RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

arm: vp9itxfm: Simplify txfm string comparisons 

Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
Martin Storsjö 2016-11-12 21:25:50 +02:00
parent 3c9546dfaf
commit 52d196fb30
1 changed files with 14 additions and 39 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

53
libavcodec/arm/vp9itxfm_neon.S
View File

@ -258,8 +258,7 @@ function ff_vp9_\txfm1\()_\txfm2\()_4x4_add_neon, export=1
.endif .endif
vmov.i16 q15, #0 vmov.i16 q15, #0
.ifc \txfm1,idct .ifc \txfm1\()_\txfm2,idct_idct
.ifc \txfm2,idct
cmp r3, #1 cmp r3, #1
bne 1f bne 1f
@ DC-only for idct/idct @ DC-only for idct/idct
@ -273,7 +272,6 @@ function ff_vp9_\txfm1\()_\txfm2\()_4x4_add_neon, export=1
vmov q3, q2 vmov q3, q2
b 2f b 2f
.endif .endif
.endif
1: 1:
vld1.16 {d4-d7}, [r2,:128] vld1.16 {d4-d7}, [r2,:128]
@ -386,29 +384,21 @@ function ff_vp9_\txfm1\()_\txfm2\()_8x8_add_neon, export=1
@ if only idct is involved. @ if only idct is involved.
@ The iadst also uses a few coefficients from @ The iadst also uses a few coefficients from
@ idct, so those always need to be loaded. @ idct, so those always need to be loaded.
.ifc \txfm1,iadst .ifc \txfm1\()_\txfm2,idct_idct
movrel r12, iadst8_coeffs
vld1.16 {q1}, [r12,:128]!
vpush {q4-q7}
vld1.16 {q0}, [r12,:128]
.else
.ifc \txfm2,iadst
movrel r12, iadst8_coeffs
vld1.16 {q1}, [r12,:128]!
vpush {q4-q7}
vld1.16 {q0}, [r12,:128]
.else
movrel r12, idct_coeffs movrel r12, idct_coeffs
vpush {q4-q5} vpush {q4-q5}
vld1.16 {q0}, [r12,:128] vld1.16 {q0}, [r12,:128]
.endif .else
movrel r12, iadst8_coeffs
vld1.16 {q1}, [r12,:128]!
vpush {q4-q7}
vld1.16 {q0}, [r12,:128]
.endif .endif
vmov.i16 q2, #0 vmov.i16 q2, #0
vmov.i16 q3, #0 vmov.i16 q3, #0
.ifc \txfm1,idct .ifc \txfm1\()_\txfm2,idct_idct
.ifc \txfm2,idct
cmp r3, #1 cmp r3, #1
bne 1f bne 1f
@ DC-only for idct/idct @ DC-only for idct/idct
@ -428,7 +418,6 @@ function ff_vp9_\txfm1\()_\txfm2\()_8x8_add_neon, export=1
vst1.16 {d4[0]}, [r2,:16] vst1.16 {d4[0]}, [r2,:16]
b 2f b 2f
.endif .endif
.endif
1: 1:
vld1.16 {q8-q9}, [r2,:128]! vld1.16 {q8-q9}, [r2,:128]!
vld1.16 {q10-q11}, [r2,:128]! vld1.16 {q10-q11}, [r2,:128]!
@ -497,14 +486,10 @@ function ff_vp9_\txfm1\()_\txfm2\()_8x8_add_neon, export=1
vst1.8 {d10}, [r3,:64], r1 vst1.8 {d10}, [r3,:64], r1
vst1.8 {d11}, [r3,:64], r1 vst1.8 {d11}, [r3,:64], r1
.ifc \txfm1,iadst .ifc \txfm1\()_\txfm2,idct_idct
vpop {q4-q7}
.else
.ifc \txfm2,iadst
vpop {q4-q7}
.else
vpop {q4-q5} vpop {q4-q5}
.endif .else
vpop {q4-q7}
.endif .endif
bx lr bx lr
endfunc endfunc
@ -798,19 +783,13 @@ itxfm16_1d_funcs iadst
.macro itxfm_func16x16 txfm1, txfm2 .macro itxfm_func16x16 txfm1, txfm2
function ff_vp9_\txfm1\()_\txfm2\()_16x16_add_neon, export=1 function ff_vp9_\txfm1\()_\txfm2\()_16x16_add_neon, export=1
.ifc \txfm1,idct .ifc \txfm1\()_\txfm2,idct_idct
.ifc \txfm2,idct
cmp r3, #1 cmp r3, #1
beq idct16x16_dc_add_neon beq idct16x16_dc_add_neon
.endif
.endif .endif
push {r4-r7,lr} push {r4-r7,lr}
.ifc \txfm1,iadst .ifnc \txfm1\()_\txfm2,idct_idct
vpush {q4-q7} vpush {q4-q7}
.else
.ifc \txfm2,iadst
vpush {q4-q7}
.endif
.endif .endif
mov r7, sp mov r7, sp
@ -850,12 +829,8 @@ A sub sp, sp, #512
.endr .endr
mov sp, r7 mov sp, r7
.ifc \txfm1,iadst .ifnc \txfm1\()_\txfm2,idct_idct
vpop {q4-q7} vpop {q4-q7}
.else
.ifc \txfm2,iadst
vpop {q4-q7}
.endif
.endif .endif
pop {r4-r7,pc} pop {r4-r7,pc}
endfunc endfunc