diff options
| author | notaz | 2012-10-02 00:47:15 +0300 |
|---|---|---|
| committer | notaz | 2012-10-02 01:06:32 +0300 |
| commit | e38fee1b96c0b904d7f221a349fb2492258d5789 (patch) | |
| tree | 6d44c56db506ddb56584e20939c8c6da08776718 /arm | |
| parent | c95affa70d7025252d63f7d931535f8c10bb483b (diff) | |
| download | picogpsp-e38fee1b96c0b904d7f221a349fb2492258d5789.tar.gz picogpsp-e38fee1b96c0b904d7f221a349fb2492258d5789.tar.bz2 picogpsp-e38fee1b96c0b904d7f221a349fb2492258d5789.zip | |
integrate M-HT's neon scalers
Diffstat (limited to 'arm')
| -rw-r--r-- | arm/neon_eagle2x.S | 337 | ||||
| -rw-r--r-- | arm/neon_eagle2x.Sinc | 761 | ||||
| -rw-r--r-- | arm/neon_eagle2x.h | 44 | ||||
| -rw-r--r-- | arm/neon_normalxx.Sinc | 665 | ||||
| -rw-r--r-- | arm/neon_scale2x.S | 306 | ||||
| -rw-r--r-- | arm/neon_scale2x.Sinc | 474 | ||||
| -rw-r--r-- | arm/neon_scale2x.h | 44 | ||||
| -rw-r--r-- | arm/neon_scale3x.S | 349 | ||||
| -rw-r--r-- | arm/neon_scale3x.Sinc | 976 | ||||
| -rw-r--r-- | arm/neon_scale3x.h | 44 | ||||
| -rw-r--r-- | arm/neon_scale_license.txt | 19 |
11 files changed, 4019 insertions, 0 deletions
diff --git a/arm/neon_eagle2x.S b/arm/neon_eagle2x.S new file mode 100644 index 0000000..aa70021 --- /dev/null +++ b/arm/neon_eagle2x.S @@ -0,0 +1,337 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+.arm
+
+#include "neon_eagle2x.Sinc"
+#include "neon_normalxx.Sinc"
+
+.global neon_eagle2x_8_8
+.global neon_eagle2x_16_16
+.global neon_eagle2x_8_16
+
+.align 4
+neon_eagle2x_8_8:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r10}
+ ldr r9, [sp, #(8*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ mov r10, sp @ oldsp = sp
+ add r5, r0, r3 @ r5 = src + srcstride
+ bic sp, sp, #31 @ align sp to 32 bytes
+ add r6, r1, ip @ r6 = dst + dststride
+ sub sp, sp, #64 @ sp -= 64
+ sub r3, r3, r2 @ r3 = srcstride - width
+ vst1.64 {d8-d11}, [sp:256] @ save q4,q5
+ add r7, sp, #32 @ r7 = sp + 32
+ sub ip, ip, r2 @ ip = dststride - width
+ vst1.64 {d12-d15}, [r7:256] @ save q6,q7
+ lsl ip, #1 @ ip = 2 * dststride - 2 * width
+ mov r7, r2 @ r7 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = counter
+@ r8 = tmpreg
+@ r9 = height
+@ r10 = oldsp
+@ ip = dstdiff (2 * dststride - 2 * width)
+
+ @ first line
+ neon_eagle2x_8_8_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+
+ @ middle lines
+ 101:
+ mov r7, r2
+
+ neon_eagle2x_8_8_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ bne 101b
+
+ @ last line
+ mov r7, r2
+
+ neon_eagle2x_8_8_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add ip, sp, #32 @ ip = sp + 32
+ vld1.64 {d8-d11}, [sp:256] @ restore q4,q5
+ mov sp, r10 @ sp = oldsp
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r10}
+ bx lr
+
+@ end procedure neon_eagle2x_8_8
+
+
+neon_eagle2x_16_16:
+
+@ r0 = const uint16_t *src
+@ r1 = uint16_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r10}
+ ldr r9, [sp, #(8*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ mov r10, sp @ oldsp = sp
+ add r5, r0, r3 @ r5 = src + srcstride
+ bic sp, sp, #31 @ align sp to 32 bytes
+ add r6, r1, ip @ r6 = dst + dststride
+ sub sp, sp, #64 @ sp -= 64
+ sub r3, r3, r2, lsl #1 @ r3 = srcstride - 2 * width
+ vst1.64 {d8-d11}, [sp:256] @ save q4,q5
+ add r7, sp, #32 @ r7 = sp + 32
+ sub ip, ip, r2, lsl #1 @ ip = dststride - 2 * width
+ vst1.64 {d12-d15}, [r7:256] @ save q6,q7
+ lsl ip, #1 @ ip = 2 * dststride - 4 * width
+ mov r7, r2 @ r7 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - 2 * width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = counter
+@ r8 = tmpreg
+@ r9 = height
+@ r10 = oldsp
+@ ip = dstdiff (2 * dststride - 4 * width)
+
+ @ first line
+ neon_eagle2x_16_16_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+
+ @ middle lines
+ 101:
+ mov r7, r2
+
+ neon_eagle2x_16_16_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ bne 101b
+
+ @ last line
+ mov r7, r2
+
+ neon_eagle2x_16_16_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add ip, sp, #32 @ ip = sp + 32
+ vld1.64 {d8-d11}, [sp:256] @ restore q4,q5
+ mov sp, r10 @ sp = oldsp
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r10}
+ bx lr
+
+@ end procedure neon_eagle2x_16_16
+
+
+neon_eagle2x_8_16:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = const uint32_t *palette
+@ r3 = unsigned int width (pixels)
+@ [sp] = unsigned int srcstride (bytes)
+@ [sp+4] = unsigned int dststride (bytes)
+@ [sp+8] = unsigned int height
+@ lr = return address
+
+@ three temporary lines
+
+ ldr ip, [sp] @ ip = srcstride
+ push {r4-r11,lr}
+ ldr r4, [sp, #(4*10)] @ r4 = dststride
+ ldr r5, [sp, #(4*11)] @ r5 = height
+ mov r6, sp @ r6 = sp
+ sub ip, ip, r3 @ ip = srcstride - width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub r7, r4, r3, lsl #1 @ r7 = dststride - 2 * width
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ sub r5, r5, #2 @ height -= 2
+ mov r10, sp @ tmpline3 = sp
+ lsl r7, #1 @ r7 = 2 * dststride - 4 * width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov r11, sp @ tmpline2 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov lr, sp @ tmpline1 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub r8, sp, #64 @ r8 = sp - 64
+ vst1.64 {d8-d11}, [r8:256] @ save q4,q5
+ sub r9, sp, #32 @ r9 = sp - 32
+ vst1.64 {d12-d15}, [r9:256] @ save q6,q7
+ sub sp, sp, #(36 + 64) @ sp -= (36 + 64)
+ str r6, [sp] @ oldsp = r6
+ str r5, [sp, #4] @ height = r5
+ str ip, [sp, #8] @ srcdiff = ip
+ str r7, [sp, #12] @ dstdiff = r7
+ str r4, [sp, #16] @ dststride = r4
+ str lr, [sp, #20] @ tmpline1 = lr
+ str r11, [sp, #24] @ tmpline2 = r11
+ str r10, [sp, #28] @ tmpline3 = r10
+ str r3, [sp, #32] @ width = r3
+
+@ r0 = src
+@ r1 = dst
+@ r2 = palette
+@ r3 = counter
+@ r4 = dst2
+
+@ r11 = bufptr1
+@ ip = bufptr2
+@ lr = bufptr3
+
+@ [sp] = oldsp
+@ [sp, #4] = height
+@ [sp, #8] = srcdiff (srcstride - width)
+@ [sp, #12] = dstdiff (2 * dststride - 4 * width)
+@ [sp, #16] = dststride
+@ [sp, #20] = tmpline1
+@ [sp, #24] = tmpline2
+@ [sp, #28] = tmpline3
+@ [sp, #32] = width
+
+ @ lr = tmpline1
+ @ r3 = counter
+
+ @ first line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r0, r0, r7 @ src += srcdiff
+
+ @ second line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r3, [sp, #32] @ counter = width
+ ldr ip, [sp, #20] @ bufptr2 = tmpline1
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+
+ @ first temporary line
+ neon_eagle2x_16_16_line first, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+
+ 100:
+
+ @ line n+1
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r4, r1, r9 @ dst2 = dst + dststride
+ ldr r3, [sp, #32] @ counter = width
+ str r11, [sp, #28] @ tmpline3 = bufptr1
+ str ip, [sp, #20] @ tmpline1 = bufptr2
+ str lr, [sp, #24] @ tmpline2 = bufptr3
+
+ @ temporary line n
+ neon_eagle2x_16_16_line middle, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+ ldr r6, [sp, #4] @ r6 = height
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ subS r6, r6, #1 @ height--
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+ str r6, [sp, #4] @ height = r6
+ bne 100b
+
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+
+ @ last temporary line
+ neon_eagle2x_16_16_line last, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+
+ add r6, sp, #36 @ r6 = sp + 36
+ ldr sp, [sp] @ sp = oldsp
+ vld1.64 {d8-d11}, [r6:256] @ restore q4,q5
+ add ip, r6, #32 @ ip = r6 + 32
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r11,lr}
+ bx lr
+
+@ end procedure neon_eagle2x_8_16
+
diff --git a/arm/neon_eagle2x.Sinc b/arm/neon_eagle2x.Sinc new file mode 100644 index 0000000..7413312 --- /dev/null +++ b/arm/neon_eagle2x.Sinc @@ -0,0 +1,761 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+
+@ S T U --\ E1 E2
+@ V C W --/ E3 E4
+@ X Y Z
+
+@ q0 = S1sl < S >
+@ q1 = S2sl < V >
+@ q2 = S3sl < X >
+@ q3 = S1sr < U >
+@ q4 = S2sr < W >
+@ q5 = S3sr < Z >
+@ q6 = E3
+@ q7 = E4
+@ q8 = S1
+@ q9 = S2
+@ q10 = S3
+@ q11 = S1prev < T >
+@ q12 = S2prev < C >
+@ q13 = S3prev < Y >
+@ q14 = E1
+@ q15 = E2
+
+
+.macro __neon_eagle2x_8_8_line src1, src2, src3, counter, dst1, dst2, reg1, qT, qY, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+
+ .ifeqs "\qT", "q11"
+ vld1.8 {d23[7]}, [\src1] @ S1prev[15] = src[-srcstride]
+ .endif
+ vld1.8 {d25[7]}, [\src2] @ S2prev[15] = src[0]
+ .ifeqs "\qY", "q13"
+ vld1.8 {d27[7]}, [\src3] @ S3prev[15] = src[srcstride]
+ .endif
+ andS \reg1, \counter, #15 @ reg1 = counter & 15
+
+ .ifnes "\qT", "q11"
+ add \src1, \src1, \counter @ src1 += counter
+ .endif
+ .ifnes "\qY", "q13"
+ add \src3, \src3, \counter @ src3 += counter
+ .endif
+ beq 1f
+
+ @ first 1-15 pixels - align counter to 16 bytes
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vld1.8 {q8}, [\src1], \reg1 @ S1 = [src - srcstride]; src1 += counter & 15
+ .endif
+
+ vld1.8 {q9}, [\src2], \reg1 @ S2 = [src ]; src2 += counter & 15
+
+ .ifeqs "\qY", "q13"
+ vld1.8 {q10}, [\src3], \reg1 @ S3 = [src + srcstride]; src3 += counter & 15
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < S >
+
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < V >
+
+ vmov q12, q9 @ S2prev = S2 < C >
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < X >
+
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q3, \qT, q8, #1 @ S1sr = (S1prev >> 8) | ... < U >
+ .endif
+
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | ... < W >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q5, \qY, q10, #1 @ S3sr = (S3prev >> 8) | ... < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ vceq.i8 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i8 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i8 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i8 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i8 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i8 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i8 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i8 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ .ifeqs "\qT", "q11"
+ sub \reg1, \src1, #1
+ .else
+ sub \reg1, \src2, #1
+ .endif
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ .ifeqs "\qT", "q11"
+ vld1.8 {d23[7]}, [\reg1] @ S1prev[15] = src[counter & 15 - 1 - srcstride]
+
+ sub \reg1, \src2, #1
+ .endif
+
+ vld1.8 {d25[7]}, [\reg1] @ S2prev[15] = src[counter & 15 - 1]
+
+ .ifeqs "\qY", "q13"
+ sub \reg1, \src3, #1
+
+ vld1.8 {d27[7]}, [\reg1] @ S3prev[15] = src[counter & 15 - 1 + srcstride]
+ .endif
+
+ ubfx \reg1, \counter, #0, #4 @ reg1 = counter & 15
+
+ lsl \reg1, #1
+
+ vst2.8 {q14-q15}, [\dst1],\reg1 @ [dst] = E1,E2; dst1 += reg1
+
+ bic \counter, \counter, #15
+
+ vst2.8 {q6-q7}, [\dst2], \reg1 @ [dst + dststride] = E3,E4; dst2 += reg1
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ .ifeqs "\qT", "q11"
+ vld1.8 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 16
+ .endif
+ vld1.8 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 16
+ .ifeqs "\qY", "q13"
+ vld1.8 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 16
+ .endif
+
+ @ inner loop (16 pixels per iteration)
+ 2:
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < S >
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < V >
+ vmov q12, q9 @ S2prev = S2 < C >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < X >
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+
+ .ifeqs "\qT", "q11"
+ vld1.8 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 16
+ vext.8 q3, \qT, q8, #1 @ S1sr = (S1prev >> 8) | S1[0] < U >
+ .endif
+
+ vld1.8 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 16
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | S2[0] < W >
+
+ .ifeqs "\qY", "q13"
+ vld1.8 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 16
+ vext.8 q5, \qY, q10, #1 @ S3sr = (S3prev >> 8) | S3[0] < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ sub \counter, \counter, #16 @ counter -= 16
+ vceq.i8 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i8 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i8 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i8 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i8 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i8 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i8 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i8 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ vst2.8 {q14-q15}, [\aldst1]! @ [dst] = E1,E2; dst1 += 2*16
+
+ cmp \counter, #16
+
+ vst2.8 {q6-q7}, [\aldst2]! @ [dst + dststride] = E3,E4; dst2 += 2*16
+ bhi 2b
+
+ @ last 16 pixels
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < S >
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < V >
+ vmov q12, q9 @ S2prev = S2 < C >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < X >
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+
+ .ifeqs "\qT", "q11"
+ vshr.u64 d16, d17, #(64-8) @ S1[0] = S1[15] | ...
+ .endif
+
+ vshr.u64 d18, d19, #(64-8) @ S2[0] = S2[15] | ...
+
+ .ifeqs "\qY", "q13"
+ vshr.u64 d20, d21, #(64-8) @ S3[0] = S3[15] | ...
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q3, \qT, q8, #1 @ S1sr = (S1prev >> 8) | S1[0] < U >
+ .endif
+
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | S2[0] < W >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q5, \qY, q10, #1 @ S3sr = (S3prev >> 8) | S3[0] < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ vceq.i8 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i8 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i8 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i8 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i8 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i8 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i8 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i8 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ vst2.8 {q14-q15}, [\aldst1]! @ [dst] = E1,E2; dst1 += 2*16
+
+ vst2.8 {q6-q7}, [\aldst2]! @ [dst + dststride] = E3,E4; dst2 += 2*16
+
+.endm
+
+.macro _neon_eagle2x_8_8_line_first src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q12, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_eagle2x_8_8_line_middle src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_eagle2x_8_8_line_last src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q12, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro neon_eagle2x_8_8_line part, src1, src2, src3, counter, dst1, dst2, reg1, srcalign16, dstalign32
+ .ifeq \srcalign16
+
+ .ifeq \dstalign32
+ _neon_eagle2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1, \dst2
+ .else
+ _neon_eagle2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1:256, \dst2:256
+ .endif
+
+ .else
+
+ .ifeq \dstalign32
+ _neon_eagle2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2
+ .else
+ _neon_eagle2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1:256, \dst2:256
+ .endif
+
+ .endif
+.endm
+
+
+.macro __neon_eagle2x_16_16_line src1, src2, src3, counter, dst1, dst2, reg1, qT, qY, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+
+ .ifeqs "\qT", "q11"
+ vld1.16 {d23[3]}, [\src1] @ S1prev[7] = src[-srcstride]
+ .endif
+ vld1.16 {d25[3]}, [\src2] @ S2prev[7] = src[0]
+ .ifeqs "\qY", "q13"
+ vld1.16 {d27[3]}, [\src3] @ S3prev[7] = src[srcstride]
+ .endif
+ andS \reg1, \counter, #7 @ reg1 = counter & 7
+
+ .ifnes "\qT", "q11"
+ add \src1, \src1, \counter, lsl #1 @ src1 += 2 * counter
+ .endif
+ .ifnes "\qY", "q13"
+ add \src3, \src3, \counter, lsl #1 @ src3 += 2 * counter
+ .endif
+ beq 1f
+
+ @ first 1-7 pixels - align counter to 16 bytes
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vld1.16 {q8}, [\src1] @ S1 = [src - srcstride]
+ add \src1, \src1, \reg1, lsl #1 @ src1 += 2 * (counter & 7)
+ .endif
+
+ vld1.16 {q9}, [\src2] @ S2 = [src ]
+ add \src2, \src2, \reg1, lsl #1 @ src2 += 2 * (counter & 7)
+
+ .ifeqs "\qY", "q13"
+ vld1.16 {q10}, [\src3] @ S3 = [src + srcstride]
+ add \src3, \src3, \reg1, lsl #1 @ src3 += 2 * (counter & 7)
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < S >
+
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < V >
+
+ vmov q12, q9 @ S2prev = S2 < C >
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < X >
+
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q3, \qT, q8, #2 @ S1sr = (S1prev >> 16) | ... < U >
+ .endif
+
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | ... < W >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q5, \qY, q10, #2 @ S3sr = (S3prev >> 16) | ... < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ vceq.i16 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i16 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i16 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i16 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i16 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i16 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i16 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i16 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ .ifeqs "\qT", "q11"
+ sub \reg1, \src1, #2
+ .else
+ sub \reg1, \src2, #2
+ .endif
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ .ifeqs "\qT", "q11"
+ vld1.16 {d23[3]}, [\reg1] @ S1prev[7] = src[2 * (counter & 7) - 2 - srcstride]
+
+ sub \reg1, \src2, #2
+ .endif
+
+ vld1.16 {d25[3]}, [\reg1] @ S2prev[7] = src[2 * (counter & 7) - 2]
+
+ .ifeqs "\qY", "q13"
+ sub \reg1, \src3, #2
+
+ vld1.16 {d27[3]}, [\reg1] @ S3prev[7] = src[2 * (counter & 7) - 2 + srcstride]
+ .endif
+
+ ubfx \reg1, \counter, #0, #3 @ reg1 = counter & 7
+
+ lsl \reg1, #2
+
+ vst2.16 {q14-q15}, [\dst1], \reg1 @ [dst] = E1,E2; dst1 += reg1
+
+ bic \counter, \counter, #7
+
+ vst2.16 {q6-q7}, [\dst2], \reg1 @ [dst + dststride] = E3,E4; dst2 += reg1
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ .ifeqs "\qT", "q11"
+ vld1.16 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 2*8
+ .endif
+ vld1.16 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 2*8
+ .ifeqs "\qY", "q13"
+ vld1.16 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 2*8
+ .endif
+
+ @ inner loop (8 pixels per iteration)
+ 2:
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < S >
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < V >
+ vmov q12, q9 @ S2prev = S2 < C >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < X >
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+
+ .ifeqs "\qT", "q11"
+ vld1.16 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 2*8
+ vext.8 q3, \qT, q8, #2 @ S1sr = (S1prev >> 16) | S1[0] < U >
+ .endif
+
+ vld1.16 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 2*8
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | S2[0] < W >
+
+ .ifeqs "\qY", "q13"
+ vld1.16 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 2*8
+ vext.8 q5, \qY, q10, #2 @ S3sr = (S3prev >> 16) | S3[0] < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ sub \counter, \counter, #8 @ counter -= 8
+ vceq.i16 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i16 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i16 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i16 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i16 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i16 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i16 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i16 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ vst2.16 {q14-q15}, [\aldst1]! @ [dst] = E1,E2; dst1 += 2*2*8
+
+ cmp \counter, #8
+
+ vst2.16 {q6-q7}, [\aldst2]! @ [dst + dststride] = E3,E4; dst2 += 2*2*8
+ bhi 2b
+
+ @ last 8 pixels
+
+@ q0 = S1sl < S >
+@ q2 = S3sl < X >
+@ q7 = tmp2
+@ q15 = tmp1
+
+ .ifeqs "\qT", "q11"
+ vext.8 q0, \qT, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < S >
+ vmov \qT, q8 @ S1prev = S1 < T >
+ .endif
+
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < V >
+ vmov q12, q9 @ S2prev = S2 < C >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q2, \qY, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < X >
+ vmov \qY, q10 @ S3prev = S3 < Y >
+ .endif
+
+ .ifeqs "\qT", "q11"
+ vshr.u64 d16, d17, #(64-16) @ S1[0] = S1[7] | ...
+ .endif
+
+ vshr.u64 d18, d19, #(64-16) @ S2[0] = S2[7] | ...
+
+ .ifeqs "\qY", "q13"
+ vshr.u64 d20, d21, #(64-16) @ S3[0] = S3[7] | ...
+ .endif
+ .ifeqs "\qT", "q11"
+ vext.8 q3, \qT, q8, #2 @ S1sr = (S1prev >> 16) | S1[0] < U >
+ .endif
+
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | S2[0] < W >
+
+ .ifeqs "\qY", "q13"
+ vext.8 q5, \qY, q10, #2 @ S3sr = (S3prev >> 16) | S3[0] < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qT", "q11"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .endif
+
+ vceq.i16 q14, q0, \qT @ E1 = < S == T >
+
+ vceq.i16 q15, q0, q1 @ tmp1 = < S == V >
+
+ vceq.i16 q6, q2, \qY @ E3 = < X == Y >
+
+ vceq.i16 q7, q2, q1 @ tmp2 = < X == V >
+
+ vand q14, q14, q15 @ E1 = < S == T && S == V >
+
+@ q0 = tmp3
+@ q15 = E2
+
+ vceq.i16 q15, q3, \qT @ E2 = < U == T >
+
+ vceq.i16 q0, q3, q4 @ tmp3 = < U == W >
+
+ vand q6, q6, q7 @ E3 = < X == Y && X == V >
+
+@ q2 = tmp4
+@ q7 = E4
+ vceq.i16 q7, q5, \qY @ E4 = < Z == Y >
+
+ vceq.i16 q2, q5, q4 @ tmp4 = < Z == W >
+
+ vand q15, q15, q0 @ E2 = < U == T && U == W >
+
+ vbsl q14, \qT, q12 @ E1 = < (S == T && S == V) ? T : C >
+
+ vbsl q15, \qT, q12 @ E2 = < (U == T && U == W) ? T : C >
+
+ vand q7, q7, q2 @ E4 = < Z == Y && Z == W >
+
+ vbsl q6, \qY, q12 @ E3 = < (X == Y && X == V) ? Y : C >
+
+ vbsl q7, \qY, q12 @ E4 = < (Z == Y && Z == W) ? Y : C >
+ vst2.16 {q14-q15}, [\aldst1]! @ [dst] = E1,E2; dst1 += 2*2*8
+
+ vst2.16 {q6-q7}, [\aldst2]! @ [dst + dststride] = E3,E4; dst2 += 2*2*8
+
+.endm
+
+.macro _neon_eagle2x_16_16_line_first src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q12, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_eagle2x_16_16_line_middle src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_eagle2x_16_16_line_last src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_eagle2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q12, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro neon_eagle2x_16_16_line part, src1, src2, src3, counter, dst1, dst2, reg1, srcalign16, dstalign32
+ .ifeq \srcalign16
+
+ .ifeq \dstalign32
+ _neon_eagle2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1, \dst2
+ .else
+ _neon_eagle2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1:256, \dst2:256
+ .endif
+
+ .else
+
+ .ifeq \dstalign32
+ _neon_eagle2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2
+ .else
+ _neon_eagle2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1:256, \dst2:256
+ .endif
+
+ .endif
+.endm
+
diff --git a/arm/neon_eagle2x.h b/arm/neon_eagle2x.h new file mode 100644 index 0000000..485a0d1 --- /dev/null +++ b/arm/neon_eagle2x.h @@ -0,0 +1,44 @@ +/** + * + * Copyright (C) 2012 Roman Pauer + * + * Permission is hereby granted, free of charge, to any person obtaining a copy of + * this software and associated documentation files (the "Software"), to deal in + * the Software without restriction, including without limitation the rights to + * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies + * of the Software, and to permit persons to whom the Software is furnished to do + * so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#if !defined(_NEON_EAGLE2X_H_INCLUDED_) +#define _NEON_EAGLE2X_H_INCLUDED_ + +#include <inttypes.h> + +#ifdef __cplusplus +extern "C" { +#endif + +extern void neon_eagle2x_8_8(const uint8_t *src, uint8_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); +extern void neon_eagle2x_16_16(const uint16_t *src, uint16_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +extern void neon_eagle2x_8_16(const uint8_t *src, uint16_t *dst, const uint32_t *palette, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +#ifdef __cplusplus +} +#endif + + +#endif /* _NEON_EAGLE2X_H_INCLUDED_ */ diff --git a/arm/neon_normalxx.Sinc b/arm/neon_normalxx.Sinc new file mode 100644 index 0000000..fcbcfd4 --- /dev/null +++ b/arm/neon_normalxx.Sinc @@ -0,0 +1,665 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+
+
+.macro _neon_normalxx_8_16_line_middle src, dst, pal, counter, reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8, reg9, dststride, dA, dB
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+
+ ldr \reg2, [\src, #4] @ reg2 = src[4-7]
+
+ ldr \reg3, [\src, #8] @ reg3 = src[8-11]
+
+ ldr \reg4, [\src, #12] @ reg4 = src[12-15]
+ ubfx \reg5, \reg1, #0, #8 @ reg5 = src[0]
+
+ ldr \reg5, [\pal, \reg5, lsl #2] @ reg5 = pal[src[0]]
+ ubfx \reg6, \reg1, #8, #8 @ reg6 = src[1]
+
+ ldr \reg6, [\pal, \reg6, lsl #2] @ reg6 = pal[src[1]]
+ ubfx \reg7, \reg1, #16, #8 @ reg7 = src[2]
+
+ ldr \reg7, [\pal, \reg7, lsl #2] @ reg7 = pal[src[2]]
+ lsr \reg1, \reg1, #24 @ reg1 = src[3]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[3]]
+ ubfx \reg8, \reg2, #0, #8 @ reg8 = src[4]
+
+ ldr \reg8, [\pal, \reg8, lsl #2] @ reg8 = pal[src[4]]
+ ubfx \reg9, \reg2, #8, #8 @ reg9 = src[5]
+
+ ldr \reg9, [\pal, \reg9, lsl #2] @ reg9 = pal[src[5]]
+ bfi \reg5, \reg6, #16, #16 @ reg5 = pal[src[0]] | pal[src[1]] << 16
+
+ bfi \reg7, \reg1, #16, #16 @ reg7 = pal[src[2]] | pal[src[3]] << 16
+ ubfx \reg6, \reg2, #16, #8 @ reg6 = src[6]
+
+ vmov d16, \reg5, \reg7 @ d16 = pal[src[0-3]]
+ lsr \reg2, \reg2, #24 @ reg2 = src[7]
+
+ ldr \reg6, [\pal, \reg6, lsl #2] @ reg6 = pal[src[6]]
+ bfi \reg8, \reg9, #16, #16 @ reg8 = pal[src[4]] | pal[src[5]] << 16
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[7]]
+ ubfx \reg1, \reg3, #0, #8 @ reg1 = src[8]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[8]]
+ ubfx \reg5, \reg3, #8, #8 @ reg5 = src[9]
+
+ ldr \reg5, [\pal, \reg5, lsl #2] @ reg5 = pal[src[9]]
+ ubfx \reg7, \reg3, #16, #8 @ reg7 = src[10]
+
+ ldr \reg7, [\pal, \reg7, lsl #2] @ reg7 = pal[src[10]]
+ bfi \reg6, \reg2, #16, #16 @ reg6 = pal[src[6]] | pal[src[7]] << 16
+
+ vmov d17, \reg8, \reg6 @ d17 = pal[src[4-7]]
+ lsr \reg3, \reg3, #24 @ reg3 = src[11]
+
+ ldr \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[11]]
+ ubfx \reg2, \reg4, #0, #8 @ reg2 = src[12]
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[12]]
+ ubfx \reg6, \reg4, #8, #8 @ reg6 = src[13]
+
+ ldr \reg6, [\pal, \reg6, lsl #2] @ reg6 = pal[src[13]]
+ ubfx \reg8, \reg4, #16, #8 @ reg8 = src[14]
+
+ ldr \reg8, [\pal, \reg8, lsl #2] @ reg8 = pal[src[14]]
+ lsr \reg4, \reg4, #24 @ reg4 = src[15]
+
+ ldr \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[src[15]]
+ bfi \reg1, \reg5, #16, #16 @ reg1 = pal[src[8]] | pal[src[9]] << 16
+
+ bfi \reg7, \reg3, #16, #16 @ reg7 = pal[src[10]] | pal[src[11]] << 16
+ bfi \reg2, \reg6, #16, #16 @ reg2 = pal[src[12]] | pal[src[13]] << 16
+
+ vmov \dA, \reg1, \reg7 @ dA = pal[src[8-11]]
+ sub \counter, \counter, #16 @ counter -= 16
+
+ bfi \reg8, \reg4, #16, #16 @ reg8 = pal[src[14]] | pal[src[15]] << 16
+ add \src, \src, #16 @ src += 16
+
+ vmov \dB, \reg2, \reg8 @ dB = pal[src[12-15]]
+ cmp \counter, #16
+.endm
+
+.macro neon_normal1x_8_16_line src, dst, pal, counter, reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8, reg9
+ @ align src to 4 bytes
+ andS \reg5, \src, #3 @ reg5 = src & 3
+ beq 10f
+
+ @ first 1-3 pixels
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+ rsb \reg5, \reg5, #4 @ reg5 = 4 - (src & 3)
+
+ add \src, \src, \reg5 @ src += reg5
+ sub \counter, \counter, \reg5 @ counter -= reg5
+
+ subS \reg5, \reg5, #1 @ reg5--
+
+ ubfx \reg2, \reg1, #0, #8 @ reg2 = src[0]
+ ubfxne \reg3, \reg1, #8, #8 @ reg3 = src[1]
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[reg2]
+
+ ldrne \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[reg3]
+
+ strh \reg2, [\dst] @ dst[0] = reg2
+
+ strneh \reg3, [\dst, #2]! @ dst[1] = reg3; dst++
+ subneS \reg5, \reg5, #1 @ reg5--
+
+ ubfxne \reg4, \reg1, #16, #8 @ reg4 = src[2]
+ add \dst, \dst, #2 @ dst++
+
+ ldrne \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[reg4]
+
+ strneh \reg4, [\dst], #2 @ dst[2] = reg4; dst++
+
+ @ middle pixels (16 per iteration)
+ 10:
+ _neon_normalxx_8_16_line_middle \src, \dst, \pal, \counter, \reg1, \reg2, \reg3, \reg4, \reg5, \reg6, \reg7, \reg8, \reg9, , d18, d19
+
+ vst1.16 {d16-d19}, [\dst]! @ dst[0-15] = d16-d19; dst += 2*16
+ bhs 10b
+
+ @ last 0-15 bytes
+
+ cmp \counter, #0
+ beq 40f
+
+ cmp \counter, #4
+ blo 30f
+
+ @ 4-12 pixels (4 pre iteration)
+ 20:
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+ sub \counter, \counter, #4 @ counter -= 4
+
+ add \src, \src, #4 @ src += 4
+ add \dst, \dst, #(2*4) @ dst += 4
+
+ ubfx \reg2, \reg1, #0, #8 @ reg2 = src[0]
+ cmp \counter, #4
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[0]]
+ ubfx \reg3, \reg1, #8, #8 @ reg3 = src[1]
+
+ ldr \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[1]]
+ ubfx \reg4, \reg1, #16, #8 @ reg4 = src[2]
+
+ ldr \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[src[2]]
+ lsr \reg1, \reg1, #24 @ reg1 = src[3]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[3]]
+
+ strh \reg2, [\dst, #-8] @ dst[0] = reg2
+
+ strh \reg3, [\dst, #-6] @ dst[1] = reg3
+
+ strh \reg4, [\dst, #-4] @ dst[2] = reg4
+
+ strh \reg1, [\dst, #-2] @ dst[3] = reg1
+ bhs 20b
+
+ cmp \counter, #0
+ beq 40f
+
+ @ last 1-3 pixels
+ 30:
+ ldrb \reg1, [\src] @ reg1 = src[0]
+ subS \counter, \counter, #1 @ counter--
+
+ ldrneb \reg2, [\src, #1]! @ reg2 = src[1]; src++
+
+ add \src, \src, #1 @ src++
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+
+ ldrne \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[1]]
+
+ strh \reg1, [\dst] @ dst[0] = reg1
+
+ strneh \reg2, [\dst, #2]! @ dst[1] = reg2; dst++
+ subneS \counter, \counter, #1 @ counter--
+
+ ldrneb \reg3, [\src], #1 @ reg3 = src[2]; src++
+ add \dst, \dst, #2 @ dst++
+
+ ldrne \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[2]]
+
+ strneh \reg3, [\dst], #2 @ dst[2] = reg3; dst++
+
+ 40:
+.endm
+
+.macro neon_normal2x_8_16_line src, dst, pal, counter, reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8, reg9, dststride
+ @ align src to 4 bytes
+ andS \reg5, \src, #3 @ reg5 = src & 3
+ beq 10f
+
+ @ first 1-3 pixels
+ rsb \reg5, \reg5, #4 @ reg5 = 4 - (src & 3)
+ 1:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \dst, \dst, #4 @ dst += 2*2
+ sub \counter, \counter, #1 @ counter--
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ subS \reg5, \reg5, #1 @ reg5--
+
+ strh \reg1, [\dst, #-4] @ dst[0] = reg1
+
+ strh \reg1, [\dst, #-2] @ dst[1] = reg1
+
+ strh \reg1, [\reg2] @ dst1[0] = reg1
+
+ strh \reg1, [\reg2, #2] @ dst1[1] = reg1
+ bne 1b
+
+ @ middle pixels (16 per iteration)
+ 10:
+ _neon_normalxx_8_16_line_middle \src, \dst, \pal, \counter, \reg1, \reg2, \reg3, \reg4, \reg5, \reg6, \reg7, \reg8, \reg9, \dststride, d20, d21
+
+ vmov q9, q8
+ add \reg1, \dst, \dststride @ reg1 = dst + dststride
+
+ vmov q11, q10
+ vst2.16 {q8,q9}, [\dst]! @ dst[0-7] = q8-q9; dst += 2*2*8
+
+ vst2.16 {q10,q11}, [\dst]! @ dst[8-15] = q10-q11; dst += 2*2*8
+
+ vst2.16 {q8,q9}, [\reg1]! @ dst1[0-7] = q8-q9; dst1 += 2*2*8
+
+ vst2.16 {q10,q11}, [\reg1]! @ dst1[8-15] = q10-q11; dst1 += 2*2*8
+ bhs 10b
+
+ @ last 0-15 bytes
+
+ cmp \counter, #0
+ beq 40f
+
+ cmp \counter, #4
+ blo 30f
+
+ @ 4-12 pixels (4 pre iteration)
+ 20:
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+ sub \counter, \counter, #4 @ counter -= 4
+
+ add \src, \src, #4 @ src += 4
+
+ ubfx \reg2, \reg1, #0, #8 @ reg2 = src[0]
+ cmp \counter, #4
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[0]]
+ ubfx \reg3, \reg1, #8, #8 @ reg3 = src[1]
+
+ ldr \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[1]]
+ ubfx \reg4, \reg1, #16, #8 @ reg4 = src[2]
+
+ ldr \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[src[2]]
+ lsr \reg1, \reg1, #24 @ reg1 = src[3]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[3]]
+
+ add \reg5, \dst, \dststride
+ bfi \reg2, \reg3, #16, #16 @ reg2 = reg2 | reg3 << 16
+
+ vmov.32 d16[0], \reg2
+
+ bfi \reg4, \reg1, #16, #16 @ reg4 = reg4 | reg1 << 16
+
+ vmov.32 d16[1], \reg4
+
+ vmov d17, d16
+
+ vst2.16 {d16,d17}, [\dst]! @ dst[0-7] = d16-d17; dst += 2*2*4
+
+ vst2.16 {d16,d17}, [\reg5] @ dst1[0-7] = d16-d17
+ bhs 20b
+
+ cmp \counter, #0
+ beq 40f
+
+ @ last 1-3 pixels
+ 30:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \dst, \dst, #4 @ dst += 2*2
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ subS \counter, \counter, #1 @ counter--
+
+ strh \reg1, [\dst, #-4] @ dst[0] = reg1
+
+ strh \reg1, [\dst, #-2] @ dst[1] = reg1
+
+ strh \reg1, [\reg2] @ dst1[0] = reg1
+
+ strh \reg1, [\reg2, #2] @ dst1[1] = reg1
+ bne 30b
+
+ 40:
+.endm
+
+.macro neon_normal3x_8_16_line src, dst, pal, counter, reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8, reg9, dststride
+ @ align src to 4 bytes
+ andS \reg5, \src, #3 @ reg5 = src & 3
+ beq 10f
+
+ @ first 1-3 pixels
+ rsb \reg5, \reg5, #4 @ reg5 = 4 - (src & 3)
+ 1:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \reg3, \reg2, \dststride
+ add \dst, \dst, #6 @ dst += 3*2
+
+ sub \counter, \counter, #1 @ counter--
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ subS \reg5, \reg5, #1 @ reg5--
+
+ strh \reg1, [\dst, #-6] @ dst[0] = reg1
+
+ strh \reg1, [\dst, #-4] @ dst[1] = reg1
+
+ strh \reg1, [\dst, #-2] @ dst[2] = reg1
+ bfi \reg1, \reg1, #16, #16 @ reg1 = reg1 | reg1 << 16
+
+ strh \reg1, [\reg2] @ dst1[0] = reg1
+
+ str \reg1, [\reg2, #2] @ dst1[1-2] = reg1
+
+ strh \reg1, [\reg3] @ dst2[0] = reg1
+
+ str \reg1, [\reg3, #2] @ dst2[1-2] = reg1
+ bne 1b
+
+ @ middle pixels (16 per iteration)
+ 10:
+ _neon_normalxx_8_16_line_middle \src, \dst, \pal, \counter, \reg1, \reg2, \reg3, \reg4, \reg5, \reg6, \reg7, \reg8, \reg9, \dststride, d22, d23
+
+ vmov q9, q8
+ add \reg1, \dst, \dststride @ reg1 = dst + dststride
+
+ vmov q10, q8
+ add \reg2, \dst, \dststride, lsl #1 @ reg1 = dst + 2 * dststride
+
+ vmov q12, q11
+ vst3.16 {d16,d18,d20}, [\dst]! @ dst[0-3] = q8-q10[0]; dst += 3*2*4
+
+ vmov q13, q11
+ vst3.16 {d17,d19,d21}, [\dst]! @ dst[4-7] = q8-q10[1]; dst += 3*2*4
+
+ vst3.16 {d22,d24,d26}, [\dst]! @ dst[8-11] = q11-q13[0]; dst += 3*2*4
+
+ vst3.16 {d23,d25,d27}, [\dst]! @ dst[12-15] = q11-q13[1]; dst += 3*2*4
+
+ vst3.16 {d16,d18,d20}, [\reg1]! @ dst1[0-3] = q8-q10[0]; dst1 += 3*2*4
+
+ vst3.16 {d17,d19,d21}, [\reg1]! @ dst1[4-7] = q8-q10[1]; dst1 += 3*2*4
+
+ vst3.16 {d22,d24,d26}, [\reg1]! @ dst1[8-11] = q11-q13[0]; dst1 += 3*2*4
+
+ vst3.16 {d23,d25,d27}, [\reg1]! @ dst1[12-15] = q11-q13[1]; dst1 += 3*2*4
+
+ vst3.16 {d16,d18,d20}, [\reg2]! @ dst2[0-3] = q8-q10[0]; dst2 += 3*2*4
+
+ vst3.16 {d17,d19,d21}, [\reg2]! @ dst2[4-7] = q8-q10[1]; dst2 += 3*2*4
+
+ vst3.16 {d22,d24,d26}, [\reg2]! @ dst2[8-11] = q11-q13[0]; dst2 += 3*2*4
+
+ vst3.16 {d23,d25,d27}, [\reg2]! @ dst2[12-15] = q11-q13[1]; dst2 += 3*2*4
+ bhs 10b
+
+ @ last 0-15 bytes
+
+ cmp \counter, #0
+ beq 40f
+
+ cmp \counter, #4
+ blo 30f
+
+ @ 4-12 pixels (4 pre iteration)
+ 20:
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+ sub \counter, \counter, #4 @ counter -= 4
+
+ add \src, \src, #4 @ src += 4
+
+ ubfx \reg2, \reg1, #0, #8 @ reg2 = src[0]
+ cmp \counter, #4
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[0]]
+ ubfx \reg3, \reg1, #8, #8 @ reg3 = src[1]
+
+ ldr \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[1]]
+ ubfx \reg4, \reg1, #16, #8 @ reg4 = src[2]
+
+ ldr \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[src[2]]
+ lsr \reg1, \reg1, #24 @ reg1 = src[3]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[3]]
+
+ add \reg5, \dst, \dststride
+ bfi \reg2, \reg3, #16, #16 @ reg2 = reg2 | reg3 << 16
+
+ vmov.32 d16[0], \reg2
+ add \reg6, \reg5, \dststride
+
+ bfi \reg4, \reg1, #16, #16 @ reg4 = reg4 | reg1 << 16
+
+ vmov.32 d16[1], \reg4
+
+ vmov d17, d16
+
+ vmov d18, d16
+
+ vst3.16 {d16,d17,d18}, [\dst]! @ dst[0-11] = d16-d18; dst += 3*2*4
+
+ vst3.16 {d16,d17,d18}, [\reg5] @ dst1[0-11] = d16-d18
+
+ vst3.16 {d16,d17,d18}, [\reg6] @ dst2[0-11] = d16-d18
+ bhs 20b
+
+ cmp \counter, #0
+ beq 40f
+
+ @ last 1-3 pixels
+ 30:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \reg3, \reg2, \dststride
+ add \dst, \dst, #6 @ dst += 3*2
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ subS \counter, \counter, #1 @ counter--
+
+ strh \reg1, [\dst, #-6] @ dst[0] = reg1
+
+ strh \reg1, [\dst, #-4] @ dst[1] = reg1
+
+ strh \reg1, [\dst, #-2] @ dst[2] = reg1
+ bfi \reg1, \reg1, #16, #16 @ reg1 = reg1 | reg1 << 16
+
+ strh \reg1, [\reg2] @ dst1[0] = reg1
+
+ str \reg1, [\reg2, #2] @ dst1[1-2] = reg1
+
+ strh \reg1, [\reg3] @ dst2[0] = reg1
+
+ str \reg1, [\reg3, #2] @ dst2[1-2] = reg1
+ bne 30b
+
+ 40:
+.endm
+
+.macro neon_normal4x_8_16_line src, dst, pal, counter, reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8, reg9, dststride
+ @ align src to 4 bytes
+ andS \reg5, \src, #3 @ reg5 = src & 3
+ beq 10f
+
+ @ first 1-3 pixels
+ rsb \reg5, \reg5, #4 @ reg5 = 4 - (src & 3)
+ 1:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \reg3, \reg2, \dststride
+ add \dst, \dst, #8 @ dst += 4*2
+
+ sub \counter, \counter, #1 @ counter--
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ add \reg4, \reg3, \dststride
+
+ strh \reg1, [\dst, #-8] @ dst[0] = reg1
+ subS \reg5, \reg5, #1 @ reg5--
+
+ strh \reg1, [\dst, #-6] @ dst[1] = reg1
+
+ bfi \reg1, \reg1, #16, #16 @ reg1 = reg1 | reg1 << 16
+ str \reg1, [\dst, #-4] @ dst[2-3] = reg1
+
+ str \reg1, [\reg2] @ dst1[0-1] = reg1
+
+ str \reg1, [\reg2, #4] @ dst1[2-3] = reg1
+
+ str \reg1, [\reg3] @ dst2[0-1] = reg1
+
+ str \reg1, [\reg3, #4] @ dst2[2-3] = reg1
+
+ str \reg1, [\reg4] @ dst3[0-1] = reg1
+
+ str \reg1, [\reg4, #4] @ dst3[2-3] = reg1
+ bne 1b
+
+ @ middle pixels (16 per iteration)
+ 10:
+ _neon_normalxx_8_16_line_middle \src, \dst, \pal, \counter, \reg1, \reg2, \reg3, \reg4, \reg5, \reg6, \reg7, \reg8, \reg9, \dststride, d24, d25
+
+ vmov q9, q8
+ add \reg1, \dst, \dststride @ reg1 = dst + dststride
+
+ vmov q10, q8
+ add \reg2, \dst, \dststride, lsl #1 @ reg2 = dst + 2 * dststride
+
+ vmov q11, q8
+ add \reg3, \reg1, \dststride,lsl #1 @ reg3 = dst + 3 * dststride
+
+ vmov q13, q12
+ vst4.16 {d16,d18,d20,d22}, [\dst]! @ dst[0-3] = q8-q11[0]; dst += 4*2*4
+
+ vmov q14, q12
+
+ vmov q15, q12
+ vst4.16 {d17,d19,d21,d23}, [\dst]! @ dst[4-7] = q8-q11[1]; dst += 4*2*4
+
+ vst4.16 {d24,d26,d28,d30}, [\dst]! @ dst[8-11] = q12-q15[0]; dst += 4*2*4
+
+ vst4.16 {d25,d27,d29,d31}, [\dst]! @ dst[12-15] = q12-q15[1]; dst += 4*2*4
+
+ vst4.16 {d16,d18,d20,d22}, [\reg1]! @ dst1[0-3] = q8-q11[0]; dst1 += 4*2*4
+
+ vst4.16 {d17,d19,d21,d23}, [\reg1]! @ dst1[4-7] = q8-q11[1]; dst1 += 4*2*4
+
+ vst4.16 {d24,d26,d28,d30}, [\reg1]! @ dst1[8-11] = q12-q15[0]; dst1 += 4*2*4
+
+ vst4.16 {d25,d27,d29,d31}, [\reg1]! @ dst1[12-15] = q12-q15[1]; dst1 += 4*2*4
+
+ vst4.16 {d16,d18,d20,d22}, [\reg2]! @ dst2[0-3] = q8-q11[0]; dst2 += 4*2*4
+
+ vst4.16 {d17,d19,d21,d23}, [\reg2]! @ dst2[4-7] = q8-q11[1]; dst2 += 4*2*4
+
+ vst4.16 {d24,d26,d28,d30}, [\reg2]! @ dst2[8-11] = q12-q15[0]; dst2 += 4*2*4
+
+ vst4.16 {d25,d27,d29,d31}, [\reg2]! @ dst2[12-15] = q12-q15[1]; dst2 += 4*2*4
+
+ vst4.16 {d16,d18,d20,d22}, [\reg3]! @ dst3[0-3] = q8-q11[0]; dst3 += 4*2*4
+
+ vst4.16 {d17,d19,d21,d23}, [\reg3]! @ dst3[4-7] = q8-q11[1]; dst3 += 4*2*4
+
+ vst4.16 {d24,d26,d28,d30}, [\reg3]! @ dst3[8-11] = q12-q15[0]; dst3 += 4*2*4
+
+ vst4.16 {d25,d27,d29,d31}, [\reg3]! @ dst3[12-15] = q12-q15[1]; dst3 += 4*2*4
+ bhs 10b
+
+ @ last 0-15 bytes
+
+ cmp \counter, #0
+ beq 40f
+
+ cmp \counter, #4
+ blo 30f
+
+ @ 4-12 pixels (4 pre iteration)
+ 20:
+ ldr \reg1, [\src] @ reg1 = src[0-3]
+ sub \counter, \counter, #4 @ counter -= 4
+
+ add \src, \src, #4 @ src += 4
+
+ ubfx \reg2, \reg1, #0, #8 @ reg2 = src[0]
+ cmp \counter, #4
+
+ ldr \reg2, [\pal, \reg2, lsl #2] @ reg2 = pal[src[0]]
+ ubfx \reg3, \reg1, #8, #8 @ reg3 = src[1]
+
+ ldr \reg3, [\pal, \reg3, lsl #2] @ reg3 = pal[src[1]]
+ ubfx \reg4, \reg1, #16, #8 @ reg4 = src[2]
+
+ ldr \reg4, [\pal, \reg4, lsl #2] @ reg4 = pal[src[2]]
+ lsr \reg1, \reg1, #24 @ reg1 = src[3]
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[3]]
+
+ add \reg5, \dst, \dststride
+ bfi \reg2, \reg3, #16, #16 @ reg2 = reg2 | reg3 << 16
+
+ vmov.32 d16[0], \reg2
+ add \reg6, \reg5, \dststride
+
+ bfi \reg4, \reg1, #16, #16 @ reg4 = reg4 | reg1 << 16
+ add \reg7, \reg6, \dststride
+
+ vmov.32 d16[1], \reg4
+
+ vmov d17, d16
+
+ vmov d18, d16
+
+ vmov d19, d16
+
+ vst4.16 {d16,d17,d18,d19}, [\dst]! @ dst[0-15] = d16-d19; dst += 4*2*4
+
+ vst4.16 {d16,d17,d18,d19}, [\reg5] @ dst1[0-15] = d16-d19
+
+ vst4.16 {d16,d17,d18,d19}, [\reg6] @ dst2[0-15] = d16-d19
+
+ vst4.16 {d16,d17,d18,d19}, [\reg7] @ dst3[0-15] = d16-d19
+ bhs 20b
+
+ cmp \counter, #0
+ beq 40f
+
+ @ last 1-3 pixels
+ 30:
+ ldrb \reg1, [\src], #1 @ reg1 = src[0]; src++
+ add \reg2, \dst, \dststride
+
+ add \reg3, \reg2, \dststride
+ add \dst, \dst, #8 @ dst += 4*2
+
+ ldr \reg1, [\pal, \reg1, lsl #2] @ reg1 = pal[src[0]]
+ add \reg4, \reg3, \dststride
+
+ strh \reg1, [\dst, #-8] @ dst[0] = reg1
+ subS \counter, \counter, #1 @ counter--
+
+ strh \reg1, [\dst, #-6] @ dst[1] = reg1
+
+ bfi \reg1, \reg1, #16, #16 @ reg1 = reg1 | reg1 << 16
+ str \reg1, [\dst, #-4] @ dst[2-3] = reg1
+
+ str \reg1, [\reg2] @ dst1[0-1] = reg1
+
+ str \reg1, [\reg2, #4] @ dst1[2-3] = reg1
+
+ str \reg1, [\reg3] @ dst2[0-1] = reg1
+
+ str \reg1, [\reg3, #4] @ dst2[2-3] = reg1
+
+ str \reg1, [\reg4] @ dst3[0-1] = reg1
+
+ str \reg1, [\reg4, #4] @ dst3[2-3] = reg1
+ bne 30b
+
+ 40:
+.endm
+
diff --git a/arm/neon_scale2x.S b/arm/neon_scale2x.S new file mode 100644 index 0000000..5c68cc6 --- /dev/null +++ b/arm/neon_scale2x.S @@ -0,0 +1,306 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+.arm
+
+#include "neon_scale2x.Sinc"
+#include "neon_normalxx.Sinc"
+
+.global neon_scale2x_8_8
+.global neon_scale2x_16_16
+.global neon_scale2x_8_16
+
+.align 4
+neon_scale2x_8_8:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r9}
+ ldr r9, [sp, #(7*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ add r5, r0, r3 @ r5 = src + srcstride
+ add r6, r1, ip @ r6 = dst + dststride
+ sub r3, r3, r2 @ r3 = srcstride - width
+ sub ip, ip, r2 @ ip = dststride - width
+ lsl ip, #1 @ ip = 2 * dststride - 2 * width
+ mov r7, r2 @ r7 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = counter
+@ r8 = tmpreg
+@ r9 = height
+@ ip = dstdiff (2 * dststride - 2 * width)
+
+ @ first line
+ neon_scale2x_8_8_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+
+ @ middle lines
+ 101:
+ mov r7, r2
+
+ neon_scale2x_8_8_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ bne 101b
+
+ @ last line
+ mov r7, r2
+
+ neon_scale2x_8_8_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ pop {r4-r9}
+ bx lr
+
+@ end procedure neon_scale2x_8_8
+
+
+neon_scale2x_16_16:
+
+@ r0 = const uint16_t *src
+@ r1 = uint16_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r9}
+ ldr r9, [sp, #(7*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ add r5, r0, r3 @ r5 = src + srcstride
+ add r6, r1, ip @ r6 = dst + dststride
+ sub r3, r3, r2, lsl #1 @ r3 = srcstride - 2 * width
+ sub ip, ip, r2, lsl #1 @ ip = dststride - 2 * width
+ lsl ip, #1 @ ip = 2 * dststride - 4 * width
+ mov r7, r2 @ r7 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - 2 * width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = counter
+@ r8 = tmpreg
+@ r9 = height
+@ ip = dstdiff (2 * dststride - 4 * width)
+
+ @ first line
+ neon_scale2x_16_16_line first, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+
+ @ middle lines
+ 101:
+ mov r7, r2
+
+ neon_scale2x_16_16_line middle, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ bne 101b
+
+ @ last line
+ mov r7, r2
+
+ neon_scale2x_16_16_line last, r4, r0, r5, r7, r1, r6, r8, 0, 0
+
+ pop {r4-r9}
+ bx lr
+
+@ end procedure neon_scale2x_16_16
+
+
+neon_scale2x_8_16:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = const uint32_t *palette
+@ r3 = unsigned int width (pixels)
+@ [sp] = unsigned int srcstride (bytes)
+@ [sp+4] = unsigned int dststride (bytes)
+@ [sp+8] = unsigned int height
+@ lr = return address
+
+@ three temporary lines
+
+ ldr ip, [sp] @ ip = srcstride
+ push {r4-r11,lr}
+ ldr r4, [sp, #(4*10)] @ r4 = dststride
+ ldr r5, [sp, #(4*11)] @ r5 = height
+ mov r6, sp @ r6 = sp
+ sub ip, ip, r3 @ ip = srcstride - width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub r7, r4, r3, lsl #1 @ r7 = dststride - 2 * width
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ sub r5, r5, #2 @ height -= 2
+ mov r10, sp @ tmpline3 = sp
+ lsl r7, #1 @ r7 = 2 * dststride - 4 * width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov r11, sp @ tmpline2 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov lr, sp @ tmpline1 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, #36
+ str r6, [sp] @ oldsp = r6
+ str r5, [sp, #4] @ height = r5
+ str ip, [sp, #8] @ srcdiff = ip
+ str r7, [sp, #12] @ dstdiff = r7
+ str r4, [sp, #16] @ dststride = r4
+ str lr, [sp, #20] @ tmpline1 = lr
+ str r11, [sp, #24] @ tmpline2 = r11
+ str r10, [sp, #28] @ tmpline3 = r10
+ str r3, [sp, #32] @ width = r3
+
+@ r0 = src
+@ r1 = dst
+@ r2 = palette
+@ r3 = counter
+@ r4 = dst2
+
+@ r11 = bufptr1
+@ ip = bufptr2
+@ lr = bufptr3
+
+@ [sp] = oldsp
+@ [sp, #4] = height
+@ [sp, #8] = srcdiff (srcstride - width)
+@ [sp, #12] = dstdiff (2 * dststride - 4 * width)
+@ [sp, #16] = dststride
+@ [sp, #20] = tmpline1
+@ [sp, #24] = tmpline2
+@ [sp, #28] = tmpline3
+@ [sp, #32] = width
+
+ @ lr = tmpline1
+ @ r3 = counter
+
+ @ first line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r0, r0, r7 @ src += srcdiff
+
+ @ second line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r3, [sp, #32] @ counter = width
+ ldr ip, [sp, #20] @ bufptr2 = tmpline1
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+
+ @ first temporary line
+ neon_scale2x_16_16_line first, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+
+ 100:
+
+ @ line n+1
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r4, r1, r9 @ dst2 = dst + dststride
+ ldr r3, [sp, #32] @ counter = width
+ str r11, [sp, #28] @ tmpline3 = bufptr1
+ str ip, [sp, #20] @ tmpline1 = bufptr2
+ str lr, [sp, #24] @ tmpline2 = bufptr3
+
+ @ temporary line n
+ neon_scale2x_16_16_line middle, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+ ldr r6, [sp, #4] @ r6 = height
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ subS r6, r6, #1 @ height--
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+ str r6, [sp, #4] @ height = r6
+ bne 100b
+
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+
+ @ last temporary line
+ neon_scale2x_16_16_line last, r11, ip, lr, r3, r1, r4, r5, 1, 0
+
+
+ ldr sp, [sp] @ sp = oldsp
+ pop {r4-r11,lr}
+ bx lr
+
+@ end procedure neon_scale2x_8_16
+
diff --git a/arm/neon_scale2x.Sinc b/arm/neon_scale2x.Sinc new file mode 100644 index 0000000..e9a80ff --- /dev/null +++ b/arm/neon_scale2x.Sinc @@ -0,0 +1,474 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+
+@ A B C --\ E0 E1
+@ D E F --/ E2 E3
+@ G H I
+
+@ q0 = E0 (tmp0)
+@ q1 = E1 (tmp1)
+@ q2 = E2 (tmp2)
+@ q3 = E3 (tmp3)
+@ q8 = S2prev
+@ q9 = S2next
+@ q10 = C0 < B == H || D == F >
+@ q11 = S1 < B >
+@ q12 = S2 < E >
+@ q13 = S3 < H >
+@ q14 = S2sl < D >
+@ q15 = S2sr < F >
+
+
+.macro __neon_scale2x_8_8_line src1, src2, src3, counter, dst1, dst2, reg1, qB, qH, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+
+ vld1.8 {d17[7]}, [\src2] @ S2prev[15] = src[0]
+ andS \reg1, \counter, #15 @ reg1 = counter & 15
+
+ .ifnes "\qB", "q11"
+ add \src1, \src1, \counter @ src1 += counter
+ .endif
+ .ifnes "\qH", "q13"
+ add \src3, \src3, \counter @ src3 += counter
+ .endif
+ beq 1f
+
+ @ first 1-15 pixels - align counter to 16 bytes
+ vld1.8 {q12}, [\src2], \reg1 @ S2 = [src] < E >; src2 += counter & 15
+
+ .ifeqs "\qB", "q11"
+ vld1.8 {\qB}, [\src1], \reg1 @ S1 = [src - srcstride] < B >; src1 += counter & 15
+ .endif
+
+ .ifeqs "\qH", "q13"
+ vld1.8 {\qH}, [\src3], \reg1 @ S3 = [src + srcstride] < H >; src3 += counter & 15
+ .endif
+ vext.8 q14, q8, q12, #15 @ S2sl = S2prev[15] | (S2 << 8) < D >
+
+ vceq.i8 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vmov.8 d17[7], \reg1 @ S2prev[15] = reg1
+ vext.8 q15, q12, q9, #1 @ S2sr = (S2 >> 8) | ... < F >
+
+ vceq.i8 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i8 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i8 q1, \qB, q15 @ tmp1 = < B == F >
+ vtbl.8 d17, {d28, d29}, d17 @ S2prev[15] = src[reg1 - 1]
+
+ lsl \reg1, #1
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+
+ vceq.i8 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i8 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.8 {q0-q1}, [\dst1], \reg1 @ [dst] = E0,E1; dst1 += reg1
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+ bic \counter, \counter, #15
+
+ vst2.8 {q2-q3}, [\dst2], \reg1 @ [dst + dststride] = E2,E3; dst2 += reg1
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ vld1.8 {q9}, [\alsrc2]! @ S2next = [src]; src2 += 16
+
+ @ inner loop (16 pixels per iteration)
+ 2:
+
+ vmov q12, q9 @ S2 = S2next < E >
+ .ifeqs "\qB", "q11"
+ vld1.8 {\qB}, [\alsrc1]! @ S1 = [src - srcstride] < B >; src1 += 16
+ .endif
+
+ .ifeqs "\qH", "q13"
+ vld1.8 {\qH}, [\alsrc3]! @ S3 = [src + srcstride] < H >; src3 += 16
+ .endif
+
+ vext.8 q14, q8, q12, #15 @ S2sl = S2prev[15] | (S2 << 8) < D >
+ vld1.8 {q9}, [\alsrc2]! @ S2next = [src]; src2 += 16
+
+ vceq.i8 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vmov q8, q12 @ S2prev = S2
+ vext.8 q15, q12, q9, #1 @ S2sr = (S2 >> 8) | S2next[0] < F >
+
+ vceq.i8 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i8 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i8 q1, \qB, q15 @ tmp1 = < B == F >
+
+ sub \counter, \counter, #16 @ counter -= 16
+
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+
+ vceq.i8 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i8 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.8 {q0-q1}, [\aldst1]! @ [dst] = E0,E1; dst1 += 2*16
+
+ cmp \counter, #16
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+
+ vst2.8 {q2-q3}, [\aldst2]! @ [dst + dststride] = E2,E3; dst2 += 2*16
+
+ bhi 2b
+
+ @ last 16 pixels
+
+ vmov q12, q9 @ S2 = S2next < E >
+
+ vshr.u64 d18, d19, #(64-8) @ S2next[0] = S2[15] | ...
+ .ifeqs "\qB", "q11"
+ vld1.8 {\qB}, [\alsrc1]! @ S1 = [src - srcstride] < B >; src1 += 16
+ .endif
+
+ vext.8 q14, q8, q12, #15 @ S2sl = S2prev[15] | (S2 << 8) < D >
+
+ vext.8 q15, q12, q9, #1 @ S2sr = (S2 >> 8) | S2next[0] < F >
+ .ifeqs "\qH", "q13"
+ vld1.8 {\qH}, [\alsrc3]! @ S3 = [src + srcstride] < H >; src3 += 16
+ .endif
+
+ vceq.i8 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i8 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vceq.i8 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i8 q1, \qB, q15 @ tmp1 = < B == F >
+
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+
+ vceq.i8 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i8 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.8 {q0-q1}, [\aldst1]! @ [dst] = E0,E1; dst1 += 2*16
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+
+ vst2.8 {q2-q3}, [\aldst2]! @ [dst + dststride] = E2,E3; dst2 += 2*16
+
+.endm
+
+.macro _neon_scale2x_8_8_line_first src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q12, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_scale2x_8_8_line_middle src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_scale2x_8_8_line_last src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q12, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro neon_scale2x_8_8_line part, src1, src2, src3, counter, dst1, dst2, reg1, srcalign16, dstalign32
+ .ifeq \srcalign16
+
+ .ifeq \dstalign32
+ _neon_scale2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1, \dst2
+ .else
+ _neon_scale2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1:256, \dst2:256
+ .endif
+
+ .else
+
+ .ifeq \dstalign32
+ _neon_scale2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2
+ .else
+ _neon_scale2x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1:256, \dst2:256
+ .endif
+
+ .endif
+.endm
+
+
+.macro __neon_scale2x_16_16_line src1, src2, src3, counter, dst1, dst2, reg1, qB, qH, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+
+ vld1.16 {d17[3]}, [\src2] @ S2prev[7] = src[0]
+ andS \reg1, \counter, #7 @ reg1 = counter & 7
+
+ .ifnes "\qB", "q11"
+ add \src1, \src1, \counter, lsl #1 @ src1 += 2 * counter
+ .endif
+ .ifnes "\qH", "q13"
+ add \src3, \src3, \counter, lsl #1 @ src3 += 2 * counter
+ .endif
+ beq 1f
+
+ @ first 1-7 pixels - align counter to 16 bytes
+ vld1.16 {q12}, [\src2] @ S2 = [src] < E >
+ lsl \reg1, #1
+
+ .ifeqs "\qB", "q11"
+ vld1.16 {\qB}, [\src1] @ S1 = [src - srcstride] < B >
+ .endif
+ bfi \reg1, \reg1, #8, #8
+
+ .ifeqs "\qH", "q13"
+ vld1.16 {\qH}, [\src3] @ S3 = [src + srcstride] < H >
+ .endif
+ vext.8 q14, q8, q12, #14 @ S2sl = S2prev[7] | (S2 << 16) < D >
+
+ add \reg1, \reg1, #256
+ vceq.i16 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vmov.16 d17[3], \reg1 @ S2prev[7] = reg1
+ vext.8 q15, q12, q9, #2 @ S2sr = (S2 >> 16) | ... < F >
+
+ vceq.i16 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i16 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i16 q1, \qB, q15 @ tmp1 = < B == F >
+ vtbl.8 d17, {d28, d29}, d17 @ S2prev[7] = src[reg1 - 1]
+
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+ and \reg1, \counter, #7
+
+ vceq.i16 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i16 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.16 {q0-q1}, [\dst1] @ [dst] = E0,E1
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+
+ bic \counter, \counter, #7
+ .ifeqs "\qB", "q11"
+ add \src1, \src1, \reg1, lsl #1
+ .endif
+ add \src2, \src2, \reg1, lsl #1
+ .ifeqs "\qH", "q13"
+ add \src3, \src3, \reg1, lsl #1
+ .endif
+
+ vst2.16 {q2-q3}, [\dst2] @ [dst + dststride] = E2,E3
+
+ add \dst1, \dst1, \reg1, lsl #2
+ add \dst2, \dst2, \reg1, lsl #2
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ vld1.16 {q9}, [\alsrc2]! @ S2next = [src]; src2 += 2*8
+
+ @ inner loop (8 pixels per iteration)
+ 2:
+
+ vmov q12, q9 @ S2 = S2next < E >
+ .ifeqs "\qB", "q11"
+ vld1.16 {\qB}, [\alsrc1]! @ S1 = [src - srcstride] < B >; src1 += 2*8
+ .endif
+
+ .ifeqs "\qH", "q13"
+ vld1.16 {\qH}, [\alsrc3]! @ S3 = [src + srcstride] < H >; src3 += 2*8
+ .endif
+
+ vext.8 q14, q8, q12, #14 @ S2sl = S2prev[7] | (S2 << 16) < D >
+ vld1.16 {q9}, [\alsrc2]! @ S2next = [src]; src2 += 2*8
+
+ vceq.i16 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vmov q8, q12 @ S2prev = S2
+ vext.8 q15, q12, q9, #2 @ S2sr = (S2 >> 16) | S2next[0] < F >
+
+ vceq.i16 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i16 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i16 q1, \qB, q15 @ tmp1 = < B == F >
+
+ sub \counter, \counter, #8 @ counter -= 8
+
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+
+ vceq.i16 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i16 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.16 {q0-q1}, [\aldst1]! @ [dst] = E0,E1; dst1 += 2*2*8
+
+ cmp \counter, #8
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+
+ vst2.16 {q2-q3}, [\aldst2]! @ [dst + dststride] = E2,E3; dst2 += 2*2*8
+
+ bhi 2b
+
+ @ last 8 pixels
+
+ vmov q12, q9 @ S2 = S2next < E >
+
+ vshr.u64 d18, d19, #(64-16) @ S2next[0] = S2[7] | ...
+ .ifeqs "\qB", "q11"
+ vld1.16 {\qB}, [\alsrc1]! @ S1 = [src - srcstride] < B >; src1 += 2*8
+ .endif
+
+ vext.8 q14, q8, q12, #14 @ S2sl = S2prev[7] | (S2 << 16) < D >
+
+ vext.8 q15, q12, q9, #2 @ S2sr = (S2 >> 16) | S2next[0] < F >
+ .ifeqs "\qH", "q13"
+ vld1.16 {\qH}, [\alsrc3]! @ S3 = [src + srcstride] < H >; src3 += 2*8
+ .endif
+
+ vceq.i16 q0, q14, \qB @ tmp0 = < D == B >
+
+ vceq.i16 q2, \qB, \qH @ tmp2 = < B == H >
+
+ vceq.i16 q3, q14, q15 @ tmp3 = < D == F >
+
+ vceq.i16 q1, \qB, q15 @ tmp1 = < B == F >
+
+ vorr q10, q2, q3 @ C0 = < B == H || D == F >
+
+ vceq.i16 q2, q14, \qH @ tmp2 = < D == H >
+
+ vceq.i16 q3, \qH, q15 @ tmp3 = < H == F >
+
+ vorn q0, q10, q0 @ tmp0 = < C0 || !(D == B) >
+
+ vorn q1, q10, q1 @ tmp1 = < C0 || !(B == F) >
+
+ vbsl q0, q12, q14 @ E0 = < (C0 || !(D == B)) ? E : D >
+
+ vbsl q1, q12, q15 @ E1 = < (C0 || !(B == F)) ? E : F >
+
+ vorn q2, q10, q2 @ tmp2 = < C0 || !(D == H) >
+
+ vorn q3, q10, q3 @ tmp3 = < C0 || !(H == F) >
+
+ vbsl q2, q12, q14 @ E2 = < (C0 || !(D == H)) ? E : D >
+ vst2.16 {q0-q1}, [\aldst1]! @ [dst] = E0,E1; dst1 += 2*2*8
+
+ vbsl q3, q12, q15 @ E3 = < (C0 || !(H == F)) ? E : F >
+
+ vst2.16 {q2-q3}, [\aldst2]! @ [dst + dststride] = E2,E3; dst2 += 2*2*8
+
+.endm
+
+.macro _neon_scale2x_16_16_line_first src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q12, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_scale2x_16_16_line_middle src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q13, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro _neon_scale2x_16_16_line_last src1, src2, src3, counter, dst1, dst2, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2
+ __neon_scale2x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, q11, q12, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2
+.endm
+
+.macro neon_scale2x_16_16_line part, src1, src2, src3, counter, dst1, dst2, reg1, srcalign16, dstalign32
+ .ifeq \srcalign16
+
+ .ifeq \dstalign32
+ _neon_scale2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1, \dst2
+ .else
+ _neon_scale2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1, \src2, \src3, \dst1:256, \dst2:256
+ .endif
+
+ .else
+
+ .ifeq \dstalign32
+ _neon_scale2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2
+ .else
+ _neon_scale2x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \reg1, \src1:128, \src2:128, \src3:128, \dst1:256, \dst2:256
+ .endif
+
+ .endif
+.endm
+
diff --git a/arm/neon_scale2x.h b/arm/neon_scale2x.h new file mode 100644 index 0000000..4db07c5 --- /dev/null +++ b/arm/neon_scale2x.h @@ -0,0 +1,44 @@ +/** + * + * Copyright (C) 2012 Roman Pauer + * + * Permission is hereby granted, free of charge, to any person obtaining a copy of + * this software and associated documentation files (the "Software"), to deal in + * the Software without restriction, including without limitation the rights to + * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies + * of the Software, and to permit persons to whom the Software is furnished to do + * so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#if !defined(_NEON_SCALE2X_H_INCLUDED_) +#define _NEON_SCALE2X_H_INCLUDED_ + +#include <inttypes.h> + +#ifdef __cplusplus +extern "C" { +#endif + +extern void neon_scale2x_8_8(const uint8_t *src, uint8_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); +extern void neon_scale2x_16_16(const uint16_t *src, uint16_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +extern void neon_scale2x_8_16(const uint8_t *src, uint16_t *dst, const uint32_t *palette, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +#ifdef __cplusplus +} +#endif + + +#endif /* _NEON_SCALE2X_H_INCLUDED_ */ diff --git a/arm/neon_scale3x.S b/arm/neon_scale3x.S new file mode 100644 index 0000000..3789488 --- /dev/null +++ b/arm/neon_scale3x.S @@ -0,0 +1,349 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+.arm
+
+#include "neon_scale3x.Sinc"
+#include "neon_normalxx.Sinc"
+
+.global neon_scale3x_8_8
+.global neon_scale3x_16_16
+.global neon_scale3x_8_16
+
+.align 4
+neon_scale3x_8_8:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r11}
+ ldr r9, [sp, #(9*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ mov r11, sp @ oldsp = sp
+ add r5, r0, r3 @ r5 = src + srcstride
+ bic sp, sp, #31 @ align sp to 32 bytes
+ add r6, r1, ip @ r6 = dst + dststride
+ sub sp, sp, #64 @ sp -= 64
+ sub r3, r3, r2 @ r3 = srcstride - width
+ vst1.64 {d8-d11}, [sp:256] @ save q4,q5
+ add r7, r1, ip, lsl #1 @ r7 = dst + 2 * dststride
+ add r8, sp, #32 @ r8 = sp + 32
+ sub ip, ip, r2 @ ip = dststride - width
+ vst1.64 {d12-d15}, [r8:256] @ save q6,q7
+ add ip, ip, ip, lsl #1 @ ip = 3 * dststride - 3 * width
+ mov r8, r2 @ r8 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = dst + 2 * dststride
+@ r8 = counter
+@ r9 = height
+@ r10 = tmpreg
+@ r11 = oldsp
+@ ip = dstdiff (3 * dststride - 3 * width)
+
+ @ first line
+ neon_scale3x_8_8_line first, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ add r7, r7, ip
+
+ @ middle lines
+ 101:
+ mov r8, r2
+
+ neon_scale3x_8_8_line middle, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ add r7, r7, ip
+ bne 101b
+
+ @ last line
+ mov r8, r2
+
+ neon_scale3x_8_8_line last, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ add ip, sp, #32 @ ip = sp + 32
+ vld1.64 {d8-d11}, [sp:256] @ restore q4,q5
+ mov sp, r11 @ sp = oldsp
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r11}
+ bx lr
+
+@ end procedure neon_scale3x_8_8
+
+
+neon_scale3x_16_16:
+
+@ r0 = const uint16_t *src
+@ r1 = uint16_t *dst
+@ r2 = unsigned int width (pixels)
+@ r3 = unsigned int srcstride (bytes)
+@ [sp] = unsigned int dststride (bytes)
+@ [sp+4] = unsigned int height
+@ lr = return address
+
+ ldr ip, [sp] @ ip = dststride
+ push {r4-r11}
+ ldr r9, [sp, #(9*4)] @ r9 = height
+ sub r4, r0, r3 @ r4 = src - srcstride
+ mov r11, sp @ oldsp = sp
+ add r5, r0, r3 @ r5 = src + srcstride
+ bic sp, sp, #31 @ align sp to 32 bytes
+ add r6, r1, ip @ r6 = dst + dststride
+ sub sp, sp, #64 @ sp -= 64
+ sub r3, r3, r2, lsl #1 @ r3 = srcstride - 2 * width
+ vst1.64 {d8-d11}, [sp:256] @ save q4,q5
+ add r7, r1, ip, lsl #1 @ r7 = dst + 2 * dststride
+ add r8, sp, #32 @ r8 = sp + 32
+ sub ip, ip, r2, lsl #1 @ ip = dststride - 2 * width
+ vst1.64 {d12-d15}, [r8:256] @ save q6,q7
+ add ip, ip, ip, lsl #1 @ ip = 3 * dststride - 6 * width
+ mov r8, r2 @ r8 = width
+ sub r9, r9, #2 @ r9 = height - 2
+
+@ r0 = src
+@ r1 = dst
+@ r2 = width
+@ r3 = srcdiff (srcstride - 2 * width)
+@ r4 = src - srcstride
+@ r5 = src + srcstride
+@ r6 = dst + dststride
+@ r7 = dst + 2 * dststride
+@ r8 = counter
+@ r9 = height
+@ r10 = tmpreg
+@ r11 = oldsp
+@ ip = dstdiff (3 * dststride - 6 * width)
+
+ @ first line
+ neon_scale3x_16_16_line first, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ add r7, r7, ip
+
+ @ middle lines
+ 101:
+ mov r8, r2
+
+ neon_scale3x_16_16_line middle, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ subS r9, r9, #1
+ add r0, r0, r3
+ add r4, r4, r3
+ add r5, r5, r3
+ add r1, r1, ip
+ add r6, r6, ip
+ add r7, r7, ip
+ bne 101b
+
+ @ last line
+ mov r8, r2
+
+ neon_scale3x_16_16_line last, r4, r0, r5, r8, r1, r6, r7, r10, 0, 0
+
+ add ip, sp, #32 @ ip = sp + 32
+ vld1.64 {d8-d11}, [sp:256] @ restore q4,q5
+ mov sp, r11 @ sp = oldsp
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r11}
+ bx lr
+
+@ end procedure neon_scale3x_16_16
+
+
+neon_scale3x_8_16:
+
+@ r0 = const uint8_t *src
+@ r1 = uint8_t *dst
+@ r2 = const uint32_t *palette
+@ r3 = unsigned int width (pixels)
+@ [sp] = unsigned int srcstride (bytes)
+@ [sp+4] = unsigned int dststride (bytes)
+@ [sp+8] = unsigned int height
+@ lr = return address
+
+@ three temporary lines
+
+ ldr ip, [sp] @ ip = srcstride
+ push {r4-r11,lr}
+ ldr r4, [sp, #(4*10)] @ r4 = dststride
+ ldr r5, [sp, #(4*11)] @ r5 = height
+ mov r6, sp @ r6 = sp
+ sub ip, ip, r3 @ ip = srcstride - width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub r7, r4, r3, lsl #1 @ r7 = dststride - 2 * width
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ sub r5, r5, #2 @ height -= 2
+ mov r10, sp @ tmpline3 = sp
+ add r7, r7, r7, lsl #1 @ r7 = 3 * dststride - 6 * width
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov r11, sp @ tmpline2 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub sp, sp, r3, lsl #1 @ sp -= 2 * width
+ mov lr, sp @ tmpline1 = sp
+ bic sp, sp, #31 @ align sp to 32 bytes
+ sub r8, sp, #64 @ r8 = sp - 64
+ vst1.64 {d8-d11}, [r8:256] @ save q4,q5
+ sub r9, sp, #32 @ r9 = sp - 32
+ vst1.64 {d12-d15}, [r9:256] @ save q6,q7
+ sub sp, sp, #(36 + 64) @ sp -= (36 + 64)
+ str r6, [sp] @ oldsp = r6
+ str r5, [sp, #4] @ height = r5
+ str ip, [sp, #8] @ srcdiff = ip
+ str r7, [sp, #12] @ dstdiff = r7
+ str r4, [sp, #16] @ dststride = r4
+ str lr, [sp, #20] @ tmpline1 = lr
+ str r11, [sp, #24] @ tmpline2 = r11
+ str r10, [sp, #28] @ tmpline3 = r10
+ str r3, [sp, #32] @ width = r3
+
+@ r0 = src
+@ r1 = dst
+@ r2 = palette
+@ r3 = counter
+@ r4 = dst2
+@ r5 = dst3
+
+@ r11 = bufptr1
+@ ip = bufptr2
+@ lr = bufptr3
+
+@ [sp] = oldsp
+@ [sp, #4] = height
+@ [sp, #8] = srcdiff (srcstride - width)
+@ [sp, #12] = dstdiff (3 * dststride - 6 * width)
+@ [sp, #16] = dststride
+@ [sp, #20] = tmpline1
+@ [sp, #24] = tmpline2
+@ [sp, #28] = tmpline3
+@ [sp, #32] = width
+
+ @ lr = tmpline1
+ @ r3 = counter
+
+ @ first line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r0, r0, r7 @ src += srcdiff
+
+ @ second line
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r3, [sp, #32] @ counter = width
+ ldr ip, [sp, #20] @ bufptr2 = tmpline1
+ ldr lr, [sp, #24] @ bufptr3 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+ add r5, r1, r9, lsl #1 @ dst3 = dst + 2 * dststride
+
+ @ first temporary line
+ neon_scale3x_16_16_line first, r11, ip, lr, r3, r1, r4, r5, r6, 1, 0
+
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+
+ 100:
+
+ @ line n+1
+ neon_normal1x_8_16_line r0, lr, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, ip
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r4, r1, r9 @ dst2 = dst + dststride
+ ldr r3, [sp, #32] @ counter = width
+ add r5, r1, r9, lsl #1 @ dst3 = dst + 2 * dststride
+ str r11, [sp, #28] @ tmpline3 = bufptr1
+ str ip, [sp, #20] @ tmpline1 = bufptr2
+ str lr, [sp, #24] @ tmpline2 = bufptr3
+
+ @ temporary line n
+ neon_scale3x_16_16_line middle, r11, ip, lr, r3, r1, r4, r5, r6, 1, 0
+
+ ldr r6, [sp, #4] @ r6 = height
+ ldr r7, [sp, #8] @ r7 = srcdiff
+ ldr r8, [sp, #12] @ r8 = dstdiff
+ ldr r3, [sp, #32] @ counter = width
+ subS r6, r6, #1 @ height--
+ ldr lr, [sp, #28] @ bufptr3 = tmpline3
+ add r0, r0, r7 @ src += srcdiff
+ add r1, r1, r8 @ dst += dstdiff
+ str r6, [sp, #4] @ height = r6
+ bne 100b
+
+
+ ldr r9, [sp, #16] @ r9 = dststride
+ ldr r11, [sp, #20] @ bufptr1 = tmpline1
+ ldr ip, [sp, #24] @ bufptr2 = tmpline2
+ add r4, r1, r9 @ dst2 = dst + dststride
+ add r5, r1, r9, lsl #1 @ dst3 = dst + 2 * dststride
+
+ @ last temporary line
+ neon_scale3x_16_16_line last, r11, ip, lr, r3, r1, r4, r5, r6, 1, 0
+
+
+ add r6, sp, #36 @ r6 = sp + 36
+ ldr sp, [sp] @ sp = oldsp
+ vld1.64 {d8-d11}, [r6:256] @ restore q4,q5
+ add ip, r6, #32 @ ip = r6 + 32
+ vld1.64 {d12-d15}, [ip:256] @ restore q6,q7
+ pop {r4-r11,lr}
+ bx lr
+
+@ end procedure neon_scale3x_8_16
+
diff --git a/arm/neon_scale3x.Sinc b/arm/neon_scale3x.Sinc new file mode 100644 index 0000000..fc72d31 --- /dev/null +++ b/arm/neon_scale3x.Sinc @@ -0,0 +1,976 @@ +@@
+@@ Copyright (C) 2012 Roman Pauer
+@@
+@@ Permission is hereby granted, free of charge, to any person obtaining a copy of
+@@ this software and associated documentation files (the "Software"), to deal in
+@@ the Software without restriction, including without limitation the rights to
+@@ use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+@@ of the Software, and to permit persons to whom the Software is furnished to do
+@@ so, subject to the following conditions:
+@@
+@@ The above copyright notice and this permission notice shall be included in all
+@@ copies or substantial portions of the Software.
+@@
+@@ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+@@ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+@@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+@@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+@@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+@@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+@@ SOFTWARE.
+@@
+
+
+@ A B C --\ E0 E1 E2
+@ D E F --/ E3 E4 E5
+@ G H I E6 E7 E8
+
+@ q0 = S1sl < A >
+@ q1 = S2sl < D >
+@ q2 = S3sl < G >
+@ q3 = S1sr < C >
+@ q4 = S2sr < F >
+@ q5 = S3sr < I >
+@ q6 =
+@ q7 =
+@ q8 = S1
+@ q9 = S2
+@ q10 = S3
+@ q11 =
+@ q12 = S2prev < E >
+@ q13 =
+@ q14 = S1prev < B >
+@ q15 = S3prev < H >
+
+
+.macro ___neon_scale3x_8_8_slice counter, dst1, dst2, dst3, reg1, dB0, dB1, dH0, dH1, first15, last16
+
+@ d12 = E0[0]
+@ d13 = E1[0]
+@ d14 = E2[0]
+@ d15 = tmp0[0]
+@ d22 = E3[0]
+@ d23 = tmp1[0]
+@ d24 = E4[0]
+@ d25 = E4[1]
+@ d26 = E5[0]
+@ d27 = C0[0]
+
+@ q0 = tmp2
+@ q3 = tmp3
+@ q2 = tmp4
+@ q5 = tmp5
+
+ vceq.i8 d15, \dB0, \dH0 @ tmp0[0] = < B == H >
+
+ vceq.i8 d23, d2, d8 @ tmp1[0] = < D == F >
+
+ .if \first15
+ cmp \reg1, #8
+ .endif
+
+ vceq.i8 q0, q12, q0 @ tmp2 = < E == A >
+
+ vceq.i8 q3, q12, q3 @ tmp3 = < E == C >
+
+ .if \first15
+ add \reg1, \reg1, \reg1, lsl #1 @ reg1 = 3 * (counter & 15)
+ .endif
+
+ vorr d27, d15, d23 @ C0[0] = < B == H || D == F >
+
+ vceq.i8 d22, d2, \dB0 @ E3[0] = < D == B >
+
+ vceq.i8 d26, \dB0, d8 @ E5[0] = < B == F >
+
+ vceq.i8 q2, q12, q2 @ tmp4 = < E == G >
+
+ vceq.i8 q5, q12, q5 @ tmp5 = < E == I >
+
+ .if \first15
+ sub \reg1, \reg1, #(3*8) @ reg1 -= 3*8
+ .endif
+
+ vorn d15, d6, d22 @ tmp0[0] = < (E == C) || !(D == B) >
+
+ vorn d23, d0, d26 @ tmp1[0] = < (E == A) || !(B == F) >
+
+ vorn d12, d27, d22 @ E0[0] = < C0 || !(D == B) >
+
+ vand d13, d15, d23 @ E1[0] = < ((E == C) || !(D == B)) && ((E == A) || !(B == F)) >
+
+ vorn d14, d27, d26 @ E2[0] = < C0 || !(B == F) >
+
+ vorr d13, d27, d13 @ E1[0] = < C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F))) >
+
+ vbsl d12, d24, d2 @ E0[0] = < (C0 || !(D == B)) ? E : D >
+
+ vbsl d14, d24, d8 @ E2[0] = < (C0 || !(B == F)) ? E : F >
+
+ vbsl d13, d24, \dB0 @ E1[0] = < (C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F)))) ? E : B >
+
+ vceq.i8 d15, d2, \dH0 @ tmp0[0] = < D == H >
+
+ vceq.i8 d23, \dH0, d8 @ tmp1[0] = < H == F >
+
+ vorn d22, d4, d22 @ E3[0] = < (E == G) || !(D == B) >
+ vst3.8 {d12-d14}, [\dst1]! @ [dst] = E0,E1,E2; dst1 += 3*8
+
+ .if \first15
+ addls \dst1, \dst1, \reg1 @ dst1 += reg1
+ .endif
+
+ vorn d26, d10, d26 @ E5[0] = < (E == I) || !(B == F) >
+
+@ d12 = tmp6[0]
+@ d13 = tmp7[0]
+
+ vorn d12, d0, d15 @ tmp6[0] = < (E == A) || !(D == H) >
+
+ vorn d13, d6, d23 @ tmp7[0] = < (E == C) || !(H == F) >
+
+ vand d22, d22, d12 @ E3[0] = < ((E == G) || !(D == B)) && ((E == A) || !(D == H)) >
+
+ vand d26, d26, d13 @ E5[0] = < ((E == I) || !(B == F)) && ((E == C) || !(H == F)) >
+
+@ d12 = E6[0]
+@ d13 = E7[0]
+@ d14 = E8[0]
+
+ vorr d22, d27, d22 @ E3[0] = < C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H))) >
+
+ vorr d26, d27, d26 @ E5[0] = < C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F))) >
+
+ vbsl d22, d24, d2 @ E3[0] = < (C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H)))) ? E : D >
+
+ vbsl d26, d24, d8 @ E5[0] = < (C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F)))) ? E : F >
+
+ vorn d13, d10, d15 @ E7[0] = < (E == I) || !(D == H) >
+
+ vorn d12, d27, d15 @ E6[0] = < C0 || !(D == H) >
+ vst3.8 {d22,d24,d26}, [\dst2]! @ [dst + dststride] = E3,E4,E5; dst2 += 3*8
+
+ .if \first15
+ addls \dst2, \dst2, \reg1 @ dst2 += reg1
+ .endif
+
+ vorn d15, d4, d23 @ tmp0[0] = < (E == G) || !(H == F) >
+
+ vorn d14, d27, d23 @ E8[0] = < C0 || !(H == F) >
+
+ vand d13, d13, d15 @ E7[0] = < ((E == I) || !(D == H)) && ((E == G) || !(H == F)) >
+
+ vbsl d12, d24, d2 @ E6[0] = < (C0 || !(D == H)) ? E : D >
+
+ vorr d13, d27, d13 @ E7[0] = < C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F))) >
+
+ vbsl d14, d24, d8 @ E8[0] = < (C0 || !(H == F)) ? E : F >
+
+ vbsl d13, d24, \dH0 @ E7[0] = < (C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F)))) ? E : H >
+
+@ d15 = tmp0[1]
+@ d22 = tmp1[1]
+@ d23 = E3[1]
+@ d24 = E4[0]
+@ d25 = E4[1]
+@ d26 = C0[1]
+@ d27 = E5[1]
+
+ vceq.i8 d15, \dB1, \dH1 @ tmp0[1] = < B == H >
+
+ vceq.i8 d22, d3, d9 @ tmp1[1] = < D == F >
+
+ vceq.i8 d23, d3, \dB1 @ E3[1] = < D == B >
+ vst3.8 {d12-d14}, [\dst3]! @ [dst + 2 * dststride] = E6,E7,E8; dst3 += 3*8
+
+ .if \first15
+ addls \dst3, \dst3, \reg1 @ dst3 += reg1
+ bls 0f
+ .endif
+
+@ d12 = E0[1]
+@ d13 = E1[1]
+@ d14 = E2[1]
+
+ vorr d26, d15, d22 @ C0[1] = < B == H || D == F >
+
+ vceq.i8 d27, \dB1, d9 @ E5[1] = < B == F >
+
+ vorn d15, d7, d23 @ tmp0[1] = < (E == C) || !(D == B) >
+
+ vorn d22, d1, d27 @ tmp1[1] = < (E == A) || !(B == F) >
+
+ vorn d12, d26, d23 @ E0[1] = < C0 || !(D == B) >
+
+ vand d13, d15, d22 @ E1[1] = < ((E == C) || !(D == B)) && ((E == A) || !(B == F)) >
+
+ vorn d14, d26, d27 @ E2[1] = < C0 || !(B == F) >
+
+ vorr d13, d26, d13 @ E1[1] = < C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F))) >
+
+ vbsl d12, d25, d3 @ E0[1] = < (C0 || !(D == B)) ? E : D >
+
+ vbsl d14, d25, d9 @ E2[1] = < (C0 || !(B == F)) ? E : F >
+
+ vbsl d13, d25, \dB1 @ E1[1] = < (C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F)))) ? E : B >
+
+ vceq.i8 d15, d3, \dH1 @ tmp0[1] = < D == H >
+
+ vceq.i8 d22, \dH1, d9 @ tmp1[1] = < H == F >
+
+ vorn d23, d5, d23 @ E3[1] = < (E == G) || !(D == B) >
+ .ifeq \first15
+ vst3.8 {d12-d14}, [\dst1]! @ [dst] = E0,E1,E2; dst1 += 3*8
+ .else
+ vst3.8 {d12-d14}, [\dst1],\reg1 @ [dst] = E0,E1,E2; dst1 += reg1
+ .endif
+
+ vorn d27, d11, d27 @ E5[1] = < (E == I) || !(B == F) >
+
+@ d12 = tmp6[1]
+@ d13 = tmp7[1]
+
+ vorn d12, d1, d15 @ tmp6[1] = < (E == A) || !(D == H) >
+
+ vorn d13, d7, d22 @ tmp7[1] = < (E == C) || !(H == F) >
+
+ vand d23, d23, d12 @ E3[1] = < ((E == G) || !(D == B)) && ((E == A) || !(D == H)) >
+
+ vand d27, d27, d13 @ E5[1] = < ((E == I) || !(B == F)) && ((E == C) || !(H == F)) >
+
+@ d12 = E6[1]
+@ d13 = E7[1]
+@ d14 = E8[1]
+
+ vorr d23, d26, d23 @ E3[1] = < C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H))) >
+
+ vorr d27, d26, d27 @ E5[1] = < C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F))) >
+
+ vbsl d23, d25, d3 @ E3[1] = < (C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H)))) ? E : D >
+
+ vbsl d27, d25, d9 @ E5[1] = < (C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F)))) ? E : F >
+
+ vorn d13, d11, d15 @ E7[1] = < (E == I) || !(D == H) >
+
+ vorn d12, d26, d15 @ E6[1] = < C0 || !(D == H) >
+ .ifeq \first15
+ vst3.8 {d23,d25,d27}, [\dst2]! @ [dst + dststride] = E3,E4,E5; dst2 += 3*8
+ .else
+ vst3.8 {d23,d25,d27}, [\dst2],\reg1 @ [dst + dststride] = E3,E4,E5; dst2 += reg1
+ .endif
+
+ vorn d15, d5, d22 @ tmp0[1] = < (E == G) || !(H == F) >
+
+ vorn d14, d26, d22 @ E8[1] = < C0 || !(H == F) >
+
+ vand d13, d13, d15 @ E7[1] = < ((E == I) || !(D == H)) && ((E == G) || !(H == F)) >
+
+ vbsl d12, d25, d3 @ E6[1] = < (C0 || !(D == H)) ? E : D >
+
+ vorr d13, d26, d13 @ E7[1] = < C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F))) >
+
+ vbsl d14, d25, d9 @ E8[1] = < (C0 || !(H == F)) ? E : F >
+
+ vbsl d13, d25, \dH1 @ E7[1] = < (C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F)))) ? E : H >
+
+ .ifeq \first15
+
+ .ifeq \last16
+ sub \counter, \counter, #16 @ counter -= 16
+ cmp \counter, #16
+ .endif
+
+ vst3.8 {d12-d14}, [\dst3]! @ [dst + 2 * dststride] = E6,E7,E8; dst3 += 3*8
+ .else
+ vst3.8 {d12-d14}, [\dst3],\reg1 @ [dst + 2 * dststride] = E6,E7,E8; dst3 += reg1
+
+ 0:
+ .endif
+
+.endm
+
+.macro __neon_scale3x_8_8_line src1, src2, src3, counter, dst1, dst2, dst3, reg1, qB, qH, dB0, dB1, dH0, dH1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+
+ .ifeqs "\qB", "q14"
+ vld1.8 {d29[7]}, [\src1] @ S1prev[15] = src[-srcstride]
+ .endif
+ vld1.8 {d25[7]}, [\src2] @ S2prev[15] = src[0]
+ .ifeqs "\qH", "q15"
+ vld1.8 {d31[7]}, [\src3] @ S3prev[15] = src[srcstride]
+ .endif
+ andS \reg1, \counter, #15 @ reg1 = counter & 15
+
+ .ifnes "\qB", "q14"
+ add \src1, \src1, \counter @ src1 += counter
+ .endif
+ .ifnes "\qH", "q15"
+ add \src3, \src3, \counter @ src3 += counter
+ .endif
+ beq 1f
+
+ @ first 1-15 pixels - align counter to 16 bytes
+
+ sub \reg1, \reg1, #1 @ reg1 = (counter & 15) - 1
+
+ .ifeqs "\qB", "q14"
+ vld1.8 {q8}, [\src1] @ S1 = [src - srcstride]
+ add \src1, \src1, \reg1 @ src1 += (counter & 15) - 1
+ .endif
+
+ vld1.8 {q9}, [\src2] @ S2 = [src ]
+ add \src2, \src2, \reg1 @ src2 += (counter & 15) - 1
+
+ .ifeqs "\qH", "q15"
+ vld1.8 {q10}, [\src3] @ S3 = [src + srcstride]
+ add \src3, \src3, \reg1 @ src3 += (counter & 15) - 1
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < S >
+
+ vmov \qB, q8 @ S1prev = S1 < T >
+ .endif
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < V >
+
+ vmov q12, q9 @ S2prev = S2 < C >
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < X >
+
+ vmov \qH, q10 @ S3prev = S3 < Y >
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q3, \qB, q8, #1 @ S1sr = (S1prev >> 8) | ... < U >
+ .endif
+
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | ... < W >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q5, \qH, q10, #1 @ S3sr = (S3prev >> 8) | ... < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .else
+ vld1.8 {d29[7]}, [\src1]! @ S1prev[15] = src[counter & 15 - 1 - srcstride]; src1++
+ .endif
+
+ add \reg1, \reg1, #1 @ reg1 = counter & 15
+
+ vld1.8 {d25[7]}, [\src2]! @ S2prev[15] = src[counter & 15 - 1]; src2++
+
+ bic \counter, \counter, #15
+
+ .ifeqs "\qH", "q15"
+ vld1.8 {d31[7]}, [\src3]! @ S3prev[15] = src[counter & 15 - 1 + srcstride]; src3++
+ .endif
+
+ ___neon_scale3x_8_8_slice \counter, \dst1, \dst2, \dst3, \reg1, \dB0, \dB1, \dH0, \dH1, 1, 0
+
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ .ifeqs "\qB", "q14"
+ vld1.8 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 16
+ .endif
+ vld1.8 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 16
+ .ifeqs "\qH", "q15"
+ vld1.8 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 16
+ .endif
+
+ @ inner loop (16 pixels per iteration)
+ 2:
+
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < A >
+ vmov \qB, q8 @ S1prev = S1 < B >
+ .endif
+
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < D >
+ vmov q12, q9 @ S2prev = S2 < E >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < G >
+ vmov \qH, q10 @ S3prev = S3 < H >
+ .endif
+
+ .ifeqs "\qB", "q14"
+ vld1.8 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 16
+ vext.8 q3, \qB, q8, #1 @ S1sr = (S1prev >> 8) | S1[0] < C >
+ .endif
+
+ vld1.8 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 16
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | S2[0] < F >
+
+ .ifeqs "\qH", "q15"
+ vld1.8 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 16
+ vext.8 q5, \qH, q10, #1 @ S3sr = (S3prev >> 8) | S3[0] < I >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < G >
+
+ vmov q5, q4 @ S3sr = S2sr < I >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < A >
+
+ vmov q3, q4 @ S1sr = S2sr < C >
+ .endif
+
+ ___neon_scale3x_8_8_slice \counter, \aldst1, \aldst2, \aldst3, \reg1, \dB0, \dB1, \dH0, \dH1, 0, 0
+
+ bhi 2b
+
+ @ last 16 pixels
+
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #15 @ S1sl = S1prev[15] | (S1 << 8) < A >
+ vmov \qB, q8 @ S1prev = S1 < B >
+ .endif
+
+ vext.8 q1, q12, q9, #15 @ S2sl = S2prev[15] | (S2 << 8) < D >
+ vmov q12, q9 @ S2prev = S2 < E >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #15 @ S3sl = S3prev[15] | (S3 << 8) < G >
+ vmov \qH, q10 @ S3prev = S3 < H >
+ .endif
+
+ .ifeqs "\qB", "q14"
+ vshr.u64 d16, d17, #(64-8) @ S1[0] = S1[15] | ...
+ .endif
+
+ vshr.u64 d18, d19, #(64-8) @ S2[0] = S2[15] | ...
+
+ .ifeqs "\qH", "q15"
+ vshr.u64 d20, d21, #(64-8) @ S3[0] = S3[15] | ...
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q3, \qB, q8, #1 @ S1sr = (S1prev >> 8) | S1[0] < C >
+ .endif
+
+ vext.8 q4, q12, q9, #1 @ S2sr = (S2prev >> 8) | S2[0] < F >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q5, \qH, q10, #1 @ S3sr = (S3prev >> 8) | S3[0] < I >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < G >
+
+ vmov q5, q4 @ S3sr = S2sr < I >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < A >
+
+ vmov q3, q4 @ S1sr = S2sr < C >
+ .endif
+
+ ___neon_scale3x_8_8_slice \counter, \aldst1, \aldst2, \aldst3, \reg1, \dB0, \dB1, \dH0, \dH1, 0, 1
+
+.endm
+
+.macro _neon_scale3x_8_8_line_first src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q12, q15, d24, d25, d30, d31, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro _neon_scale3x_8_8_line_middle src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q14, q15, d28, d29, d30, d31, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro _neon_scale3x_8_8_line_last src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_8_8_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q14, q12, d28, d29, d24, d25, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro neon_scale3x_8_8_line part, src1, src2, src3, counter, dst1, dst2, dst3, reg1, srcalign16, dstalign8
+ .ifeq \srcalign16
+
+ .ifeq \dstalign8
+ _neon_scale3x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1, \src2, \src3, \dst1, \dst2, \dst3
+ .else
+ _neon_scale3x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1, \src2, \src3, \dst1:64, \dst2:64, \dst3:64
+ .endif
+
+ .else
+
+ .ifeq \dstalign8
+ _neon_scale3x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2, \dst3
+ .else
+ _neon_scale3x_8_8_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1:128, \src2:128, \src3:128, \dst1:64, \dst2:64, \dst3:64
+ .endif
+
+ .endif
+.endm
+
+
+.macro ___neon_scale3x_16_16_slice counter, dst1, dst2, dst3, reg1, dB0, dB1, dH0, dH1, first7, last8
+
+@ d12 = E0[0]
+@ d13 = E1[0]
+@ d14 = E2[0]
+@ d15 = tmp0[0]
+@ d22 = E3[0]
+@ d23 = tmp1[0]
+@ d24 = E4[0]
+@ d25 = E4[1]
+@ d26 = E5[0]
+@ d27 = C0[0]
+
+@ q0 = tmp2
+@ q3 = tmp3
+@ q2 = tmp4
+@ q5 = tmp5
+
+ vceq.i16 d15, \dB0, \dH0 @ tmp0[0] = < B == H >
+
+ vceq.i16 d23, d2, d8 @ tmp1[0] = < D == F >
+
+ .if \first7
+ cmp \reg1, #4
+ .endif
+
+ vceq.i16 q0, q12, q0 @ tmp2 = < E == A >
+
+ vceq.i16 q3, q12, q3 @ tmp3 = < E == C >
+
+ .if \first7
+ lsl \reg1, #1 @ reg1 = 2 * (counter & 7)
+ .endif
+
+ vorr d27, d15, d23 @ C0[0] = < B == H || D == F >
+
+ vceq.i16 d22, d2, \dB0 @ E3[0] = < D == B >
+
+ vceq.i16 d26, \dB0, d8 @ E5[0] = < B == F >
+
+ vceq.i16 q2, q12, q2 @ tmp4 = < E == G >
+
+ vceq.i16 q5, q12, q5 @ tmp5 = < E == I >
+
+ .if \first7
+ add \reg1, \reg1, \reg1, lsl #1 @ reg1 = 2 * 3 * (counter & 7)
+ .endif
+
+ vorn d15, d6, d22 @ tmp0[0] = < (E == C) || !(D == B) >
+
+ vorn d23, d0, d26 @ tmp1[0] = < (E == A) || !(B == F) >
+
+ vorn d12, d27, d22 @ E0[0] = < C0 || !(D == B) >
+
+ .if \first7
+ sub \reg1, \reg1, #(3*2*4) @ reg1 -= 3*2*4
+ .endif
+
+ vand d13, d15, d23 @ E1[0] = < ((E == C) || !(D == B)) && ((E == A) || !(B == F)) >
+
+ vorn d14, d27, d26 @ E2[0] = < C0 || !(B == F) >
+
+ vorr d13, d27, d13 @ E1[0] = < C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F))) >
+
+ vbsl d12, d24, d2 @ E0[0] = < (C0 || !(D == B)) ? E : D >
+
+ vbsl d14, d24, d8 @ E2[0] = < (C0 || !(B == F)) ? E : F >
+
+ vbsl d13, d24, \dB0 @ E1[0] = < (C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F)))) ? E : B >
+
+ vceq.i16 d15, d2, \dH0 @ tmp0[0] = < D == H >
+
+ vceq.i16 d23, \dH0, d8 @ tmp1[0] = < H == F >
+
+ vorn d22, d4, d22 @ E3[0] = < (E == G) || !(D == B) >
+ vst3.16 {d12-d14}, [\dst1]! @ [dst] = E0,E1,E2; dst1 += 3*2*4
+
+ .if \first7
+ addls \dst1, \dst1, \reg1 @ dst1 += reg1
+ .endif
+
+ vorn d26, d10, d26 @ E5[0] = < (E == I) || !(B == F) >
+
+@ d12 = tmp6[0]
+@ d13 = tmp7[0]
+
+ vorn d12, d0, d15 @ tmp6[0] = < (E == A) || !(D == H) >
+
+ vorn d13, d6, d23 @ tmp7[0] = < (E == C) || !(H == F) >
+
+ vand d22, d22, d12 @ E3[0] = < ((E == G) || !(D == B)) && ((E == A) || !(D == H)) >
+
+ vand d26, d26, d13 @ E5[0] = < ((E == I) || !(B == F)) && ((E == C) || !(H == F)) >
+
+@ d12 = E6[0]
+@ d13 = E7[0]
+@ d14 = E8[0]
+
+ vorr d22, d27, d22 @ E3[0] = < C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H))) >
+
+ vorr d26, d27, d26 @ E5[0] = < C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F))) >
+
+ vbsl d22, d24, d2 @ E3[0] = < (C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H)))) ? E : D >
+
+ vbsl d26, d24, d8 @ E5[0] = < (C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F)))) ? E : F >
+
+ vorn d13, d10, d15 @ E7[0] = < (E == I) || !(D == H) >
+
+ vorn d12, d27, d15 @ E6[0] = < C0 || !(D == H) >
+ vst3.16 {d22,d24,d26}, [\dst2]! @ [dst + dststride] = E3,E4,E5; dst2 += 3*2*4
+
+ .if \first7
+ addls \dst2, \dst2, \reg1 @ dst2 += reg1
+ .endif
+
+ vorn d15, d4, d23 @ tmp0[0] = < (E == G) || !(H == F) >
+
+ vorn d14, d27, d23 @ E8[0] = < C0 || !(H == F) >
+
+ vand d13, d13, d15 @ E7[0] = < ((E == I) || !(D == H)) && ((E == G) || !(H == F)) >
+
+ vbsl d12, d24, d2 @ E6[0] = < (C0 || !(D == H)) ? E : D >
+
+ vorr d13, d27, d13 @ E7[0] = < C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F))) >
+
+ vbsl d14, d24, d8 @ E8[0] = < (C0 || !(H == F)) ? E : F >
+
+ vbsl d13, d24, \dH0 @ E7[0] = < (C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F)))) ? E : H >
+
+@ d15 = tmp0[1]
+@ d22 = tmp1[1]
+@ d23 = E3[1]
+@ d24 = E4[0]
+@ d25 = E4[1]
+@ d26 = C0[1]
+@ d27 = E5[1]
+
+ vceq.i16 d15, \dB1, \dH1 @ tmp0[1] = < B == H >
+
+ vceq.i16 d22, d3, d9 @ tmp1[1] = < D == F >
+
+ vceq.i16 d23, d3, \dB1 @ E3[1] = < D == B >
+ vst3.16 {d12-d14}, [\dst3]! @ [dst + 2 * dststride] = E6,E7,E8; dst3 += 3*2*4
+
+ .if \first7
+ addls \dst3, \dst3, \reg1 @ dst3 += reg1
+ bls 0f
+ .endif
+
+@ d12 = E0[1]
+@ d13 = E1[1]
+@ d14 = E2[1]
+
+ vorr d26, d15, d22 @ C0[1] = < B == H || D == F >
+
+ vceq.i16 d27, \dB1, d9 @ E5[1] = < B == F >
+
+ vorn d15, d7, d23 @ tmp0[1] = < (E == C) || !(D == B) >
+
+ vorn d22, d1, d27 @ tmp1[1] = < (E == A) || !(B == F) >
+
+ vorn d12, d26, d23 @ E0[1] = < C0 || !(D == B) >
+
+ vand d13, d15, d22 @ E1[1] = < ((E == C) || !(D == B)) && ((E == A) || !(B == F)) >
+
+ vorn d14, d26, d27 @ E2[1] = < C0 || !(B == F) >
+
+ vorr d13, d26, d13 @ E1[1] = < C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F))) >
+
+ vbsl d12, d25, d3 @ E0[1] = < (C0 || !(D == B)) ? E : D >
+
+ vbsl d14, d25, d9 @ E2[1] = < (C0 || !(B == F)) ? E : F >
+
+ vbsl d13, d25, \dB1 @ E1[1] = < (C0 || (((E == C) || !(D == B)) && ((E == A) || !(B == F)))) ? E : B >
+
+ vceq.i16 d15, d3, \dH1 @ tmp0[1] = < D == H >
+
+ vceq.i16 d22, \dH1, d9 @ tmp1[1] = < H == F >
+
+ vorn d23, d5, d23 @ E3[1] = < (E == G) || !(D == B) >
+ .ifeq \first7
+ vst3.16 {d12-d14}, [\dst1]! @ [dst] = E0,E1,E2; dst1 += 3*2*4
+ .else
+ vst3.16 {d12-d14}, [\dst1], \reg1 @ [dst] = E0,E1,E2; dst1 += reg1
+ .endif
+
+ vorn d27, d11, d27 @ E5[1] = < (E == I) || !(B == F) >
+
+@ d12 = tmp6[1]
+@ d13 = tmp7[1]
+
+ vorn d12, d1, d15 @ tmp6[1] = < (E == A) || !(D == H) >
+
+ vorn d13, d7, d22 @ tmp7[1] = < (E == C) || !(H == F) >
+
+ vand d23, d23, d12 @ E3[1] = < ((E == G) || !(D == B)) && ((E == A) || !(D == H)) >
+
+ vand d27, d27, d13 @ E5[1] = < ((E == I) || !(B == F)) && ((E == C) || !(H == F)) >
+
+@ d12 = E6[1]
+@ d13 = E7[1]
+@ d14 = E8[1]
+
+ vorr d23, d26, d23 @ E3[1] = < C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H))) >
+
+ vorr d27, d26, d27 @ E5[1] = < C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F))) >
+
+ vbsl d23, d25, d3 @ E3[1] = < (C0 || (((E == G) || !(D == B)) && ((E == A) || !(D == H)))) ? E : D >
+
+ vbsl d27, d25, d9 @ E5[1] = < (C0 || (((E == I) || !(B == F)) && ((E == C) || !(H == F)))) ? E : F >
+
+ vorn d13, d11, d15 @ E7[1] = < (E == I) || !(D == H) >
+
+ vorn d12, d26, d15 @ E6[1] = < C0 || !(D == H) >
+ .ifeq \first7
+ vst3.16 {d23,d25,d27}, [\dst2]! @ [dst + dststride] = E3,E4,E5; dst2 += 3*2*4
+ .else
+ vst3.16 {d23,d25,d27}, [\dst2], \reg1 @ [dst + dststride] = E3,E4,E5; dst2 += reg1
+ .endif
+
+ vorn d15, d5, d22 @ tmp0[1] = < (E == G) || !(H == F) >
+
+ vorn d14, d26, d22 @ E8[1] = < C0 || !(H == F) >
+
+ vand d13, d13, d15 @ E7[1] = < ((E == I) || !(D == H)) && ((E == G) || !(H == F)) >
+
+ vbsl d12, d25, d3 @ E6[1] = < (C0 || !(D == H)) ? E : D >
+
+ vorr d13, d26, d13 @ E7[1] = < C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F))) >
+
+ vbsl d14, d25, d9 @ E8[1] = < (C0 || !(H == F)) ? E : F >
+
+ vbsl d13, d25, \dH1 @ E7[1] = < (C0 || (((E == I) || !(D == H)) && ((E == G) || !(H == F)))) ? E : H >
+
+ .ifeq \first7
+
+ .ifeq \last8
+ sub \counter, \counter, #8 @ counter -= 8
+ cmp \counter, #8
+ .endif
+
+ vst3.16 {d12-d14}, [\dst3]! @ [dst + 2 * dststride] = E6,E7,E8; dst3 += 3*2*4
+ .else
+ vst3.16 {d12-d14}, [\dst3], \reg1 @ [dst + 2 * dststride] = E6,E7,E8; dst3 += reg1
+
+ 0:
+ .endif
+
+.endm
+
+.macro __neon_scale3x_16_16_line src1, src2, src3, counter, dst1, dst2, dst3, reg1, qB, qH, dB0, dB1, dH0, dH1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+
+ .ifeqs "\qB", "q14"
+ vld1.16 {d29[3]}, [\src1] @ S1prev[7] = src[-srcstride]
+ .endif
+ vld1.16 {d25[3]}, [\src2] @ S2prev[7] = src[0]
+ .ifeqs "\qH", "q15"
+ vld1.16 {d31[3]}, [\src3] @ S3prev[7] = src[srcstride]
+ .endif
+ andS \reg1, \counter, #7 @ reg1 = counter & 7
+
+ .ifnes "\qB", "q14"
+ add \src1, \src1, \counter, lsl #1 @ src1 += 2 * counter
+ .endif
+ .ifnes "\qH", "q15"
+ add \src3, \src3, \counter, lsl #1 @ src3 += 2 * counter
+ .endif
+ beq 1f
+
+ @ first 1-7 pixels - align counter to 16 bytes
+
+ sub \reg1, \reg1, #1 @ reg1 = (counter & 7) - 1
+
+ .ifeqs "\qB", "q14"
+ vld1.16 {q8}, [\src1] @ S1 = [src - srcstride]
+ add \src1, \src1, \reg1, lsl #1 @ src1 += 2 * ((counter & 7) - 1)
+ .endif
+
+ vld1.16 {q9}, [\src2] @ S2 = [src ]
+ add \src2, \src2, \reg1, lsl #1 @ src2 += 2 * ((counter & 7) - 1)
+
+ .ifeqs "\qH", "q15"
+ vld1.16 {q10}, [\src3] @ S3 = [src + srcstride]
+ add \src3, \src3, \reg1, lsl #1 @ src3 += 2 * ((counter & 7) - 1)
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < S >
+
+ vmov \qB, q8 @ S1prev = S1 < T >
+ .endif
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < V >
+
+ vmov q12, q9 @ S2prev = S2 < C >
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < X >
+
+ vmov \qH, q10 @ S3prev = S3 < Y >
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q3, \qB, q8, #2 @ S1sr = (S1prev >> 16) | ... < U >
+ .endif
+
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | ... < W >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q5, \qH, q10, #2 @ S3sr = (S3prev >> 16) | ... < Z >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < X >
+
+ vmov q5, q4 @ S3sr = S2sr < Z >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < S >
+
+ vmov q3, q4 @ S1sr = S2sr < U >
+ .else
+ vld1.16 {d29[3]}, [\src1]! @ S1prev[7] = src[counter & 7 - 1 - srcstride]; src1 += 2
+ .endif
+
+ add \reg1, \reg1, #1 @ reg1 = counter & 7
+
+ vld1.16 {d25[3]}, [\src2]! @ S2prev[7] = src[counter & 7 - 1]; src2 += 2
+
+ bic \counter, \counter, #7
+
+ .ifeqs "\qH", "q15"
+ vld1.16 {d31[3]}, [\src3]! @ S3prev[7] = src[counter & 7 - 1 + srcstride]; src3 += 2
+ .endif
+
+ ___neon_scale3x_16_16_slice \counter, \dst1, \dst2, \dst3, \reg1, \dB0, \dB1, \dH0, \dH1, 1, 0
+
+
+ @ counter is aligned to 16 bytes
+
+ 1:
+ .ifeqs "\qB", "q14"
+ vld1.16 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 2*8
+ .endif
+ vld1.16 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 2*8
+ .ifeqs "\qH", "q15"
+ vld1.16 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 2*8
+ .endif
+
+ @ inner loop (8 pixels per iteration)
+ 2:
+
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < A >
+ vmov \qB, q8 @ S1prev = S1 < B >
+ .endif
+
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < D >
+ vmov q12, q9 @ S2prev = S2 < E >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < G >
+ vmov \qH, q10 @ S3prev = S3 < H >
+ .endif
+
+ .ifeqs "\qB", "q14"
+ vld1.16 {q8}, [\alsrc1]! @ S1 = [src - srcstride]; src1 += 2*8
+ vext.8 q3, \qB, q8, #2 @ S1sr = (S1prev >> 16) | S1[0] < C >
+ .endif
+
+ vld1.16 {q9}, [\alsrc2]! @ S2 = [src ]; src2 += 2*8
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | S2[0] < F >
+
+ .ifeqs "\qH", "q15"
+ vld1.16 {q10}, [\alsrc3]! @ S3 = [src + srcstride]; src3 += 2*8
+ vext.8 q5, \qH, q10, #2 @ S3sr = (S3prev >> 16) | S3[0] < I >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < G >
+
+ vmov q5, q4 @ S3sr = S2sr < I >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < A >
+
+ vmov q3, q4 @ S1sr = S2sr < C >
+ .endif
+
+ ___neon_scale3x_16_16_slice \counter, \aldst1, \aldst2, \aldst3, \reg1, \dB0, \dB1, \dH0, \dH1, 0, 0
+
+ bhi 2b
+
+ @ last 8 pixels
+
+ .ifeqs "\qB", "q14"
+ vext.8 q0, \qB, q8, #14 @ S1sl = S1prev[7] | (S1 << 16) < A >
+ vmov \qB, q8 @ S1prev = S1 < B >
+ .endif
+
+ vext.8 q1, q12, q9, #14 @ S2sl = S2prev[7] | (S2 << 16) < D >
+ vmov q12, q9 @ S2prev = S2 < E >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q2, \qH, q10, #14 @ S3sl = S3prev[7] | (S3 << 16) < G >
+ vmov \qH, q10 @ S3prev = S3 < H >
+ .endif
+
+ .ifeqs "\qB", "q14"
+ vshr.u64 d16, d17, #(64-16) @ S1[0] = S1[7] | ...
+ .endif
+
+ vshr.u64 d18, d19, #(64-16) @ S2[0] = S2[7] | ...
+
+ .ifeqs "\qH", "q15"
+ vshr.u64 d20, d21, #(64-16) @ S3[0] = S3[7] | ...
+ .endif
+ .ifeqs "\qB", "q14"
+ vext.8 q3, \qB, q8, #2 @ S1sr = (S1prev >> 16) | S1[0] < C >
+ .endif
+
+ vext.8 q4, q12, q9, #2 @ S2sr = (S2prev >> 16) | S2[0] < F >
+
+ .ifeqs "\qH", "q15"
+ vext.8 q5, \qH, q10, #2 @ S3sr = (S3prev >> 16) | S3[0] < I >
+ .else
+ vmov q2, q1 @ S3sl = S2sl < G >
+
+ vmov q5, q4 @ S3sr = S2sr < I >
+ .endif
+
+ .ifnes "\qB", "q14"
+ vmov q0, q1 @ S1sl = S2sl < A >
+
+ vmov q3, q4 @ S1sr = S2sr < C >
+ .endif
+
+ ___neon_scale3x_16_16_slice \counter, \aldst1, \aldst2, \aldst3, \reg1, \dB0, \dB1, \dH0, \dH1, 0, 1
+
+.endm
+
+.macro _neon_scale3x_16_16_line_first src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q12, q15, d24, d25, d30, d31, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro _neon_scale3x_16_16_line_middle src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q14, q15, d28, d29, d30, d31, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro _neon_scale3x_16_16_line_last src1, src2, src3, counter, dst1, dst2, dst3, reg1, alsrc1, alsrc2, alsrc3, aldst1, aldst2, aldst3
+ __neon_scale3x_16_16_line \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, q14, q12, d28, d29, d24, d25, \alsrc1, \alsrc2, \alsrc3, \aldst1, \aldst2, \aldst3
+.endm
+
+.macro neon_scale3x_16_16_line part, src1, src2, src3, counter, dst1, dst2, dst3, reg1, srcalign16, dstalign8
+ .ifeq \srcalign16
+
+ .ifeq \dstalign8
+ _neon_scale3x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1, \src2, \src3, \dst1, \dst2, \dst3
+ .else
+ _neon_scale3x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1, \src2, \src3, \dst1:64, \dst2:64, \dst3:64
+ .endif
+
+ .else
+
+ .ifeq \dstalign8
+ _neon_scale3x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1:128, \src2:128, \src3:128, \dst1, \dst2, \dst3
+ .else
+ _neon_scale3x_16_16_line_\part \src1, \src2, \src3, \counter, \dst1, \dst2, \dst3, \reg1, \src1:128, \src2:128, \src3:128, \dst1:64, \dst2:64, \dst3:64
+ .endif
+
+ .endif
+.endm
+
diff --git a/arm/neon_scale3x.h b/arm/neon_scale3x.h new file mode 100644 index 0000000..6abd150 --- /dev/null +++ b/arm/neon_scale3x.h @@ -0,0 +1,44 @@ +/** + * + * Copyright (C) 2012 Roman Pauer + * + * Permission is hereby granted, free of charge, to any person obtaining a copy of + * this software and associated documentation files (the "Software"), to deal in + * the Software without restriction, including without limitation the rights to + * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies + * of the Software, and to permit persons to whom the Software is furnished to do + * so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#if !defined(_NEON_SCALE3X_H_INCLUDED_) +#define _NEON_SCALE3X_H_INCLUDED_ + +#include <inttypes.h> + +#ifdef __cplusplus +extern "C" { +#endif + +extern void neon_scale3x_8_8(const uint8_t *src, uint8_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); +extern void neon_scale3x_16_16(const uint16_t *src, uint16_t *dst, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +extern void neon_scale3x_8_16(const uint8_t *src, uint16_t *dst, const uint32_t *palette, unsigned int width, unsigned int srcstride, unsigned int dststride, unsigned int height); + +#ifdef __cplusplus +} +#endif + + +#endif /* _NEON_SCALE3X_H_INCLUDED_ */ diff --git a/arm/neon_scale_license.txt b/arm/neon_scale_license.txt new file mode 100644 index 0000000..73e622b --- /dev/null +++ b/arm/neon_scale_license.txt @@ -0,0 +1,19 @@ +Copyright (C) 2012 Roman Pauer + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. |