@ r1 = dst\r
@ r2 = width\r
@ r3 = srcdiff (srcstride - 2 * width)\r
@ r4 = src - srcstride\r
@ r5 = src + srcstride\r
@ r6 = dst + dststride\r
@ r7 = counter\r
@ r8 = tmpreg\r
@ r9 = height\r
@ ip = dstdiff (2 * dststride - 4 * width)\r
\r
    @ first line\r
        neon_scale2x_16_16_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        add r0, r0, r3\r
        add r4, r4, r3\r
        add r5, r5, r3\r
        add r1, r1, ip\r
        add r6, r6, ip\r
\r
    @ middle lines\r
    101:\r
        mov r7, r2\r
\r
        neon_scale2x_16_16_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        subS r9, r9, #1\r
        add r0, r0, r3\r
        add r4, r4, r3\r
        add r5, r5, r3\r
        add r1, r1, ip\r
        add r6, r6, ip\r
        bne 101b\r
\r
    @ last line\r
        mov r7, r2\r
\r
        neon_scale2x_16_16_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
\r
        pop {r4-r9}\r
        bx lr\r
\r
@ end procedure neon_scale2x_16_16\r
\r
\r
neon_scale2x_8_16:\r
\r
@ r0     = const uint8_t *src\r
@ r1     = uint8_t *dst\r
@ r2     = const uint32_t *palette\r
@ r3     = unsigned int width (pixels)\r
@ [sp]   = unsigned int srcstride (bytes)\r
@ [sp+4] = unsigned int dststride (bytes)\r
@ [sp+8] = unsigned int height\r
@ lr     = return address\r
\r
@ three temporary lines\r
\r
        ldr	ip, [sp]                @ ip = srcstride\r
        push {r4-r11,lr}\r
        ldr r4, [sp, #(4*10)]       @ r4 = dststride\r
        ldr r5, [sp, #(4*11)]       @ r5 = height\r
        mov r6, sp                  @ r6 = sp\r
        sub ip, ip, r3              @ ip = srcstride - width\r
        bic sp, sp, #31             @ align sp to 32 bytes\r
        sub r7, r4, r3, lsl #1      @ r7 = dststride - 2 * width\r
        sub sp, sp, r3, lsl #1      @ sp -= 2 * width\r
        sub r5, r5, #2              @ height -= 2\r
        mov r10, sp                 @ tmpline3 = sp\r
        lsl r7, #1                  @ r7 = 2 * dststride - 4 * width\r
        bic sp, sp, #31             @ align sp to 32 bytes\r
        sub sp, sp, r3, lsl #1      @ sp -= 2 * width\r
        mov r11, sp                 @ tmpline2 = sp\r
        bic sp, sp, #31             @ align sp to 32 bytes\r
        sub sp, sp, r3, lsl #1      @ sp -= 2 * width\r
        mov lr, sp                  @ tmpline1 = sp\r
        bic sp, sp, #31             @ align sp to 32 bytes\r
        sub sp, sp, #36\r
        str r6, [sp]                @ oldsp = r6\r
        str r5, [sp, #4]            @ height = r5\r
        str ip, [sp, #8]            @ srcdiff = ip\r
        str r7, [sp, #12]           @ dstdiff = r7\r
        str r4, [sp, #16]           @ dststride = r4\r
        str lr, [sp, #20]           @ tmpline1 = lr\r
        str r11, [sp, #24]          @ tmpline2 = r11\r
        str r10, [sp, #28]          @ tmpline3 = r10\r
        str r3, [sp, #32]           @ width = r3\r
\r
@ r0  = src\r
@ r1  = dst\r
@ r2  = palette\r
@ r3  = counter\r
@ r4  = dst2\r
\r
@ r11 = bufptr1\r
@ ip  = bufptr2\r
@ lr  = bufptr3\r
\r
@ [sp]      = oldsp\r
@ [sp, #4]  = height\r
@ [sp, #8]  = srcdiff (srcstride - width)\r
@ [sp, #12] = dstdiff (2 * dststride - 4 * width)\r
@ [sp, #16] = dststride\r
@ [sp, #20] = tmpline1\r
@ [sp, #24] = tmpline2\r
@ [sp, #28] = tmpline3\r
@ [sp, #32] = width\r
\r
    @ lr = tmpline1\r
    @ r3 = counter\r
\r
    @ first line\r
        neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
\r
        ldr r7, [sp, #8]            @ r7 = srcdiff\r
        ldr r3, [sp, #32]           @ counter = width\r
        ldr lr, [sp, #24]           @ bufptr3 = tmpline2\r
        add r0, r0, r7              @ src += srcdiff\r
\r
    @ second line\r
        neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
\r
        ldr r9, [sp, #16]           @ r9 = dststride\r
        ldr r3, [sp, #32]           @ counter = width\r
        ldr ip, [sp, #20]           @ bufptr2 = tmpline1\r
        ldr lr, [sp, #24]           @ bufptr3 = tmpline2\r
        add r4, r1, r9              @ dst2 = dst + dststride\r
\r
    @ first temporary line\r
        neon_scale2x_16_16_line first, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
\r
        ldr r7, [sp, #8]            @ r7 = srcdiff\r
        ldr r8, [sp, #12]           @ r8 = dstdiff\r
        ldr r3, [sp, #32]           @ counter = width\r
        ldr lr, [sp, #28]           @ bufptr3 = tmpline3\r
        add r0, r0, r7              @ src += srcdiff\r
        add r1, r1, r8              @ dst += dstdiff\r
\r
    100:\r
\r
    @ line n+1\r
        neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
\r
        ldr r9, [sp, #16]           @ r9 = dststride\r
        ldr r11, [sp, #20]          @ bufptr1 = tmpline1\r
        ldr ip, [sp, #24]           @ bufptr2 = tmpline2\r
        ldr lr, [sp, #28]           @ bufptr3 = tmpline3\r
        add r4, r1, r9              @ dst2 = dst + dststride\r
        ldr r3, [sp, #32]           @ counter = width\r
        str r11, [sp, #28]          @ tmpline3 = bufptr1\r
        str ip, [sp, #20]           @ tmpline1 = bufptr2\r
        str lr, [sp, #24]           @ tmpline2 = bufptr3\r
\r
    @ temporary line n\r
        neon_scale2x_16_16_line middle, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
\r
        ldr r6, [sp, #4]            @ r6 = height\r
        ldr r7, [sp, #8]            @ r7 = srcdiff\r
        ldr r8, [sp, #12]           @ r8 = dstdiff\r
        ldr r3, [sp, #32]           @ counter = width\r
        subS r6, r6, #1             @ height--\r
        ldr lr, [sp, #28]           @ bufptr3 = tmpline3\r
        add r0, r0, r7              @ src += srcdiff\r
        add r1, r1, r8              @ dst += dstdiff\r
        str r6, [sp, #4]            @ height = r6\r
        bne 100b\r
\r
\r
        ldr r9, [sp, #16]           @ r9 = dststride\r
        ldr r11, [sp, #20]          @ bufptr1 = tmpline1\r
        ldr ip, [sp, #24]           @ bufptr2 = tmpline2\r
        add r4, r1, r9              @ dst2 = dst + dststride\r
\r
    @ last temporary line\r
        neon_scale2x_16_16_line last, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
\r
\r
        ldr sp, [sp]                @ sp = oldsp\r
        pop {r4-r11,lr}\r
        bx lr\r
\r
@ end procedure neon_scale2x_8_16\r
\r
Commit	Line	Data
7127faf3	1	@@\r
	2	@@ Copyright (C) 2012 Roman Pauer\r
	3	@@\r
	4	@@ Permission is hereby granted, free of charge, to any person obtaining a copy of\r
	5	@@ this software and associated documentation files (the "Software"), to deal in\r
	6	@@ the Software without restriction, including without limitation the rights to\r
	7	@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies\r
	8	@@ of the Software, and to permit persons to whom the Software is furnished to do\r
	9	@@ so, subject to the following conditions:\r
	10	@@\r
	11	@@ The above copyright notice and this permission notice shall be included in all\r
	12	@@ copies or substantial portions of the Software.\r
	13	@@\r
	14	@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\r
	15	@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\r
	16	@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\r
	17	@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\r
	18	@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\r
	19	@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\r
	20	@@ SOFTWARE.\r
	21	@@\r
	22	\r
	23	.arm\r
	24	\r
88e59df3	25	#include "neon_scale2x.Sinc"\r
88e59df3	26	#include "neon_normalxx.Sinc"\r
7127faf3	27	\r
	28	.global neon_scale2x_8_8\r
	29	.global neon_scale2x_16_16\r
	30	.global neon_scale2x_8_16\r
	31	\r
	32	.align 4\r
	33	neon_scale2x_8_8:\r
	34	\r
	35	@ r0 = const uint8_t *src\r
	36	@ r1 = uint8_t *dst\r
	37	@ r2 = unsigned int width (pixels)\r
	38	@ r3 = unsigned int srcstride (bytes)\r
	39	@ [sp] = unsigned int dststride (bytes)\r
	40	@ [sp+4] = unsigned int height\r
	41	@ lr = return address\r
	42	\r
	43	ldr ip, [sp] @ ip = dststride\r
	44	push {r4-r9}\r
	45	ldr r9, [sp, #(7*4)] @ r9 = height\r
	46	sub r4, r0, r3 @ r4 = src - srcstride\r
	47	add r5, r0, r3 @ r5 = src + srcstride\r
	48	add r6, r1, ip @ r6 = dst + dststride\r
	49	sub r3, r3, r2 @ r3 = srcstride - width\r
	50	sub ip, ip, r2 @ ip = dststride - width\r
	51	lsl ip, #1 @ ip = 2 * dststride - 2 * width\r
	52	mov r7, r2 @ r7 = width\r
	53	sub r9, r9, #2 @ r9 = height - 2\r
	54	\r
	55	@ r0 = src\r
	56	@ r1 = dst\r
	57	@ r2 = width\r
	58	@ r3 = srcdiff (srcstride - width)\r
	59	@ r4 = src - srcstride\r
	60	@ r5 = src + srcstride\r
	61	@ r6 = dst + dststride\r
	62	@ r7 = counter\r
	63	@ r8 = tmpreg\r
	64	@ r9 = height\r
	65	@ ip = dstdiff (2 * dststride - 2 * width)\r
	66	\r
	67	@ first line\r
	68	neon_scale2x_8_8_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
	69	\r
	70	add r0, r0, r3\r
	71	add r4, r4, r3\r
	72	add r5, r5, r3\r
	73	add r1, r1, ip\r
	74	add r6, r6, ip\r
	75	\r
	76	@ middle lines\r
	77	101:\r
	78	mov r7, r2\r
	79	\r
	80	neon_scale2x_8_8_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
	81	\r
	82	subS r9, r9, #1\r
	83	add r0, r0, r3\r
	84	add r4, r4, r3\r
	85	add r5, r5, r3\r
	86	add r1, r1, ip\r
	87	add r6, r6, ip\r
	88	bne 101b\r
	89	\r
	90	@ last line\r
91	mov r7, r2\r
92	\r
93	neon_scale2x_8_8_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
94	\r
95	pop {r4-r9}\r
96	bx lr\r
97	\r
98	@ end procedure neon_scale2x_8_8\r
99	\r
100	\r
101	neon_scale2x_16_16:\r
102	\r
103	@ r0 = const uint16_t *src\r
104	@ r1 = uint16_t *dst\r
105	@ r2 = unsigned int width (pixels)\r
106	@ r3 = unsigned int srcstride (bytes)\r
107	@ [sp] = unsigned int dststride (bytes)\r
108	@ [sp+4] = unsigned int height\r
109	@ lr = return address\r
110	\r
111	ldr ip, [sp] @ ip = dststride\r
112	push {r4-r9}\r
113	ldr r9, [sp, #(7*4)] @ r9 = height\r
114	sub r4, r0, r3 @ r4 = src - srcstride\r
115	add r5, r0, r3 @ r5 = src + srcstride\r
116	add r6, r1, ip @ r6 = dst + dststride\r
117	sub r3, r3, r2, lsl #1 @ r3 = srcstride - 2 * width\r
118	sub ip, ip, r2, lsl #1 @ ip = dststride - 2 * width\r
119	lsl ip, #1 @ ip = 2 * dststride - 4 * width\r
120	mov r7, r2 @ r7 = width\r
121	sub r9, r9, #2 @ r9 = height - 2\r
122	\r
123	@ r0 = src\r
124	@ r1 = dst\r
125	@ r2 = width\r
126	@ r3 = srcdiff (srcstride - 2 * width)\r
127	@ r4 = src - srcstride\r
128	@ r5 = src + srcstride\r
129	@ r6 = dst + dststride\r
130	@ r7 = counter\r
131	@ r8 = tmpreg\r
132	@ r9 = height\r
133	@ ip = dstdiff (2 * dststride - 4 * width)\r
134	\r
135	@ first line\r
136	neon_scale2x_16_16_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
137	\r
138	add r0, r0, r3\r
139	add r4, r4, r3\r
140	add r5, r5, r3\r
141	add r1, r1, ip\r
142	add r6, r6, ip\r
143	\r
144	@ middle lines\r
145	101:\r
146	mov r7, r2\r
147	\r
148	neon_scale2x_16_16_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
149	\r
150	subS r9, r9, #1\r
151	add r0, r0, r3\r
152	add r4, r4, r3\r
153	add r5, r5, r3\r
154	add r1, r1, ip\r
155	add r6, r6, ip\r
156	bne 101b\r
157	\r
158	@ last line\r
159	mov r7, r2\r
160	\r
161	neon_scale2x_16_16_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0\r
162	\r
163	pop {r4-r9}\r
164	bx lr\r
165	\r
166	@ end procedure neon_scale2x_16_16\r
167	\r
168	\r
169	neon_scale2x_8_16:\r
170	\r
171	@ r0 = const uint8_t *src\r
172	@ r1 = uint8_t *dst\r
173	@ r2 = const uint32_t *palette\r
174	@ r3 = unsigned int width (pixels)\r
175	@ [sp] = unsigned int srcstride (bytes)\r
176	@ [sp+4] = unsigned int dststride (bytes)\r
177	@ [sp+8] = unsigned int height\r
178	@ lr = return address\r
179	\r
180	@ three temporary lines\r
181	\r
182	ldr ip, [sp] @ ip = srcstride\r
183	push {r4-r11,lr}\r
184	ldr r4, [sp, #(4*10)] @ r4 = dststride\r
185	ldr r5, [sp, #(4*11)] @ r5 = height\r
186	mov r6, sp @ r6 = sp\r
187	sub ip, ip, r3 @ ip = srcstride - width\r
188	bic sp, sp, #31 @ align sp to 32 bytes\r
189	sub r7, r4, r3, lsl #1 @ r7 = dststride - 2 * width\r
190	sub sp, sp, r3, lsl #1 @ sp -= 2 * width\r
191	sub r5, r5, #2 @ height -= 2\r
192	mov r10, sp @ tmpline3 = sp\r
193	lsl r7, #1 @ r7 = 2 * dststride - 4 * width\r
194	bic sp, sp, #31 @ align sp to 32 bytes\r
195	sub sp, sp, r3, lsl #1 @ sp -= 2 * width\r
196	mov r11, sp @ tmpline2 = sp\r
197	bic sp, sp, #31 @ align sp to 32 bytes\r
198	sub sp, sp, r3, lsl #1 @ sp -= 2 * width\r
199	mov lr, sp @ tmpline1 = sp\r
200	bic sp, sp, #31 @ align sp to 32 bytes\r
201	sub sp, sp, #36\r
202	str r6, [sp] @ oldsp = r6\r
203	str r5, [sp, #4] @ height = r5\r
204	str ip, [sp, #8] @ srcdiff = ip\r
205	str r7, [sp, #12] @ dstdiff = r7\r
206	str r4, [sp, #16] @ dststride = r4\r
207	str lr, [sp, #20] @ tmpline1 = lr\r
208	str r11, [sp, #24] @ tmpline2 = r11\r
209	str r10, [sp, #28] @ tmpline3 = r10\r
210	str r3, [sp, #32] @ width = r3\r
211	\r
212	@ r0 = src\r
213	@ r1 = dst\r
214	@ r2 = palette\r
215	@ r3 = counter\r
216	@ r4 = dst2\r
217	\r
218	@ r11 = bufptr1\r
219	@ ip = bufptr2\r
220	@ lr = bufptr3\r
221	\r
222	@ [sp] = oldsp\r
223	@ [sp, #4] = height\r
224	@ [sp, #8] = srcdiff (srcstride - width)\r
225	@ [sp, #12] = dstdiff (2 * dststride - 4 * width)\r
226	@ [sp, #16] = dststride\r
227	@ [sp, #20] = tmpline1\r
228	@ [sp, #24] = tmpline2\r
229	@ [sp, #28] = tmpline3\r
230	@ [sp, #32] = width\r
231	\r
232	@ lr = tmpline1\r
233	@ r3 = counter\r
234	\r
235	@ first line\r
236	neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
237	\r
238	ldr r7, [sp, #8] @ r7 = srcdiff\r
239	ldr r3, [sp, #32] @ counter = width\r
240	ldr lr, [sp, #24] @ bufptr3 = tmpline2\r
241	add r0, r0, r7 @ src += srcdiff\r
242	\r
243	@ second line\r
244	neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
245	\r
246	ldr r9, [sp, #16] @ r9 = dststride\r
247	ldr r3, [sp, #32] @ counter = width\r
248	ldr ip, [sp, #20] @ bufptr2 = tmpline1\r
249	ldr lr, [sp, #24] @ bufptr3 = tmpline2\r
250	add r4, r1, r9 @ dst2 = dst + dststride\r
251	\r
252	@ first temporary line\r
253	neon_scale2x_16_16_line first, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
254	\r
255	ldr r7, [sp, #8] @ r7 = srcdiff\r
256	ldr r8, [sp, #12] @ r8 = dstdiff\r
257	ldr r3, [sp, #32] @ counter = width\r
258	ldr lr, [sp, #28] @ bufptr3 = tmpline3\r
259	add r0, r0, r7 @ src += srcdiff\r
260	add r1, r1, r8 @ dst += dstdiff\r
261	\r
262	100:\r
263	\r
264	@ line n+1\r
265	neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip\r
266	\r
267	ldr r9, [sp, #16] @ r9 = dststride\r
268	ldr r11, [sp, #20] @ bufptr1 = tmpline1\r
269	ldr ip, [sp, #24] @ bufptr2 = tmpline2\r
270	ldr lr, [sp, #28] @ bufptr3 = tmpline3\r
271	add r4, r1, r9 @ dst2 = dst + dststride\r
272	ldr r3, [sp, #32] @ counter = width\r
273	str r11, [sp, #28] @ tmpline3 = bufptr1\r
274	str ip, [sp, #20] @ tmpline1 = bufptr2\r
275	str lr, [sp, #24] @ tmpline2 = bufptr3\r
276	\r
277	@ temporary line n\r
278	neon_scale2x_16_16_line middle, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
279	\r
280	ldr r6, [sp, #4] @ r6 = height\r
281	ldr r7, [sp, #8] @ r7 = srcdiff\r
282	ldr r8, [sp, #12] @ r8 = dstdiff\r
283	ldr r3, [sp, #32] @ counter = width\r
284	subS r6, r6, #1 @ height--\r
285	ldr lr, [sp, #28] @ bufptr3 = tmpline3\r
286	add r0, r0, r7 @ src += srcdiff\r
287	add r1, r1, r8 @ dst += dstdiff\r
288	str r6, [sp, #4] @ height = r6\r
289	bne 100b\r
290	\r
291	\r
292	ldr r9, [sp, #16] @ r9 = dststride\r
293	ldr r11, [sp, #20] @ bufptr1 = tmpline1\r
294	ldr ip, [sp, #24] @ bufptr2 = tmpline2\r
295	add r4, r1, r9 @ dst2 = dst + dststride\r
296	\r
297	@ last temporary line\r
298	neon_scale2x_16_16_line last, r11, ip, lr, r3, r1, r4, r5, 1, 0\r
299	\r
300	\r
301	ldr sp, [sp] @ sp = oldsp\r
302	pop {r4-r11,lr}\r
303	bx lr\r
304	\r
305	@ end procedure neon_scale2x_8_16\r
306	\r