Skip to content
GitLab
Commit 769f2424 authored by Fritz Koenig's avatar Fritz Koenig
Browse files

Removed unnecessary pxor. 

There is no need to make sure that the lower byte of the
register is 0 because the downshift by 11 overwrites that byte.

Change-Id: I89cbf004b2ff532a2c68e0dc399c45a49cdad5a1
parent 71a1c197
v1.14.0-linphone 1.4.X aylesbury bali cayuga eider experimental feature/update_to_v1.9.0-linphone feature/uwp_nuget forest frame_parallel highbitdepth indianrunnerduck javanwhistlingduck khakicampbell linphone linphone-android linphone-old longtailedduck m29-baseline m31-baseline m49-2623 m52-2743 m54-2840 m56-2924 m66-3359 m68-3440 mandarinduck mcw mcw2 nextgen nextgenv2 pcs-2013 playground sandbox/Jingning/experimental sandbox/Jingning/transcode sandbox/Jingning/vpx sandbox/aconverse@google.com/ansbench sandbox/atna/dec_sem_sync sandbox/awatry/initial_opencl_implementation sandbox/debargha/playground sandbox/hkuang/frame_parallel sandbox/hkuang@google.com/decode sandbox/hlundin/error-concealment sandbox/holmer/error-concealment sandbox/jimbankoski@google.com/proposed-aom sandbox/jingning@google.com/decoder_test_suite sandbox/jingning@google.com/experimental sandbox/jkoleszar/cached-multibit sandbox/jkoleszar/experimental-knobs sandbox/jkoleszar/new-rate-control sandbox/jkoleszar/new-rtcd sandbox/jkoleszar/reuse-modemv sandbox/jkoleszar/use-memcpy sandbox/jzern@google.com/test sandbox/slavarnway/test sandbox/wangch@google.com/vp9 sandbox/yaowu@google.com/mergeaom stable-vp9-decoder v1.12.0-linphone v1.6.1_linphone v1.7.0-linphone v1.9.0-linphone vp9-preview v1.9.0 v1.9.0-rc1 v1.8.2 v1.8.1 v1.8.0 v1.7.0 v1.6.1 v1.6.0 v1.5.0 v1.4.0 v1.3.0 v1.2.0 v1.1.0 v1.0.0 v0.9.7 v0.9.7-p1 v0.9.6 v0.9.5
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 42 additions and 62 deletions
vp8/common/x86/loopfilter_sse2.asm
+ 42
− 62
View file @ 769f2424
......@@ -196,12 +196,12 @@
pxor xmm7, [t80 GLOBAL] ; q1 offset to convert to signed values
psubsb xmm2, xmm7 ; p1 - q1
pand xmm2, xmm4 ; high var mask (hvm)(p1 - q1)
pxor xmm6, [t80 GLOBAL] ; offset to convert to signed values
pand xmm2, xmm4 ; high var mask (hvm)(p1 - q1)
pxor xmm0, [t80 GLOBAL] ; offset to convert to signed values
movdqa xmm3, xmm0 ; q0
movdqa xmm3, xmm0 ; q0
psubsb xmm0, xmm6 ; q0 - p0
paddsb xmm2, xmm0 ; 1 * (q0 - p0) + hvm(p1 - q1)
paddsb xmm2, xmm0 ; 2 * (q0 - p0) + hvm(p1 - q1)
......@@ -211,29 +211,28 @@
paddsb xmm1, [t4 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb xmm2, [t3 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
pxor xmm0, xmm0
pxor xmm5, xmm5
punpcklbw xmm0, xmm2
punpckhbw xmm5, xmm2
psraw xmm0, 11
psraw xmm5, 11
packsswb xmm0, xmm5
movdqa xmm2, xmm0 ; (3* (q0 - p0) + hvm(p1 - q1) + 3) >> 3;
pxor xmm0, xmm0 ; 0
movdqa xmm5, xmm1 ; abcdefgh
punpcklbw xmm0, xmm1 ; e0f0g0h0
punpckhbw xmm5, xmm2 ; axbxcxdx
punpcklbw xmm2, xmm2 ; exfxgxhx
psraw xmm5, 11 ; sign extended shift right by 3
psraw xmm2, 11 ; sign extended shift right by 3
packsswb xmm2, xmm5 ; (3* (q0 - p0) + hvm(p1 - q1) + 3) >> 3;
punpcklbw xmm0, xmm1 ; exfxgxhx
punpckhbw xmm1, xmm1 ; axbxcxdx
psraw xmm0, 11 ; sign extended shift right by 3
pxor xmm1, xmm1 ; 0
punpckhbw xmm1, xmm5 ; a0b0c0d0
psraw xmm1, 11 ; sign extended shift right by 3
movdqa xmm5, xmm0 ; save results
movdqa xmm5, xmm0 ; save results
packsswb xmm0, xmm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>3
paddsw xmm5, [ones GLOBAL]
paddsw xmm1, [ones GLOBAL]
psraw xmm5, 1 ; partial shifted one more time for 2nd tap
psraw xmm1, 1 ; partial shifted one more time for 2nd tap
packsswb xmm5, xmm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>4
pandn xmm4, xmm5 ; high edge variance additive
%endmacro
......@@ -433,29 +432,27 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
pand xmm2, xmm4; ; Filter2 = vp8_filter & hev
movdqa xmm5, xmm2
paddsb xmm5, [t3 GLOBAL]
paddsb xmm5, [t3 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 3)
punpckhbw xmm7, xmm5 ; axbxcxdx
punpcklbw xmm5, xmm5 ; exfxgxhx
pxor xmm0, xmm0 ; 0
pxor xmm7, xmm7 ; 0
punpcklbw xmm0, xmm5 ; e0f0g0h0
psraw xmm0, 11 ; sign extended shift right by 3
punpckhbw xmm7, xmm5 ; a0b0c0d0
psraw xmm7, 11 ; sign extended shift right by 3
packsswb xmm0, xmm7 ; Filter2 >>=3;
movdqa xmm5, xmm0 ; Filter2
paddsb xmm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
psraw xmm5, 11 ; sign extended shift right by 3
packsswb xmm5, xmm7 ; Filter2 >>=3;
paddsb xmm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
punpckhbw xmm7, xmm2 ; axbxcxdx
punpcklbw xmm0, xmm2 ; exfxgxhx
pxor xmm0, xmm0 ; 0
pxor xmm7, xmm7 ; 0
punpcklbw xmm0, xmm2 ; e0f0g0h0
psraw xmm0, 11 ; sign extended shift right by 3
punpckhbw xmm7, xmm2 ; a0b0c0d0
psraw xmm7, 11 ; sign extended shift right by 3
packsswb xmm0, xmm7 ; Filter2 >>=3;
psraw xmm0, 11 ; sign extended shift right by 3
psubsb xmm3, xmm0 ; qs0 =qs0 - filter1
packsswb xmm0, xmm7 ; Filter2 >>=3;
paddsb xmm6, xmm5 ; ps0 =ps0 + Fitler2
psubsb xmm3, xmm0 ; qs0 =qs0 - filter1
pandn xmm4, xmm1 ; vp8_filter&=~hev
%endmacro
......@@ -465,7 +462,6 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
; *oq0 = s^0x80;
; s = vp8_signed_char_clamp(ps0 + u);
; *op0 = s^0x80;
pxor xmm0, xmm0
pxor xmm1, xmm1
pxor xmm2, xmm2
......@@ -1022,28 +1018,19 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
paddsb xmm1, [t4 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb xmm2, [t3 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
pxor xmm0, xmm0
pxor xmm5, xmm5
punpcklbw xmm0, xmm2
punpckhbw xmm5, xmm2
psraw xmm0, 11
punpcklbw xmm2, xmm2
psraw xmm5, 11
packsswb xmm0, xmm5
psraw xmm2, 11
movdqa xmm2, xmm0 ; (3* (q0 - p0) + hvm(p1 - q1) + 3) >> 3;
packsswb xmm2, xmm5 ; (3* (q0 - p0) + hvm(p1 - q1) + 3) >> 3;
punpcklbw xmm0, xmm1 ; exfxgxhx
pxor xmm0, xmm0 ; 0
movdqa xmm5, xmm1 ; abcdefgh
punpcklbw xmm0, xmm1 ; e0f0g0h0
punpckhbw xmm1, xmm1 ; axbxcxdx
psraw xmm0, 11 ; sign extended shift right by 3
pxor xmm1, xmm1 ; 0
punpckhbw xmm1, xmm5 ; a0b0c0d0
psraw xmm1, 11 ; sign extended shift right by 3
movdqa xmm5, xmm0 ; save results
......@@ -1308,28 +1295,22 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
movdqa xmm5, xmm2
paddsb xmm5, [t3 GLOBAL]
pxor xmm0, xmm0 ; 0
pxor xmm7, xmm7 ; 0
punpcklbw xmm0, xmm5 ; e0f0g0h0
psraw xmm0, 11 ; sign extended shift right by 3
punpckhbw xmm7, xmm5 ; axbxcxdx
punpcklbw xmm5, xmm5 ; exfxgxhx
punpckhbw xmm7, xmm5 ; a0b0c0d0
psraw xmm7, 11 ; sign extended shift right by 3
psraw xmm5, 11 ; sign extended shift right by 3
packsswb xmm0, xmm7 ; Filter2 >>=3;
movdqa xmm5, xmm0 ; Filter2
packsswb xmm5, xmm7 ; Filter2 >>=3;
paddsb xmm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
pxor xmm0, xmm0 ; 0
pxor xmm7, xmm7 ; 0
punpcklbw xmm0, xmm2 ; e0f0g0h0
punpcklbw xmm0, xmm2 ; exfxgxhx
punpckhbw xmm7, xmm2 ; axbxcxdx
psraw xmm0, 11 ; sign extended shift right by 3
punpckhbw xmm7, xmm2 ; a0b0c0d0
psraw xmm7, 11 ; sign extended shift right by 3
packsswb xmm0, xmm7 ; Filter2 >>=3;
psubsb xmm3, xmm0 ; qs0 =qs0 - filter1
......@@ -1344,7 +1325,6 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
; *oq0 = s^0x80;
; s = vp8_signed_char_clamp(ps0 + u);
; *op0 = s^0x80;
pxor xmm0, xmm0
pxor xmm1, xmm1
pxor xmm2, xmm2
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment