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

Fix building the deblocking aarch64 assembly with gnu binutils

Browse Source
This commit is contained in:
Martin Storsjö 2014-06-17 10:10:50 +03:00
parent b9477cdb94
commit 720f8dcc52
1 changed files with 117 additions and 117 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

234
codec/common/arm64/deblocking_aarch64_neon.S
View File

@ -295,166 +295,166 @@ bs_mv_check_jump1:
#else
.macro MASK_MATRIX arg0, arg1, arg2, arg3, arg4, arg5, arg6
uabd \arg6.16b, \arg1.16b, \arg2.16b
cmhi \arg6.16b, \arg4.16b, \arg6.16b
uabd \arg6\().16b, \arg1\().16b, \arg2\().16b
cmhi \arg6\().16b, \arg4\().16b, \arg6\().16b
uabd \arg4.16b, \arg0.16b, \arg1.16b
cmhi \arg4.16b, \arg5.16b, \arg4.16b
and \arg6.16b, \arg6.16b, \arg4.16b
uabd \arg4\().16b, \arg0\().16b, \arg1\().16b
cmhi \arg4\().16b, \arg5\().16b, \arg4\().16b
and \arg6\().16b, \arg6\().16b, \arg4\().16b
uabd \arg4.16b, \arg3.16b, \arg2.16b
cmhi \arg4.16b, \arg5.16b, \arg4.16b
and \arg6.16b, \arg6.16b, \arg4.16b
uabd \arg4\().16b, \arg3\().16b, \arg2\().16b
cmhi \arg4\().16b, \arg5\().16b, \arg4\().16b
and \arg6\().16b, \arg6\().16b, \arg4\().16b
.endm
.macro DIFF_LUMA_LT4_P1_Q1 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8, arg9
//v0, v1, v2, v3, v17(beta), v18(-Tc0), v6(Tc0), v7(flag), v19, v20
urhadd \arg8.16b, \arg2.16b, \arg3.16b
uhadd \arg8.16b, \arg0.16b, \arg8.16b
usubl \arg9.8h, \arg8.8b, \arg1.8b
sqxtn \arg9.8b, \arg9.8h
usubl2 \arg8.8h, \arg8.16b, \arg1.16b
sqxtn2 \arg9.16b, \arg8.8h
smax \arg8.16b, \arg9.16b, \arg5.16b
urhadd \arg8\().16b, \arg2\().16b, \arg3\().16b
uhadd \arg8\().16b, \arg0\().16b, \arg8\().16b
usubl \arg9\().8h, \arg8\().8b, \arg1\().8b
sqxtn \arg9\().8b, \arg9\().8h
usubl2 \arg8\().8h, \arg8\().16b, \arg1\().16b
sqxtn2 \arg9\().16b, \arg8\().8h
smax \arg8\().16b, \arg9\().16b, \arg5\().16b
//
smin \arg8.16b, \arg8.16b, \arg6.16b
uabd \arg9.16b, \arg0.16b, \arg2.16b
cmhi \arg9.16b, \arg4.16b, \arg9.16b
and \arg8.16b, \arg8.16b, \arg9.16b
and \arg8.16b, \arg8.16b, \arg7.16b
add \arg8.16b, \arg1.16b, \arg8.16b
abs \arg9.16b, \arg9.16b
smin \arg8\().16b, \arg8\().16b, \arg6\().16b
uabd \arg9\().16b, \arg0\().16b, \arg2\().16b
cmhi \arg9\().16b, \arg4\().16b, \arg9\().16b
and \arg8\().16b, \arg8\().16b, \arg9\().16b
and \arg8\().16b, \arg8\().16b, \arg7\().16b
add \arg8\().16b, \arg1\().16b, \arg8\().16b
abs \arg9\().16b, \arg9\().16b
.endm
.macro DIFF_LUMA_LT4_P0_Q0_1 arg0, arg1, arg2, arg3, arg4, arg5, arg6
usubl \arg5.8h, \arg0.8b, \arg3.8b
usubl \arg6.8h, \arg2.8b, \arg1.8b
shl \arg6.8h, \arg6.8h, #2
add \arg5.8h, \arg5.8h, \arg6.8h
sqrshrn \arg4.8b, \arg5.8h, #3
usubl \arg5\().8h, \arg0\().8b, \arg3\().8b
usubl \arg6\().8h, \arg2\().8b, \arg1\().8b
shl \arg6\().8h, \arg6\().8h, #2
add \arg5\().8h, \arg5\().8h, \arg6\().8h
sqrshrn \arg4\().8b, \arg5\().8h, #3
.endm
.macro DIFF_LUMA_LT4_P0_Q0_2 arg0, arg1, arg2, arg3, arg4, arg5, arg6
usubl2 \arg5.8h, \arg0.16b, \arg3.16b
usubl2 \arg6.8h, \arg2.16b, \arg1.16b
shl \arg6.8h, \arg6.8h, #2
add \arg5.8h, \arg5.8h, \arg6.8h
sqrshrn2 \arg4.16b, \arg5.8h, #3
usubl2 \arg5\().8h, \arg0\().16b, \arg3\().16b
usubl2 \arg6\().8h, \arg2\().16b, \arg1\().16b
shl \arg6\().8h, \arg6\().8h, #2
add \arg5\().8h, \arg5\().8h, \arg6\().8h
sqrshrn2 \arg4\().16b, \arg5\().8h, #3
.endm
.macro EXTRACT_DELTA_INTO_TWO_PART arg0, arg1
cmge \arg1.16b, \arg0.16b, #0
and \arg1.16b, \arg0.16b, \arg1.16b
sub \arg0.16b, \arg1.16b, \arg0.16b
cmge \arg1\().16b, \arg0\().16b, #0
and \arg1\().16b, \arg0\().16b, \arg1\().16b
sub \arg0\().16b, \arg1\().16b, \arg0\().16b
.endm
.macro DIFF_LUMA_EQ4_P2P1P0_1 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8, arg9
uaddl \arg8.8h, \arg1.8b, \arg2.8b
uaddl \arg9.8h, \arg3.8b, \arg4.8b
add \arg9.8h, \arg9.8h, \arg8.8h
uaddl \arg8\().8h, \arg1\().8b, \arg2\().8b
uaddl \arg9\().8h, \arg3\().8b, \arg4\().8b
add \arg9\().8h, \arg9\().8h, \arg8\().8h
uaddl \arg8.8h, \arg0.8b, \arg1.8b
shl \arg8.8h, \arg8.8h, #1
add \arg8.8h, \arg9.8h, \arg8.8h
uaddl \arg8\().8h, \arg0\().8b, \arg1\().8b
shl \arg8\().8h, \arg8\().8h, #1
add \arg8\().8h, \arg9\().8h, \arg8\().8h
rshrn \arg0.8b, \arg9.8h, #2
rshrn \arg7.8b, \arg8.8h, #3
shl \arg9.8h, \arg9.8h, #1
usubl \arg8.8h, \arg5.8b, \arg1.8b
add \arg9.8h, \arg8.8h, \arg9.8h
rshrn \arg0\().8b, \arg9\().8h, #2
rshrn \arg7\().8b, \arg8\().8h, #3
shl \arg9\().8h, \arg9\().8h, #1
usubl \arg8\().8h, \arg5\().8b, \arg1\().8b
add \arg9\().8h, \arg8\().8h, \arg9\().8h
uaddl \arg8.8h, \arg2.8b, \arg5.8b
uaddw \arg8.8h, \arg8.8h, \arg2.8b
uaddw \arg8.8h, \arg8.8h, \arg3.8b
uaddl \arg8\().8h, \arg2\().8b, \arg5\().8b
uaddw \arg8\().8h, \arg8\().8h, \arg2\().8b
uaddw \arg8\().8h, \arg8\().8h, \arg3\().8b
rshrn \arg9.8b, \arg9.8h, #3
rshrn \arg8.8b, \arg8.8h, #2
bsl \arg6.8b, \arg9.8b, \arg8.8b
rshrn \arg9\().8b, \arg9\().8h, #3
rshrn \arg8\().8b, \arg8\().8h, #2
bsl \arg6\().8b, \arg9\().8b, \arg8\().8b
.endm
.macro DIFF_LUMA_EQ4_P2P1P0_2 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8, arg9
uaddl2 \arg8.8h, \arg1.16b, \arg2.16b
uaddl2 \arg9.8h, \arg3.16b, \arg4.16b
add \arg9.8h, \arg9.8h, \arg8.8h
uaddl2 \arg8\().8h, \arg1\().16b, \arg2\().16b
uaddl2 \arg9\().8h, \arg3\().16b, \arg4\().16b
add \arg9\().8h, \arg9\().8h, \arg8\().8h
uaddl2 \arg8.8h, \arg0.16b, \arg1.16b
shl \arg8.8h, \arg8.8h, #1
add \arg8.8h, \arg9.8h, \arg8.8h
uaddl2 \arg8\().8h, \arg0\().16b, \arg1\().16b
shl \arg8\().8h, \arg8\().8h, #1
add \arg8\().8h, \arg9\().8h, \arg8\().8h
rshrn2 \arg0.16b, \arg9.8h, #2
rshrn2 \arg7.16b, \arg8.8h, #3
shl \arg9.8h, \arg9.8h, #1
usubl2 \arg8.8h, \arg5.16b, \arg1.16b
add \arg9.8h, \arg8.8h, \arg9.8h
rshrn2 \arg0\().16b, \arg9\().8h, #2
rshrn2 \arg7\().16b, \arg8\().8h, #3
shl \arg9\().8h, \arg9\().8h, #1
usubl2 \arg8\().8h, \arg5\().16b, \arg1\().16b
add \arg9\().8h, \arg8\().8h, \arg9\().8h
uaddl2 \arg8.8h, \arg2.16b, \arg5.16b
uaddw2 \arg8.8h, \arg8.8h, \arg2.16b
uaddw2 \arg8.8h, \arg8.8h, \arg3.16b
uaddl2 \arg8\().8h, \arg2\().16b, \arg5\().16b
uaddw2 \arg8\().8h, \arg8\().8h, \arg2\().16b
uaddw2 \arg8\().8h, \arg8\().8h, \arg3\().16b
rshrn2 \arg9.16b, \arg9.8h, #3
rshrn2 \arg8.16b, \arg8.8h, #2
bsl \arg6.16b, \arg9.16b, \arg8.16b
rshrn2 \arg9\().16b, \arg9\().8h, #3
rshrn2 \arg8\().16b, \arg8\().8h, #2
bsl \arg6\().16b, \arg9\().16b, \arg8\().16b
.endm
.macro DIFF_CHROMA_EQ4_P0Q0_1 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7
uaddl \arg4.8h, \arg0.8b, \arg3.8b
shl \arg4.8h, \arg4.8h, #1
usubl \arg5.8h, \arg1.8b, \arg3.8b
add \arg5.8h, \arg5.8h, \arg4.8h
rshrn \arg6.8b, \arg5.8h, #2
usubl \arg5.8h, \arg2.8b, \arg0.8b
add \arg5.8h, \arg5.8h, \arg4.8h
rshrn \arg7.8b, \arg5.8h, #2
uaddl \arg4\().8h, \arg0\().8b, \arg3\().8b
shl \arg4\().8h, \arg4\().8h, #1
usubl \arg5\().8h, \arg1\().8b, \arg3\().8b
add \arg5\().8h, \arg5\().8h, \arg4\().8h
rshrn \arg6\().8b, \arg5\().8h, #2
usubl \arg5\().8h, \arg2\().8b, \arg0\().8b
add \arg5\().8h, \arg5\().8h, \arg4\().8h
rshrn \arg7\().8b, \arg5\().8h, #2
.endm
.macro DIFF_CHROMA_EQ4_P0Q0_2 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7
uaddl2 \arg4.8h, \arg0.16b, \arg3.16b
shl \arg4.8h, \arg4.8h, #1
usubl2 \arg5.8h, \arg1.16b, \arg3.16b
add \arg5.8h, \arg5.8h, \arg4.8h
rshrn2 \arg6.16b, \arg5.8h, #2
usubl2 \arg5.8h, \arg2.16b, \arg0.16b
add \arg5.8h, \arg5.8h, \arg4.8h
rshrn2 \arg7.16b, \arg5.8h, #2
uaddl2 \arg4\().8h, \arg0\().16b, \arg3\().16b
shl \arg4\().8h, \arg4\().8h, #1
usubl2 \arg5\().8h, \arg1\().16b, \arg3\().16b
add \arg5\().8h, \arg5\().8h, \arg4\().8h
rshrn2 \arg6\().16b, \arg5\().8h, #2
usubl2 \arg5\().8h, \arg2\().16b, \arg0\().16b
add \arg5\().8h, \arg5\().8h, \arg4\().8h
rshrn2 \arg7\().16b, \arg5\().8h, #2
.endm
.macro DIFF_LUMA_EQ4_MASK arg0, arg1, arg2, arg3
mov.16b \arg3, \arg2
bsl \arg3.16b, \arg0.16b, \arg1.16b
bsl \arg3\().16b, \arg0\().16b, \arg1\().16b
.endm
.macro LOAD_LUMA_DATA_3 arg0, arg1, arg2, arg3, arg4, arg5, arg6
ld3 {\arg0.b, \arg1.b, \arg2.b} [\arg6], [x2], x1
ld3 {\arg3.b, \arg4.b, \arg5.b} [\arg6], [x0], x1
ld3 {\arg0\().b, \arg1\().b, \arg2\().b} [\arg6], [x2], x1
ld3 {\arg3\().b, \arg4\().b, \arg5\().b} [\arg6], [x0], x1
.endm
.macro LOAD_LUMA_DATA_4 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8
ld4 {\arg0.b, \arg1.b, \arg2.b, \arg3.b} [\arg8], [x3], x1
ld4 {\arg4.b, \arg5.b, \arg6.b, \arg7.b} [\arg8], [x0], x1
ld4 {\arg0\().b, \arg1\().b, \arg2\().b, \arg3\().b} [\arg8], [x3], x1
ld4 {\arg4\().b, \arg5\().b, \arg6\().b, \arg7\().b} [\arg8], [x0], x1
.endm
.macro STORE_LUMA_DATA_4 arg0, arg1, arg2, arg3, arg4, arg5
st4 {\arg0.b, \arg1.b, \arg2.b, \arg3.b} [\arg4], [x0], x1
st4 {\arg0.b, \arg1.b, \arg2.b, \arg3.b} [\arg5], [x2], x1
st4 {\arg0\().b, \arg1\().b, \arg2\().b, \arg3\().b} [\arg4], [x0], x1
st4 {\arg0\().b, \arg1\().b, \arg2\().b, \arg3\().b} [\arg5], [x2], x1
.endm
.macro STORE_LUMA_DATA_3 arg0, arg1, arg2, arg3, arg4, arg5, arg6
st3 {\arg0.b, \arg1.b, \arg2.b} [\arg6], [x3], x1
st3 {\arg3.b, \arg4.b, \arg5.b} [\arg6], [x0], x1
st3 {\arg0\().b, \arg1\().b, \arg2\().b} [\arg6], [x3], x1
st3 {\arg3\().b, \arg4\().b, \arg5\().b} [\arg6], [x0], x1
.endm
.macro LOAD_CHROMA_DATA_4 arg0, arg1, arg2, arg3, arg4, arg5
ld4 {\arg0.b, \arg1.b, \arg2.b, \arg3.b} [\arg5], [\arg4], x2
ld4 {\arg0\().b, \arg1\().b, \arg2\().b, \arg3\().b} [\arg5], [\arg4], x2
.endm
.macro STORE_CHROMA_DATA_2 arg0, arg1, arg2, arg3
st2 {\arg0.b, \arg1.b} [\arg3], [\arg2], x2
st2 {\arg0\().b, \arg1\().b} [\arg3], [\arg2], x2
.endm
.macro ZERO_JUMP_END arg0, arg1, arg2, arg3
mov \arg1, \arg0.d[0]
mov \arg2, \arg0.d[1]
mov \arg1, \arg0\().d[0]
mov \arg2, \arg0\().d[1]
orr \arg1, \arg1, \arg2
cbz \arg1, \arg3
.endm
@ -471,7 +471,7 @@ bs_mv_check_jump1:
bs_nzc_check_jump0:
ext.16b v1, v1, v0, #12
add \arg3.16b, v0.16b, v1.16b
add \arg3\().16b, v0.16b, v1.16b
// Arrange the input data --- LEFT
ands x6, \arg1, #1
@ -492,28 +492,28 @@ bs_nzc_check_jump1:
ins v2.d[0], v0.d[1]
zip1 v0.16b, v0.16b, v2.16b
ext.16b v1, v1, v0, #12
add \arg4.16b, v0.16b, v1.16b
add \arg4\().16b, v0.16b, v1.16b
.endm
.macro BS_COMPARE_MV arg0, arg1, arg2, arg3, arg4, arg5
//in: \arg0,\arg1(const),\arg2(const),\arg3(const),\arg4(const); out:\arg5
mov w6, #4
sabd v20.8h, \arg0.8h, \arg1.8h
sabd v21.8h, \arg1.8h, \arg2.8h
dup \arg0.8h, w6
sabd v22.8h, \arg2.8h, \arg3.8h
sabd v23.8h, \arg3.8h, \arg4.8h
sabd v20.8h, \arg0\().8h, \arg1\().8h
sabd v21.8h, \arg1\().8h, \arg2\().8h
dup \arg0\().8h, w6
sabd v22.8h, \arg2\().8h, \arg3\().8h
sabd v23.8h, \arg3\().8h, \arg4\().8h
cmge v20.8h, v20.8h, \arg0.8h
cmge v21.8h, v21.8h, \arg0.8h
cmge v22.8h, v22.8h, \arg0.8h
cmge v23.8h, v23.8h, \arg0.8h
cmge v20.8h, v20.8h, \arg0\().8h
cmge v21.8h, v21.8h, \arg0\().8h
cmge v22.8h, v22.8h, \arg0\().8h
cmge v23.8h, v23.8h, \arg0\().8h
addp v20.8h, v20.8h, v21.8h
addp v21.8h, v22.8h, v23.8h
addhn \arg5.8b, v20.8h, v20.8h
addhn2 \arg5.16b, v21.8h, v21.8h
addhn \arg5\().8b, v20.8h, v20.8h
addhn2 \arg5\().16b, v21.8h, v21.8h
.endm
.macro BS_MV_CHECK arg0, arg1, arg2, arg3, arg4, arg5, arg6
@ -540,14 +540,14 @@ bs_mv_check_jump0:
ld1 {v4.s} [2], [x6]
ld1 {v4.s} [3], [x7]
bs_mv_check_jump1:
zip1 \arg5.4s, v0.4s, v2.4s
zip2 \arg6.4s, v0.4s, v2.4s
zip1 \arg5\().4s, v0.4s, v2.4s
zip2 \arg6\().4s, v0.4s, v2.4s
zip1 v0.4s, v1.4s, v3.4s
zip2 v2.4s, v1.4s, v3.4s
zip2 v1.4s, \arg5.4s, v0.4s
zip1 v0.4s, \arg5.4s, v0.4s
zip2 v3.4s, \arg6.4s, v2.4s
zip1 v2.4s, \arg6.4s, v2.4s
zip2 v1.4s, \arg5\().4s, v0.4s
zip1 v0.4s, \arg5\().4s, v0.4s
zip2 v3.4s, \arg6\().4s, v2.4s
zip1 v2.4s, \arg6\().4s, v2.4s
BS_COMPARE_MV v4, v0, v1, v2, v3, \arg4
.endm
#endif