generic-library/openh264
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge pull request #1863 from mstorsjo/aarch64-official-instructions

Browse Source 
Use the correct, official syntax for aarch64 mov instructions
This commit is contained in:
sijchen 2015-03-18 11:52:02 +08:00
commit 499be24cda
6 changed files with 172 additions and 180 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
codec/common/arm64/arm_arch64_common_macro.S
View File

@ -59,12 +59,4 @@ ret
.endfunc
.endm
.macro mov.16b arg0, arg1
mov \arg0\().16b, \arg1\().16b
.endm
.macro mov.8b arg0, arg1
mov \arg0\().8b, \arg1\().8b
.endm
#endif

52
codec/common/arm64/deblocking_aarch64_neon.S
View File

@ -162,7 +162,7 @@
.endm
.macro DIFF_LUMA_EQ4_MASK
mov.16b $3, $2
mov $3.16b, $2.16b
bsl $3.16b, $0.16b, $1.16b
.endm
@ -420,7 +420,7 @@ bs_mv_check_jump1:
.endm
.macro DIFF_LUMA_EQ4_MASK arg0, arg1, arg2, arg3
mov.16b \arg3, \arg2
mov \arg3\().16b, \arg2\().16b
bsl \arg3\().16b, \arg0\().16b, \arg1\().16b
.endm
@ -652,10 +652,10 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockLumaEq4V_AArch64_neon
and v22.16b, v22.16b, v20.16b //(iDetaP0Q0 < ((iAlpha >> 2) + 2))&&bDetaQ2Q0
and v20.16b, v20.16b, v18.16b //(iDetaP0Q0 < iAlpha) && bDetaP1P0 && bDetaQ1Q0&&(iDetaP0Q0 < ((iAlpha >> 2) + 2))
mov.16b v23, v21
mov.16b v24, v21
mov v23.16b, v21.16b
mov v24.16b, v21.16b
mov.16b v25, v0
mov v25.16b, v0.16b
DIFF_LUMA_EQ4_P2P1P0_1 v0, v1, v2, v3, v4, v5, v23, v19, v17, v16
DIFF_LUMA_EQ4_P2P1P0_2 v25, v1, v2, v3, v4, v5, v24, v19, v17, v16
ins v0.d[1], v25.d[1]
@ -669,9 +669,9 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockLumaEq4V_AArch64_neon
st1 {v17.16b}, [x3], x1
mov.16b v23, v22
mov.16b v24, v22
mov.16b v25, v7
mov v23.16b, v22.16b
mov v24.16b, v22.16b
mov v25.16b, v7.16b
DIFF_LUMA_EQ4_P2P1P0_1 v7, v6, v5, v4, v3, v2, v23, v19, v17, v16
DIFF_LUMA_EQ4_P2P1P0_2 v25, v6, v5, v4, v3, v2, v24, v19, v17, v16
ins v7.d[1], v25.d[1]
@ -728,7 +728,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockLumaLt4H_AArch64_neon //uint8_t* pPix, int32_
sub v18.16b, v18.16b, v6.16b // -iTc0: 0000, 1111, 2222, 3333
DIFF_LUMA_LT4_P1_Q1 v0, v1, v2, v3, v17, v18, v6, v7, v19, v20 //Use Tmp v23,v24
mov.16b v25, v19
mov v25.16b, v19.16b
DIFF_LUMA_LT4_P1_Q1 v5, v4, v3, v2, v17, v18, v6, v7, v21, v22 //Use Tmp v23,v24
@ -749,11 +749,11 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockLumaLt4H_AArch64_neon //uint8_t* pPix, int32_
EXTRACT_DELTA_INTO_TWO_PART v19, v20
uqadd v2.16b, v2.16b, v20.16b
uqsub v2.16b, v2.16b, v19.16b
mov.16b v26, v2
mov v26.16b, v2.16b
uqsub v3.16b, v3.16b, v20.16b
uqadd v3.16b, v3.16b, v19.16b
mov.16b v27, v3
mov.16b v28, v21
mov v27.16b, v3.16b
mov v28.16b, v21.16b
sub x0, x0, #2
add x2, x0, x1
@ -816,37 +816,37 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockLumaEq4H_AArch64_neon
and v22.16b, v22.16b, v20.16b //(iDetaP0Q0 < ((iAlpha >> 2) + 2))&&bDetaQ2Q0
and v20.16b, v20.16b, v18.16b //(iDetaP0Q0 < iAlpha) && bDetaP1P0 && bDetaQ1Q0&&(iDetaP0Q0 < ((iAlpha >> 2) + 2))
mov.16b v23, v21
mov.16b v24, v21
mov v23.16b, v21.16b
mov v24.16b, v21.16b
mov.16b v25, v0
mov v25.16b, v0.16b
DIFF_LUMA_EQ4_P2P1P0_1 v0, v1, v2, v3, v4, v5, v23, v19, v17, v16
DIFF_LUMA_EQ4_P2P1P0_2 v25, v1, v2, v3, v4, v5, v24, v19, v17, v16
ins v0.d[1], v25.d[1]
ins v23.d[1], v24.d[1]
and v21.16b, v20.16b, v21.16b
DIFF_LUMA_EQ4_MASK v19, v1, v21, v17
mov.16b v26, v17
mov v26.16b, v17.16b
DIFF_LUMA_EQ4_MASK v0, v2, v21, v17
mov.16b v27, v17
mov v27.16b, v17.16b
DIFF_LUMA_EQ4_MASK v23, v3, v18, v17
mov.16b v28, v17
mov v28.16b, v17.16b
mov.16b v23, v22
mov.16b v24, v22
mov.16b v25, v7
mov v23.16b, v22.16b
mov v24.16b, v22.16b
mov v25.16b, v7.16b
DIFF_LUMA_EQ4_P2P1P0_1 v7, v6, v5, v4, v3, v2, v23, v19, v17, v16
DIFF_LUMA_EQ4_P2P1P0_2 v25, v6, v5, v4, v3, v2, v24, v19, v17, v16
ins v7.d[1], v25.d[1]
ins v23.d[1], v24.d[1]
and v22.16b, v20.16b, v22.16b
DIFF_LUMA_EQ4_MASK v23, v4, v18, v17
mov.16b v29, v17
mov v29.16b, v17.16b
DIFF_LUMA_EQ4_MASK v7, v5, v22, v17
mov.16b v30, v17
mov v30.16b, v17.16b
DIFF_LUMA_EQ4_MASK v19, v6, v22, v17
mov.16b v31, v17
mov v31.16b, v17.16b
STORE_LUMA_DATA_3 v26, v27, v28, v29, v30, v31, 0
STORE_LUMA_DATA_3 v26, v27, v28, v29, v30, v31, 1
@ -1013,7 +1013,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockChromaEq4V_AArch64_neon //uint8_t* pPixCb, ui
DIFF_CHROMA_EQ4_P0Q0_1 v0, v1, v2, v3, v18, v19, v20, v21
DIFF_CHROMA_EQ4_P0Q0_2 v0, v1, v2, v3, v18, v19, v20, v21
mov.16b v6, v7
mov v6.16b, v7.16b
bsl v6.16b, v20.16b, v1.16b
bsl v7.16b, v21.16b, v2.16b
@ -1059,7 +1059,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN DeblockChromaEq4H_AArch64_neon //uint8_t* pPixCb, ui
DIFF_CHROMA_EQ4_P0Q0_1 v0, v1, v2, v3, v18, v19, v20, v21
DIFF_CHROMA_EQ4_P0Q0_2 v0, v1, v2, v3, v18, v19, v20, v21
mov.16b v6, v7
mov v6.16b, v7.16b
bsl v6.16b, v20.16b, v1.16b
bsl v7.16b, v21.16b, v2.16b

4
codec/common/arm64/expand_picture_aarch64_neon.S
View File

@ -46,8 +46,8 @@ _expand_picture_luma_loop2:
add x6, x4, #1
ld1r {v0.16b}, [x7], x1
ld1r {v2.16b}, [x4], x1
mov.16b v1, v0
mov.16b v3, v2
mov v1.16b, v0.16b
mov v3.16b, v2.16b
st2 {v0.16b, v1.16b}, [x5]
st2 {v2.16b, v3.16b}, [x6]
sub x8, x8, #1

260
codec/common/arm64/mc_aarch64_neon.S
View File

@ -696,12 +696,12 @@ w16_xy_01_mc_luma_loop:
FILTER_6TAG_8BITS2_AVERAGE_WITH_0 v3, v4, v5, v6, v7, v2, v20, v0, v1
st1 {v20.16b}, [x2], x3 //write 16Byte : 7 line
mov.16b v3, v5
mov.16b v5, v7
mov.16b v7, v2
mov.16b v2, v4
mov.16b v4, v6
mov.16b v6, v7
mov v3.16b, v5.16b
mov v5.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v4.16b
mov v4.16b, v6.16b
mov v6.16b, v7.16b
sub x4, x4, #8
cbnz x4, w16_xy_01_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -746,12 +746,12 @@ w8_xy_01_mc_luma_loop:
FILTER_6TAG_8BITS1_AVERAGE_WITH_0 v5, v6, v7, v2, v3, v4, v20, v0, v1
st1 {v20.8b}, [x2], x3 //write 8Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v7, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v7
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v7.16b
sub x4, x4, #4
cbnz x4, w8_xy_01_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -799,14 +799,14 @@ w4_xy_01_mc_luma_loop:
st1 {v20.s}[1], [x2], x3 //write 4Byte : 3 line
mov v4.s[0], v3.s[1]
mov.8b v21, v6
mov.8b v6, v4
mov.8b v4, v2
mov.8b v2, v21
mov.8b v21, v3
mov.8b v3, v7
mov.8b v7, v5
mov.8b v5, v21
mov v21.8b, v6.8b
mov v6.8b, v4.8b
mov v4.8b, v2.8b
mov v2.8b, v21.8b
mov v21.8b, v3.8b
mov v3.8b, v7.8b
mov v7.8b, v5.8b
mov v5.8b, v21.8b
sub x4, x4, #4
cbnz x4, w4_xy_01_mc_luma_loop
@ -885,12 +885,12 @@ w16_xy_03_mc_luma_loop:
FILTER_6TAG_8BITS2_AVERAGE_WITH_1 v3, v4, v5, v6, v7, v2, v20, v0, v1
st1 {v20.16b}, [x2], x3 //write 16Byte : 7 line
mov.16b v3, v5
mov.16b v5, v7
mov.16b v7, v2
mov.16b v2, v4
mov.16b v4, v6
mov.16b v6, v7
mov v3.16b, v5.16b
mov v5.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v4.16b
mov v4.16b, v6.16b
mov v6.16b, v7.16b
sub x4, x4, #8
cbnz x4, w16_xy_03_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -935,12 +935,12 @@ w8_xy_03_mc_luma_loop:
FILTER_6TAG_8BITS1_AVERAGE_WITH_1 v5, v6, v7, v2, v3, v4, v20, v0, v1
st1 {v20.8b}, [x2], x3 //write 8Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v7, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v7
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v7.16b
sub x4, x4, #4
cbnz x4, w8_xy_03_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -988,14 +988,14 @@ w4_xy_03_mc_luma_loop:
st1 {v20.s}[1], [x2], x3 //write 4Byte : 3 line
mov v4.s[0], v3.s[1]
mov.8b v21, v6
mov.8b v6, v4
mov.8b v4, v2
mov.8b v2, v21
mov.8b v21, v3
mov.8b v3, v7
mov.8b v7, v5
mov.8b v5, v21
mov v21.8b, v6.8b
mov v6.8b, v4.8b
mov v4.8b, v2.8b
mov v2.8b, v21.8b
mov v21.8b, v3.8b
mov v3.8b, v7.8b
mov v7.8b, v5.8b
mov v5.8b, v21.8b
sub x4, x4, #4
cbnz x4, w4_xy_03_mc_luma_loop
@ -1074,12 +1074,12 @@ w16_xy_02_mc_luma_loop:
FILTER_6TAG_8BITS2 v3, v4, v5, v6, v7, v2, v20, v0, v1
st1 {v20.16b}, [x2], x3 //write 16Byte : 7 line
mov.16b v3, v5
mov.16b v5, v7
mov.16b v7, v2
mov.16b v2, v4
mov.16b v4, v6
mov.16b v6, v7
mov v3.16b, v5.16b
mov v5.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v4.16b
mov v4.16b, v6.16b
mov v6.16b, v7.16b
sub x4, x4, #8
cbnz x4, w16_xy_02_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -1124,12 +1124,12 @@ w8_xy_02_mc_luma_loop:
FILTER_6TAG_8BITS1 v5, v6, v7, v2, v3, v4, v20, v0, v1
st1 {v20.8b}, [x2], x3 //write 8Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v7, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v7
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v7.16b
sub x4, x4, #4
cbnz x4, w8_xy_02_mc_luma_loop
WELS_ASM_AARCH64_FUNC_END
@ -1177,14 +1177,14 @@ w4_xy_02_mc_luma_loop:
st1 {v20.s}[1], [x2], x3 //write 4Byte : 3 line
mov v4.s[0], v3.s[1]
mov.8b v21, v6
mov.8b v6, v4
mov.8b v4, v2
mov.8b v2, v21
mov.8b v21, v3
mov.8b v3, v7
mov.8b v7, v5
mov.8b v5, v21
mov v21.8b, v6.8b
mov v6.8b, v4.8b
mov v4.8b, v2.8b
mov v2.8b, v21.8b
mov v21.8b, v3.8b
mov v3.8b, v7.8b
mov v7.8b, v5.8b
mov v5.8b, v21.8b
sub x4, x4, #4
cbnz x4, w4_xy_02_mc_luma_loop
@ -1326,26 +1326,26 @@ w16_hv_mc_luma_loop:
FILTER_3_IN_16BITS_TO_8BITS2 v23, v24, v25, v26 //output to v26[1]
st1 {v26.16b}, [x2], x3 //write 16Byte : 7 line
mov.16b v5, v11
mov.16b v11, v17
mov.16b v30, v2
mov.16b v2, v8
mov.16b v8, v14
mov.16b v14, v30
mov v5.16b, v11.16b
mov v11.16b, v17.16b
mov v30.16b, v2.16b
mov v2.16b, v8.16b
mov v8.16b, v14.16b
mov v14.16b, v30.16b
mov.16b v6, v12
mov.16b v12, v18
mov.16b v30, v3
mov.16b v3, v9
mov.16b v9, v15
mov.16b v15, v30
mov v6.16b, v12.16b
mov v12.16b, v18.16b
mov v30.16b, v3.16b
mov v3.16b, v9.16b
mov v9.16b, v15.16b
mov v15.16b, v30.16b
mov.16b v7, v13
mov.16b v13, v19
mov.16b v30, v4
mov.16b v4, v10
mov.16b v10, v16
mov.16b v16, v30
mov v7.16b, v13.16b
mov v13.16b, v19.16b
mov v30.16b, v4.16b
mov v4.16b, v10.16b
mov v10.16b, v16.16b
mov v16.16b, v30.16b
sub x4, x4, #8
cbnz x4, w16_hv_mc_luma_loop
@ -1416,12 +1416,12 @@ w8_hv_mc_luma_loop:
st1 {v26.8b}, [x2], x3 //write 8Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v30, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v30
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v30.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v30.16b
sub x4, x4, #4
cbnz x4, w8_hv_mc_luma_loop
@ -1487,12 +1487,12 @@ w4_hv_mc_luma_loop:
st1 {v27.s}[0], [x2], x3 //write 4Byte : 2 line
st1 {v27.s}[1], [x2], x3 //write 4Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v30, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v30
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v30.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v30.16b
sub x4, x4, #4
cbnz x4, w4_hv_mc_luma_loop
@ -1735,8 +1735,8 @@ w8_mc_chroma_loop:
rshrn v17.8b, v16.8h, #6
st1 {v17.8b}, [x2], x3
mov.16b v0, v18
mov.16b v1, v19
mov v0.16b, v18.16b
mov v1.16b, v19.16b
sub x5, x5, #2
cbnz x5, w8_mc_chroma_loop
WELS_ASM_AARCH64_FUNC_END
@ -1764,8 +1764,8 @@ w4_mc_chroma_loop:
st1 {v17.s}[0], [x2], x3
st1 {v17.s}[1], [x2], x3
mov.8b v0, v18
mov.8b v1, v19
mov v0.8b, v18.8b
mov v1.8b, v19.8b
sub x5, x5, #2
cbnz x5, w4_mc_chroma_loop
WELS_ASM_AARCH64_FUNC_END
@ -1985,26 +1985,26 @@ w17_hv_mc_luma_loop:
UNPACK_FILTER_SINGLE_TAG_16BITS v26, v22, v29, v27, v28, d26
st1 {v26.b}[0], [x2], x3 //write 16th Byte : 7 line
mov.16b v5, v11
mov.16b v11, v17
mov.16b v30, v2
mov.16b v2, v8
mov.16b v8, v14
mov.16b v14, v30
mov v5.16b, v11.16b
mov v11.16b, v17.16b
mov v30.16b, v2.16b
mov v2.16b, v8.16b
mov v8.16b, v14.16b
mov v14.16b, v30.16b
mov.16b v6, v12
mov.16b v12, v18
mov.16b v30, v3
mov.16b v3, v9
mov.16b v9, v15
mov.16b v15, v30
mov v6.16b, v12.16b
mov v12.16b, v18.16b
mov v30.16b, v3.16b
mov v3.16b, v9.16b
mov v9.16b, v15.16b
mov v15.16b, v30.16b
mov.16b v7, v13
mov.16b v13, v19
mov.16b v30, v4
mov.16b v4, v10
mov.16b v10, v16
mov.16b v16, v30
mov v7.16b, v13.16b
mov v13.16b, v19.16b
mov v30.16b, v4.16b
mov v4.16b, v10.16b
mov v10.16b, v16.16b
mov v16.16b, v30.16b
sub x4, x4, #8
cbnz x4, w17_hv_mc_luma_loop
@ -2104,12 +2104,12 @@ w9_hv_mc_luma_loop:
st1 {v26.b}[0], [x2], x3 //write 8th Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v30, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v30
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v30.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v30.16b
sub x4, x4, #4
cbnz x4, w9_hv_mc_luma_loop
@ -2200,12 +2200,12 @@ w17_v_mc_luma_loop:
FILTER_6TAG_8BITS2 v3, v4, v5, v6, v7, v2, v20, v0, v1
st1 {v20.16b}, [x2], x3 //write 16Byte : 7 line
mov.16b v3, v5
mov.16b v5, v7
mov.16b v7, v2
mov.16b v2, v4
mov.16b v4, v6
mov.16b v6, v7
mov v3.16b, v5.16b
mov v5.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v4.16b
mov v4.16b, v6.16b
mov v6.16b, v7.16b
sub x4, x4, #8
cbnz x4, w17_v_mc_luma_loop
@ -2255,12 +2255,12 @@ w9_v_mc_luma_loop:
FILTER_6TAG_8BITS1 v5, v6, v7, v2, v3, v4, v20, v0, v1
st1 {v20.8b}, [x2], x3 //write 8Byte : 3 line
mov.16b v5, v3
mov.16b v3, v7
mov.16b v7, v2
mov.16b v2, v6
mov.16b v6, v4
mov.16b v4, v7
mov v5.16b, v3.16b
mov v3.16b, v7.16b
mov v7.16b, v2.16b
mov v2.16b, v6.16b
mov v6.16b, v4.16b
mov v4.16b, v7.16b
sub x4, x4, #4
cbnz x4, w9_v_mc_luma_loop

26
codec/encoder/core/arm64/reconstruct_aarch64_neon.S
View File

@ -73,7 +73,7 @@
cmgt $4.8h, $0.8h, #0 // if true, location of coef == 11111111
bif $3.16b, $1.16b, $4.16b // if (x<0) reserved part; else keep 0 untouched
shl $3.8h, $3.8h, #1
mov.16b $6, $1
mov $6.16b, $1.16b
sub $1.8h, $1.8h, $3.8h // if x > 0, -= 0; else x-= 2x
// }
.endm
@ -533,7 +533,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN WelsQuant4x4_AArch64_neon
ld1 {v2.8h}, [x1]
ld1 {v0.8h, v1.8h}, [x0]
ld1 {v3.8h}, [x2]
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS v0, v2, v3, v5, v6, v7
st1 {v2.8h}, [x0], #16
NEWQUANT_COEF_EACH_16BITS v1, v4, v3, v5, v6, v7
@ -545,7 +545,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN WelsQuant4x4Dc_AArch64_neon
ld1 {v0.8h, v1.8h}, [x0]
dup v2.8h, w1 // even ff range [0, 768]
dup v3.8h, w2
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS v0, v2, v3, v5, v6, v7
st1 {v2.8h}, [x0], #16
NEWQUANT_COEF_EACH_16BITS v1, v4, v3, v5, v6, v7
@ -559,10 +559,10 @@ WELS_ASM_AARCH64_FUNC_BEGIN WelsQuantFour4x4_AArch64_neon
.rept 4
ld1 {v0.8h, v1.8h}, [x0], #32
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS v0, v4, v3, v5, v6, v7
st1 {v4.8h}, [x1], #16
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS v1, v4, v3, v5, v6, v7
st1 {v4.8h}, [x1], #16
.endr
@ -575,36 +575,36 @@ WELS_ASM_AARCH64_FUNC_BEGIN WelsQuantFour4x4Max_AArch64_neon
mov x1, x0
ld1 {v0.8h, v1.8h}, [x0], #32
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v0, v4, v3, v5, v6, v7, v16
st1 {v4.8h}, [x1], #16
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v1, v4, v3, v5, v6, v7, v17
st1 {v4.8h}, [x1], #16 // then 1st 16 elem in v16 & v17
ld1 {v0.8h, v1.8h}, [x0], #32
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v0, v4, v3, v5, v6, v7, v18
st1 {v4.8h}, [x1], #16
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v1, v4, v3, v5, v6, v7, v19
st1 {v4.8h}, [x1], #16 // then 2st 16 elem in v18 & v19
SELECT_MAX_IN_ABS_COEF v16, v17, v18, v19, h20, h21
ld1 {v0.8h, v1.8h}, [x0], #32
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v0, v4, v3, v5, v6, v7, v16
st1 {v4.8h}, [x1], #16
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v1, v4, v3, v5, v6, v7, v17
st1 {v4.8h}, [x1], #16 // then 1st 16 elem in v16 & v17
ld1 {v0.8h, v1.8h}, [x0], #32
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v0, v4, v3, v5, v6, v7, v18
st1 {v4.8h}, [x1], #16
mov.16b v4, v2
mov v4.16b, v2.16b
NEWQUANT_COEF_EACH_16BITS_MAX v1, v4, v3, v5, v6, v7, v19
st1 {v4.8h}, [x1], #16 // then 2st 16 elem in v18 & v19

2
codec/encoder/core/arm64/svc_motion_estimation_aarch64_neon.S
View File

@ -288,7 +288,7 @@ WELS_ASM_AARCH64_FUNC_BEGIN FillQpelLocationByFeatureValue_AArch64_neon
_hash_height_loop:
mov x7, x1
mov.16b v2, v5 //mx_x_offset_x4
mov v2.16b, v5.16b //mx_x_offset_x4
_hash_width_loop:
ld1 {v0.d}[0], [x0], #8