[FFmpeg-devel] [PATCH 18/21] aarch64: hevc: Produce plain neon versions of qpel_hv
Martin Storsjö
martin at martin.st
Mon Mar 25 17:02:40 EET 2024
As the plain neon qpel_h functions process two rows at a time,
we need to allocate storage for h+8 rows instead of h+7.
By allocating storage for h+8 rows, incrementing the stack
pointer won't end up at the right spot in the end. Store the
intended final stack pointer value in a register x14 which we
store on the stack.
AWS Graviton 3:
put_hevc_qpel_hv4_8_c: 386.0
put_hevc_qpel_hv4_8_neon: 125.7
put_hevc_qpel_hv4_8_i8mm: 83.2
put_hevc_qpel_hv6_8_c: 749.0
put_hevc_qpel_hv6_8_neon: 207.0
put_hevc_qpel_hv6_8_i8mm: 166.0
put_hevc_qpel_hv8_8_c: 1305.2
put_hevc_qpel_hv8_8_neon: 216.5
put_hevc_qpel_hv8_8_i8mm: 213.0
put_hevc_qpel_hv12_8_c: 2570.5
put_hevc_qpel_hv12_8_neon: 480.0
put_hevc_qpel_hv12_8_i8mm: 398.2
put_hevc_qpel_hv16_8_c: 4158.7
put_hevc_qpel_hv16_8_neon: 659.7
put_hevc_qpel_hv16_8_i8mm: 593.5
put_hevc_qpel_hv24_8_c: 8626.7
put_hevc_qpel_hv24_8_neon: 1653.5
put_hevc_qpel_hv24_8_i8mm: 1398.7
put_hevc_qpel_hv32_8_c: 14646.0
put_hevc_qpel_hv32_8_neon: 2566.2
put_hevc_qpel_hv32_8_i8mm: 2287.5
put_hevc_qpel_hv48_8_c: 31072.5
put_hevc_qpel_hv48_8_neon: 6228.5
put_hevc_qpel_hv48_8_i8mm: 5291.0
put_hevc_qpel_hv64_8_c: 53847.2
put_hevc_qpel_hv64_8_neon: 9856.7
put_hevc_qpel_hv64_8_i8mm: 8831.0
---
libavcodec/aarch64/hevcdsp_init_aarch64.c | 6 +
libavcodec/aarch64/hevcdsp_qpel_neon.S | 166 +++++++++++++---------
2 files changed, 104 insertions(+), 68 deletions(-)
diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c b/libavcodec/aarch64/hevcdsp_init_aarch64.c
index ea0d26c019..105c26017b 100644
--- a/libavcodec/aarch64/hevcdsp_init_aarch64.c
+++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c
@@ -265,6 +265,10 @@ NEON8_FNPROTO(qpel_v, (int16_t *dst,
const uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width),);
+NEON8_FNPROTO(qpel_hv, (int16_t *dst,
+ const uint8_t *src, ptrdiff_t srcstride,
+ int height, intptr_t mx, intptr_t my, int width),);
+
NEON8_FNPROTO(qpel_hv, (int16_t *dst,
const uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width), _i8mm);
@@ -436,6 +440,8 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
NEON8_FNASSIGN_SHARED_32(c->put_hevc_qpel_uni_w, 0, 1, qpel_uni_w_h,);
+ NEON8_FNASSIGN(c->put_hevc_qpel, 1, 1, qpel_hv,);
+
if (have_i8mm(cpu_flags)) {
NEON8_FNASSIGN(c->put_hevc_epel, 0, 1, epel_h, _i8mm);
NEON8_FNASSIGN(c->put_hevc_epel, 1, 1, epel_hv, _i8mm);
diff --git a/libavcodec/aarch64/hevcdsp_qpel_neon.S b/libavcodec/aarch64/hevcdsp_qpel_neon.S
index ad568e415b..7bffb991a7 100644
--- a/libavcodec/aarch64/hevcdsp_qpel_neon.S
+++ b/libavcodec/aarch64/hevcdsp_qpel_neon.S
@@ -3804,7 +3804,8 @@ function hevc_put_hevc_qpel_hv4_8_end_neon
.endm
1: calc_all
.purgem calc
-2: ret
+2: mov sp, x14
+ ret
endfunc
function hevc_put_hevc_qpel_hv6_8_end_neon
@@ -3831,7 +3832,8 @@ function hevc_put_hevc_qpel_hv6_8_end_neon
.endm
1: calc_all
.purgem calc
-2: ret
+2: mov sp, x14
+ ret
endfunc
function hevc_put_hevc_qpel_hv8_8_end_neon
@@ -3857,7 +3859,8 @@ function hevc_put_hevc_qpel_hv8_8_end_neon
.endm
1: calc_all
.purgem calc
-2: ret
+2: mov sp, x14
+ ret
endfunc
function hevc_put_hevc_qpel_hv12_8_end_neon
@@ -3882,7 +3885,8 @@ function hevc_put_hevc_qpel_hv12_8_end_neon
.endm
1: calc_all2
.purgem calc
-2: ret
+2: mov sp, x14
+ ret
endfunc
function hevc_put_hevc_qpel_hv16_8_end_neon
@@ -3906,7 +3910,8 @@ function hevc_put_hevc_qpel_hv16_8_end_neon
.endm
1: calc_all2
.purgem calc
-2: ret
+2: mov sp, x14
+ ret
endfunc
function hevc_put_hevc_qpel_hv32_8_end_neon
@@ -3937,162 +3942,187 @@ function hevc_put_hevc_qpel_hv32_8_end_neon
add sp, sp, #32
subs w6, w6, #16
b.hi 0b
- add w10, w3, #6
- add sp, sp, #64 // discard rest of first line
- lsl x10, x10, #7
- add sp, sp, x10 // tmp_array without first line
+ mov sp, x14
ret
endfunc
-#if HAVE_I8MM
-ENABLE_I8MM
-function ff_hevc_put_hevc_qpel_hv4_8_neon_i8mm, export=1
- add w10, w3, #7
+.macro qpel_hv suffix
+function ff_hevc_put_hevc_qpel_hv4_8_\suffix, export=1
+ add w10, w3, #8
mov x7, #128
lsl x10, x10, #7
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
- add x0, sp, #32
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
+ add x0, sp, #48
sub x1, x1, x2, lsl #1
add x3, x3, #7
sub x1, x1, x2
- bl X(ff_hevc_put_hevc_qpel_h4_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ bl X(ff_hevc_put_hevc_qpel_h4_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv4_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv6_8_neon_i8mm, export=1
- add w10, w3, #7
+function ff_hevc_put_hevc_qpel_hv6_8_\suffix, export=1
+ add w10, w3, #8
mov x7, #128
lsl x10, x10, #7
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
- add x0, sp, #32
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
+ add x0, sp, #48
sub x1, x1, x2, lsl #1
add x3, x3, #7
sub x1, x1, x2
- bl X(ff_hevc_put_hevc_qpel_h6_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ bl X(ff_hevc_put_hevc_qpel_h6_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv6_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv8_8_neon_i8mm, export=1
- add w10, w3, #7
+function ff_hevc_put_hevc_qpel_hv8_8_\suffix, export=1
+ add w10, w3, #8
lsl x10, x10, #7
sub x1, x1, x2, lsl #1
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
- add x0, sp, #32
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
+ add x0, sp, #48
add x3, x3, #7
sub x1, x1, x2
- bl X(ff_hevc_put_hevc_qpel_h8_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ bl X(ff_hevc_put_hevc_qpel_h8_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv8_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv12_8_neon_i8mm, export=1
- add w10, w3, #7
+function ff_hevc_put_hevc_qpel_hv12_8_\suffix, export=1
+ add w10, w3, #8
lsl x10, x10, #7
sub x1, x1, x2, lsl #1
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
- add x0, sp, #32
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
+ add x0, sp, #48
add x3, x3, #7
sub x1, x1, x2
- bl X(ff_hevc_put_hevc_qpel_h12_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ mov w6, #12
+ bl X(ff_hevc_put_hevc_qpel_h12_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv12_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv16_8_neon_i8mm, export=1
- add w10, w3, #7
+function ff_hevc_put_hevc_qpel_hv16_8_\suffix, export=1
+ add w10, w3, #8
lsl x10, x10, #7
sub x1, x1, x2, lsl #1
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
add x3, x3, #7
- add x0, sp, #32
+ add x0, sp, #48
sub x1, x1, x2
- bl X(ff_hevc_put_hevc_qpel_h16_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ bl X(ff_hevc_put_hevc_qpel_h16_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv16_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv24_8_neon_i8mm, export=1
+function ff_hevc_put_hevc_qpel_hv24_8_\suffix, export=1
stp x4, x5, [sp, #-64]!
stp x2, x3, [sp, #16]
stp x0, x1, [sp, #32]
str x30, [sp, #48]
- bl X(ff_hevc_put_hevc_qpel_hv12_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv12_8_\suffix)
ldp x0, x1, [sp, #32]
ldp x2, x3, [sp, #16]
ldp x4, x5, [sp], #48
add x1, x1, #12
add x0, x0, #24
- bl X(ff_hevc_put_hevc_qpel_hv12_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv12_8_\suffix)
ldr x30, [sp], #16
ret
endfunc
-function ff_hevc_put_hevc_qpel_hv32_8_neon_i8mm, export=1
- add w10, w3, #7
+function ff_hevc_put_hevc_qpel_hv32_8_\suffix, export=1
+ add w10, w3, #8
sub x1, x1, x2, lsl #1
lsl x10, x10, #7
sub x1, x1, x2
+ mov x14, sp
sub sp, sp, x10 // tmp_array
- stp x5, x30, [sp, #-32]!
- stp x0, x3, [sp, #16]
+ stp x5, x30, [sp, #-48]!
+ stp x0, x3, [sp, #16]
+ str x14, [sp, #32]
add x3, x3, #7
- add x0, sp, #32
- bl X(ff_hevc_put_hevc_qpel_h32_8_neon_i8mm)
- ldp x0, x3, [sp, #16]
- ldp x5, x30, [sp], #32
+ add x0, sp, #48
+ mov w6, #32
+ bl X(ff_hevc_put_hevc_qpel_h32_8_\suffix)
+ ldr x14, [sp, #32]
+ ldp x0, x3, [sp, #16]
+ ldp x5, x30, [sp], #48
b hevc_put_hevc_qpel_hv32_8_end_neon
endfunc
-function ff_hevc_put_hevc_qpel_hv48_8_neon_i8mm, export=1
+function ff_hevc_put_hevc_qpel_hv48_8_\suffix, export=1
stp x4, x5, [sp, #-64]!
stp x2, x3, [sp, #16]
stp x0, x1, [sp, #32]
str x30, [sp, #48]
- bl X(ff_hevc_put_hevc_qpel_hv24_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv24_8_\suffix)
ldp x0, x1, [sp, #32]
ldp x2, x3, [sp, #16]
ldp x4, x5, [sp], #48
add x1, x1, #24
add x0, x0, #48
- bl X(ff_hevc_put_hevc_qpel_hv24_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv24_8_\suffix)
ldr x30, [sp], #16
ret
endfunc
-function ff_hevc_put_hevc_qpel_hv64_8_neon_i8mm, export=1
+function ff_hevc_put_hevc_qpel_hv64_8_\suffix, export=1
stp x4, x5, [sp, #-64]!
stp x2, x3, [sp, #16]
stp x0, x1, [sp, #32]
str x30, [sp, #48]
mov x6, #32
- bl X(ff_hevc_put_hevc_qpel_hv32_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv32_8_\suffix)
ldp x0, x1, [sp, #32]
ldp x2, x3, [sp, #16]
ldp x4, x5, [sp], #48
add x1, x1, #32
add x0, x0, #64
mov x6, #32
- bl X(ff_hevc_put_hevc_qpel_hv32_8_neon_i8mm)
+ bl X(ff_hevc_put_hevc_qpel_hv32_8_\suffix)
ldr x30, [sp], #16
ret
endfunc
+.endm
+
+qpel_hv neon
+
+#if HAVE_I8MM
+ENABLE_I8MM
+
+qpel_hv neon_i8mm
+
DISABLE_I8MM
#endif
--
2.39.3 (Apple Git-146)
More information about the ffmpeg-devel
mailing list