diff options
Diffstat (limited to 'config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S')
| -rw-r--r-- | config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S | 56 |
1 files changed, 25 insertions, 31 deletions
diff --git a/config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S b/config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S index 712a534cf..6a308cb74 100644 --- a/config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S +++ b/config/arm-neon/vpx_dsp/arm/loopfilter_4_neon.asm.S @@ -1,6 +1,6 @@ @ This file was created from a .asm file @ using the ads2gas.pl script. - .syntax unified +.syntax unified @ @ Copyright (c) 2013 The WebM project authors. All Rights Reserved. @ @@ -11,18 +11,18 @@ @ be found in the AUTHORS file in the root of the source tree. @ - .global vpx_lpf_horizontal_4_neon - .type vpx_lpf_horizontal_4_neon, function - .global vpx_lpf_vertical_4_neon - .type vpx_lpf_vertical_4_neon, function - .global vpx_lpf_horizontal_4_dual_neon - .type vpx_lpf_horizontal_4_dual_neon, function - .global vpx_lpf_vertical_4_dual_neon - .type vpx_lpf_vertical_4_dual_neon, function - .arm + .global vpx_lpf_horizontal_4_neon + .type vpx_lpf_horizontal_4_neon, function + .global vpx_lpf_vertical_4_neon + .type vpx_lpf_vertical_4_neon, function + .global vpx_lpf_horizontal_4_dual_neon + .type vpx_lpf_horizontal_4_dual_neon, function + .global vpx_lpf_vertical_4_dual_neon + .type vpx_lpf_vertical_4_dual_neon, function + .arm -.text -.p2align 2 + .text + .p2align 2 @ Currently vpx only works on iterations 8 at a time. The vp8 loop filter @ works on 16 iterations at a time. @@ -38,8 +38,7 @@ @ r2 const uint8_t *blimit, @ r3 const uint8_t *limit, @ sp const uint8_t *thresh, -_vpx_lpf_horizontal_4_neon: - vpx_lpf_horizontal_4_neon: @ PROC +vpx_lpf_horizontal_4_neon: @ PROC push {lr} vld1.8 {d0[]}, [r2] @ duplicate *blimit @@ -72,7 +71,7 @@ _vpx_lpf_horizontal_4_neon: vst1.u8 {d7}, [r3,:64], r1 @ store oq1 pop {pc} - .size vpx_lpf_horizontal_4_neon, .-vpx_lpf_horizontal_4_neon @ ENDP @ |vpx_lpf_horizontal_4_neon| +.size vpx_lpf_horizontal_4_neon, .-vpx_lpf_horizontal_4_neon @ ENDP @ |vpx_lpf_horizontal_4_neon| @ Currently vpx only works on iterations 8 at a time. The vp8 loop filter @ works on 16 iterations at a time. @@ -88,8 +87,7 @@ _vpx_lpf_horizontal_4_neon: @ r2 const uint8_t *blimit, @ r3 const uint8_t *limit, @ sp const uint8_t *thresh, -_vpx_lpf_vertical_4_neon: - vpx_lpf_vertical_4_neon: @ PROC +vpx_lpf_vertical_4_neon: @ PROC push {lr} vld1.8 {d0[]}, [r2] @ duplicate *blimit @@ -140,7 +138,7 @@ _vpx_lpf_vertical_4_neon: vst4.8 {d4[7], d5[7], d6[7], d7[7]}, [r0] pop {pc} - .size vpx_lpf_vertical_4_neon, .-vpx_lpf_vertical_4_neon @ ENDP @ |vpx_lpf_vertical_4_neon| +.size vpx_lpf_vertical_4_neon, .-vpx_lpf_vertical_4_neon @ ENDP @ |vpx_lpf_vertical_4_neon| @ void filter4_8(); @ This is a helper function for the loopfilters. The invidual functions do the @@ -166,8 +164,7 @@ _vpx_lpf_vertical_4_neon: @ d5 op0 @ d6 oq0 @ d7 oq1 -_filter4_8: - filter4_8: @ PROC +filter4_8: @ PROC @ filter_mask vabd.u8 d19, d3, d4 @ m1 = abs(p3 - p2) vabd.u8 d20, d4, d5 @ m2 = abs(p2 - p1) @@ -257,7 +254,7 @@ _filter4_8: veor d7, d20, d18 @ *oq1 = u^0x80 bx lr - .size filter4_8, .-filter4_8 @ ENDP @ |filter4_8| +.size filter4_8, .-filter4_8 @ ENDP @ |filter4_8| @void vpx_lpf_horizontal_4_dual_neon(uint8_t *s, int p, @ const uint8_t *blimit0, @@ -275,8 +272,7 @@ _filter4_8: @ sp+8 const uint8_t *limit1, @ sp+12 const uint8_t *thresh1, -_vpx_lpf_horizontal_4_dual_neon: - vpx_lpf_horizontal_4_dual_neon: @ PROC +vpx_lpf_horizontal_4_dual_neon: @ PROC push {lr} ldr r12, [sp, #4] @ load thresh0 @@ -323,7 +319,7 @@ _vpx_lpf_horizontal_4_dual_neon: vpop {d8-d15} @ restore neon registers pop {pc} - .size vpx_lpf_horizontal_4_dual_neon, .-vpx_lpf_horizontal_4_dual_neon @ ENDP @ |vpx_lpf_horizontal_4_dual_neon| +.size vpx_lpf_horizontal_4_dual_neon, .-vpx_lpf_horizontal_4_dual_neon @ ENDP @ |vpx_lpf_horizontal_4_dual_neon| @void vpx_lpf_vertical_4_dual_neon(uint8_t *s, int p, @ const uint8_t *blimit0, @@ -341,8 +337,7 @@ _vpx_lpf_horizontal_4_dual_neon: @ sp+8 const uint8_t *limit1, @ sp+12 const uint8_t *thresh1, -_vpx_lpf_vertical_4_dual_neon: - vpx_lpf_vertical_4_dual_neon: @ PROC +vpx_lpf_vertical_4_dual_neon: @ PROC push {lr} ldr r12, [sp, #4] @ load thresh0 @@ -439,7 +434,7 @@ _vpx_lpf_vertical_4_dual_neon: vpop {d8-d15} @ restore neon registers pop {pc} - .size vpx_lpf_vertical_4_dual_neon, .-vpx_lpf_vertical_4_dual_neon @ ENDP @ |vpx_lpf_vertical_4_dual_neon| +.size vpx_lpf_vertical_4_dual_neon, .-vpx_lpf_vertical_4_dual_neon @ ENDP @ |vpx_lpf_vertical_4_dual_neon| @ void filter4_16(); @ This is a helper function for the loopfilters. The invidual functions do the @@ -464,8 +459,7 @@ _vpx_lpf_vertical_4_dual_neon: @ q6 op0 @ q7 oq0 @ q8 oq1 -_filter4_16: - filter4_16: @ PROC +filter4_16: @ PROC @ filter_mask vabd.u8 q11, q3, q4 @ m1 = abs(p3 - p2) @@ -558,6 +552,6 @@ _filter4_16: veor q8, q12, q10 @ *oq1 = u^0x80 bx lr - .size filter4_16, .-filter4_16 @ ENDP @ |filter4_16| +.size filter4_16, .-filter4_16 @ ENDP @ |filter4_16| - .section .note.GNU-stack,"",%progbits + .section .note.GNU-stack,"",%progbits |