* [FFmpeg-devel] [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
@ 2024-03-02 7:42 flow gg
2024-03-03 2:01 ` flow gg
0 siblings, 1 reply; 3+ messages in thread
From: flow gg @ 2024-03-02 7:42 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 1 bytes --]
[-- Attachment #2: 0004-lavc-vp9dsp-R-V-V-ipred-tm.patch --]
[-- Type: text/x-patch, Size: 6536 bytes --]
From 3128765d298f5a44fd13be7b3da2ef88c96083f9 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Sat, 2 Mar 2024 09:35:22 +0800
Subject: [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
C908:
vp9_tm_4x4_8bpp_c: 116.5
vp9_tm_4x4_8bpp_rvv_i32: 43.5
vp9_tm_8x8_8bpp_c: 416.2
vp9_tm_8x8_8bpp_rvv_i32: 86.0
vp9_tm_16x16_8bpp_c: 1665.5
vp9_tm_16x16_8bpp_rvv_i32: 187.2
vp9_tm_32x32_8bpp_c: 6974.2
vp9_tm_32x32_8bpp_rvv_i32: 625.7
---
libavcodec/riscv/vp9_intra_rvv.S | 143 +++++++++++++++++++++++++++++++
libavcodec/riscv/vp9dsp.h | 8 ++
libavcodec/riscv/vp9dsp_init.c | 4 +
3 files changed, 155 insertions(+)
diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S
index 35fb0ebe10..a8265255c4 100644
--- a/libavcodec/riscv/vp9_intra_rvv.S
+++ b/libavcodec/riscv/vp9_intra_rvv.S
@@ -290,3 +290,146 @@ func ff_h_8x8_rvv, zve32x
ret
endfunc
+
+.macro tm_sum dst, top, offset
+ lbu t3, \offset(a2)
+ sub t3, t3, a4
+ vadd.vx \dst, \top, t3
+.endm
+
+func ff_tm_32x32_rvv, zve32x
+ lbu a4, -1(a3)
+ li t5, 32
+
+ .macro tm_sum32 n1,n2,n3,n4,n5,n6,n7,n8
+
+ vsetvli zero, t5, e16, m4, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+
+ tm_sum v0, v28, \n1
+ tm_sum v4, v28, \n2
+ tm_sum v8, v28, \n3
+ tm_sum v12, v28, \n4
+ tm_sum v16, v28, \n5
+ tm_sum v20, v28, \n6
+ tm_sum v24, v28, \n7
+ tm_sum v28, v28, \n8
+
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m2, ta, ma
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+
+ .endm
+
+ tm_sum32 31, 30, 29, 28, 27, 26, 25, 24
+ tm_sum32 23, 22, 21, 20, 19, 18, 17, 16
+ tm_sum32 15, 14, 13, 12, 11, 10, 9, 8
+ tm_sum32 7, 6, 5, 4, 3, 2, 1, 0
+
+ ret
+endfunc
+
+func ff_tm_16x16_rvv, zve32x
+ vsetivli zero, 16, e16, m2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v30, v8
+ lbu a4, -1(a3)
+
+ tm_sum v0, v30, 15
+ tm_sum v2, v30, 14
+ tm_sum v4, v30, 13
+ tm_sum v6, v30, 12
+ tm_sum v8, v30, 11
+ tm_sum v10, v30, 10
+ tm_sum v12, v30, 9
+ tm_sum v14, v30, 8
+ tm_sum v16, v30, 7
+ tm_sum v18, v30, 6
+ tm_sum v20, v30, 5
+ tm_sum v22, v30, 4
+ tm_sum v24, v30, 3
+ tm_sum v26, v30, 2
+ tm_sum v28, v30, 1
+ tm_sum v30, v30, 0
+
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m1, ta, ma
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v30, v30, 0
+ vse8.v v30, (a0)
+
+ ret
+endfunc
+
+func ff_tm_8x8_rvv, zve32x
+ vsetivli zero, 8, e16, m1, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 7
+ tm_sum v17, v28, 6
+ tm_sum v18, v28, 5
+ tm_sum v19, v28, 4
+ tm_sum v20, v28, 3
+ tm_sum v21, v28, 2
+ tm_sum v22, v28, 1
+ tm_sum v23, v28, 0
+
+ .irp n 16, 17, 18, 19, 20, 21, 22, 23
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf2, ta, ma
+ .irp n 16, 17, 18, 19, 20, 21, 22
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v23, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
+
+func ff_tm_4x4_rvv, zve32x
+ vsetivli zero, 4, e16, mf2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 3
+ tm_sum v17, v28, 2
+ tm_sum v18, v28, 1
+ tm_sum v19, v28, 0
+
+ .irp n 16, 17, 18, 19
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf4, ta, ma
+ .irp n 16, 17, 18
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v19, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h
index 2b2e0db0d8..cff2310e5c 100644
--- a/libavcodec/riscv/vp9dsp.h
+++ b/libavcodec/riscv/vp9dsp.h
@@ -72,5 +72,13 @@ void ff_h_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
void ff_h_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
+void ff_tm_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_4x4_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
#endif // #ifndef AVCODEC_RISCV_VP9DSP_RISCV_H
diff --git a/libavcodec/riscv/vp9dsp_init.c b/libavcodec/riscv/vp9dsp_init.c
index 16aeeb260a..f08c8f6a42 100644
--- a/libavcodec/riscv/vp9dsp_init.c
+++ b/libavcodec/riscv/vp9dsp_init.c
@@ -57,6 +57,10 @@ static av_cold void vp9dsp_intrapred_init_rvv(VP9DSPContext *dsp, int bpp)
dsp->intra_pred[TX_32X32][HOR_PRED] = ff_h_32x32_rvv;
dsp->intra_pred[TX_16X16][HOR_PRED] = ff_h_16x16_rvv;
dsp->intra_pred[TX_8X8][HOR_PRED] = ff_h_8x8_rvv;
+ dsp->intra_pred[TX_32X32][TM_VP8_PRED] = ff_tm_32x32_rvv;
+ dsp->intra_pred[TX_16X16][TM_VP8_PRED] = ff_tm_16x16_rvv;
+ dsp->intra_pred[TX_8X8][TM_VP8_PRED] = ff_tm_8x8_rvv;
+ dsp->intra_pred[TX_4X4][TM_VP8_PRED] = ff_tm_4x4_rvv;
}
#endif
}
--
2.44.0
[-- Attachment #3: Type: text/plain, Size: 251 bytes --]
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [FFmpeg-devel] [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
2024-03-02 7:42 [FFmpeg-devel] [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm flow gg
@ 2024-03-03 2:01 ` flow gg
2024-03-22 6:03 ` flow gg
0 siblings, 1 reply; 3+ messages in thread
From: flow gg @ 2024-03-03 2:01 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 135 bytes --]
Due to the PATCH 1/4 update, updates are made here.
flow gg <hlefthleft@gmail.com> 于2024年3月2日周六 15:42写道：
>
>
[-- Attachment #2: 0004-lavc-vp9dsp-R-V-V-ipred-tm.patch --]
[-- Type: text/x-patch, Size: 6536 bytes --]
From d7aa14940f52b627baf0ae4905e8af6038dc16fc Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Sat, 2 Mar 2024 09:35:22 +0800
Subject: [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
C908:
vp9_tm_4x4_8bpp_c: 116.5
vp9_tm_4x4_8bpp_rvv_i32: 43.5
vp9_tm_8x8_8bpp_c: 416.2
vp9_tm_8x8_8bpp_rvv_i32: 86.0
vp9_tm_16x16_8bpp_c: 1665.5
vp9_tm_16x16_8bpp_rvv_i32: 187.2
vp9_tm_32x32_8bpp_c: 6974.2
vp9_tm_32x32_8bpp_rvv_i32: 625.7
---
libavcodec/riscv/vp9_intra_rvv.S | 143 +++++++++++++++++++++++++++++++
libavcodec/riscv/vp9dsp.h | 8 ++
libavcodec/riscv/vp9dsp_init.c | 4 +
3 files changed, 155 insertions(+)
diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S
index 6644c68e67..d3231c9836 100644
--- a/libavcodec/riscv/vp9_intra_rvv.S
+++ b/libavcodec/riscv/vp9_intra_rvv.S
@@ -260,3 +260,146 @@ func ff_h_8x8_rvv, zve32x
ret
endfunc
+
+.macro tm_sum dst, top, offset
+ lbu t3, \offset(a2)
+ sub t3, t3, a4
+ vadd.vx \dst, \top, t3
+.endm
+
+func ff_tm_32x32_rvv, zve32x
+ lbu a4, -1(a3)
+ li t5, 32
+
+ .macro tm_sum32 n1,n2,n3,n4,n5,n6,n7,n8
+
+ vsetvli zero, t5, e16, m4, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+
+ tm_sum v0, v28, \n1
+ tm_sum v4, v28, \n2
+ tm_sum v8, v28, \n3
+ tm_sum v12, v28, \n4
+ tm_sum v16, v28, \n5
+ tm_sum v20, v28, \n6
+ tm_sum v24, v28, \n7
+ tm_sum v28, v28, \n8
+
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m2, ta, ma
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+
+ .endm
+
+ tm_sum32 31, 30, 29, 28, 27, 26, 25, 24
+ tm_sum32 23, 22, 21, 20, 19, 18, 17, 16
+ tm_sum32 15, 14, 13, 12, 11, 10, 9, 8
+ tm_sum32 7, 6, 5, 4, 3, 2, 1, 0
+
+ ret
+endfunc
+
+func ff_tm_16x16_rvv, zve32x
+ vsetivli zero, 16, e16, m2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v30, v8
+ lbu a4, -1(a3)
+
+ tm_sum v0, v30, 15
+ tm_sum v2, v30, 14
+ tm_sum v4, v30, 13
+ tm_sum v6, v30, 12
+ tm_sum v8, v30, 11
+ tm_sum v10, v30, 10
+ tm_sum v12, v30, 9
+ tm_sum v14, v30, 8
+ tm_sum v16, v30, 7
+ tm_sum v18, v30, 6
+ tm_sum v20, v30, 5
+ tm_sum v22, v30, 4
+ tm_sum v24, v30, 3
+ tm_sum v26, v30, 2
+ tm_sum v28, v30, 1
+ tm_sum v30, v30, 0
+
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m1, ta, ma
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v30, v30, 0
+ vse8.v v30, (a0)
+
+ ret
+endfunc
+
+func ff_tm_8x8_rvv, zve32x
+ vsetivli zero, 8, e16, m1, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 7
+ tm_sum v17, v28, 6
+ tm_sum v18, v28, 5
+ tm_sum v19, v28, 4
+ tm_sum v20, v28, 3
+ tm_sum v21, v28, 2
+ tm_sum v22, v28, 1
+ tm_sum v23, v28, 0
+
+ .irp n 16, 17, 18, 19, 20, 21, 22, 23
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf2, ta, ma
+ .irp n 16, 17, 18, 19, 20, 21, 22
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v23, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
+
+func ff_tm_4x4_rvv, zve32x
+ vsetivli zero, 4, e16, mf2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 3
+ tm_sum v17, v28, 2
+ tm_sum v18, v28, 1
+ tm_sum v19, v28, 0
+
+ .irp n 16, 17, 18, 19
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf4, ta, ma
+ .irp n 16, 17, 18
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v19, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h
index 2b2e0db0d8..cff2310e5c 100644
--- a/libavcodec/riscv/vp9dsp.h
+++ b/libavcodec/riscv/vp9dsp.h
@@ -72,5 +72,13 @@ void ff_h_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
void ff_h_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
+void ff_tm_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_4x4_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
#endif // #ifndef AVCODEC_RISCV_VP9DSP_RISCV_H
diff --git a/libavcodec/riscv/vp9dsp_init.c b/libavcodec/riscv/vp9dsp_init.c
index 16aeeb260a..f08c8f6a42 100644
--- a/libavcodec/riscv/vp9dsp_init.c
+++ b/libavcodec/riscv/vp9dsp_init.c
@@ -57,6 +57,10 @@ static av_cold void vp9dsp_intrapred_init_rvv(VP9DSPContext *dsp, int bpp)
dsp->intra_pred[TX_32X32][HOR_PRED] = ff_h_32x32_rvv;
dsp->intra_pred[TX_16X16][HOR_PRED] = ff_h_16x16_rvv;
dsp->intra_pred[TX_8X8][HOR_PRED] = ff_h_8x8_rvv;
+ dsp->intra_pred[TX_32X32][TM_VP8_PRED] = ff_tm_32x32_rvv;
+ dsp->intra_pred[TX_16X16][TM_VP8_PRED] = ff_tm_16x16_rvv;
+ dsp->intra_pred[TX_8X8][TM_VP8_PRED] = ff_tm_8x8_rvv;
+ dsp->intra_pred[TX_4X4][TM_VP8_PRED] = ff_tm_4x4_rvv;
}
#endif
}
--
2.44.0
[-- Attachment #3: Type: text/plain, Size: 251 bytes --]
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [FFmpeg-devel] [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
2024-03-03 2:01 ` flow gg
@ 2024-03-22 6:03 ` flow gg
0 siblings, 0 replies; 3+ messages in thread
From: flow gg @ 2024-03-22 6:03 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 295 bytes --]
Because the previous patch was updated, so it was updated in this response
flow gg <hlefthleft@gmail.com> 于2024年3月3日周日 10:01写道：
> Due to the PATCH 1/4 update, updates are made here.
>
> flow gg <hlefthleft@gmail.com> 于2024年3月2日周六 15:42写道：
>
>>
>>
[-- Attachment #2: 0004-lavc-vp9dsp-R-V-V-ipred-tm.patch --]
[-- Type: text/x-patch, Size: 6649 bytes --]
From 9561d35be25c330a0be3a371269289ce21f5ada3 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Thu, 21 Mar 2024 21:42:25 +0800
Subject: [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm
C908:
vp9_tm_4x4_8bpp_c: 116.5
vp9_tm_4x4_8bpp_rvv_i32: 43.5
vp9_tm_8x8_8bpp_c: 416.2
vp9_tm_8x8_8bpp_rvv_i32: 86.0
vp9_tm_16x16_8bpp_c: 1665.5
vp9_tm_16x16_8bpp_rvv_i32: 187.2
vp9_tm_32x32_8bpp_c: 6974.2
vp9_tm_32x32_8bpp_rvv_i32: 625.7
---
libavcodec/riscv/vp9_intra_rvv.S | 143 +++++++++++++++++++++++++++++++
libavcodec/riscv/vp9dsp.h | 8 ++
libavcodec/riscv/vp9dsp_init.c | 4 +
3 files changed, 155 insertions(+)
diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S
index bf7f6d8692..81590e4f0f 100644
--- a/libavcodec/riscv/vp9_intra_rvv.S
+++ b/libavcodec/riscv/vp9_intra_rvv.S
@@ -213,3 +213,146 @@ func ff_h_8x8_rvv, zve32x
ret
endfunc
+
+.macro tm_sum dst, top, offset
+ lbu t3, \offset(a2)
+ sub t3, t3, a4
+ vadd.vx \dst, \top, t3
+.endm
+
+func ff_tm_32x32_rvv, zve32x
+ lbu a4, -1(a3)
+ li t5, 32
+
+ .macro tm_sum32 n1,n2,n3,n4,n5,n6,n7,n8
+
+ vsetvli zero, t5, e16, m4, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+
+ tm_sum v0, v28, \n1
+ tm_sum v4, v28, \n2
+ tm_sum v8, v28, \n3
+ tm_sum v12, v28, \n4
+ tm_sum v16, v28, \n5
+ tm_sum v20, v28, \n6
+ tm_sum v24, v28, \n7
+ tm_sum v28, v28, \n8
+
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m2, ta, ma
+ .irp n 0, 4, 8, 12, 16, 20, 24, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+
+ .endm
+
+ tm_sum32 31, 30, 29, 28, 27, 26, 25, 24
+ tm_sum32 23, 22, 21, 20, 19, 18, 17, 16
+ tm_sum32 15, 14, 13, 12, 11, 10, 9, 8
+ tm_sum32 7, 6, 5, 4, 3, 2, 1, 0
+
+ ret
+endfunc
+
+func ff_tm_16x16_rvv, zve32x
+ vsetivli zero, 16, e16, m2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v30, v8
+ lbu a4, -1(a3)
+
+ tm_sum v0, v30, 15
+ tm_sum v2, v30, 14
+ tm_sum v4, v30, 13
+ tm_sum v6, v30, 12
+ tm_sum v8, v30, 11
+ tm_sum v10, v30, 10
+ tm_sum v12, v30, 9
+ tm_sum v14, v30, 8
+ tm_sum v16, v30, 7
+ tm_sum v18, v30, 6
+ tm_sum v20, v30, 5
+ tm_sum v22, v30, 4
+ tm_sum v24, v30, 3
+ tm_sum v26, v30, 2
+ tm_sum v28, v30, 1
+ tm_sum v30, v30, 0
+
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, m1, ta, ma
+ .irp n 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v30, v30, 0
+ vse8.v v30, (a0)
+
+ ret
+endfunc
+
+func ff_tm_8x8_rvv, zve32x
+ vsetivli zero, 8, e16, m1, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 7
+ tm_sum v17, v28, 6
+ tm_sum v18, v28, 5
+ tm_sum v19, v28, 4
+ tm_sum v20, v28, 3
+ tm_sum v21, v28, 2
+ tm_sum v22, v28, 1
+ tm_sum v23, v28, 0
+
+ .irp n 16, 17, 18, 19, 20, 21, 22, 23
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf2, ta, ma
+ .irp n 16, 17, 18, 19, 20, 21, 22
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v23, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
+
+func ff_tm_4x4_rvv, zve32x
+ vsetivli zero, 4, e16, mf2, ta, ma
+ vle8.v v8, (a3)
+ vzext.vf2 v28, v8
+ lbu a4, -1(a3)
+
+ tm_sum v16, v28, 3
+ tm_sum v17, v28, 2
+ tm_sum v18, v28, 1
+ tm_sum v19, v28, 0
+
+ .irp n 16, 17, 18, 19
+ vmax.vx v\n, v\n, zero
+ .endr
+
+ vsetvli zero, zero, e8, mf4, ta, ma
+ .irp n 16, 17, 18
+ vnclipu.wi v\n, v\n, 0
+ vse8.v v\n, (a0)
+ add a0, a0, a1
+ .endr
+ vnclipu.wi v24, v19, 0
+ vse8.v v24, (a0)
+
+ ret
+endfunc
diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h
index 51d9bc3957..0bf7d18a7a 100644
--- a/libavcodec/riscv/vp9dsp.h
+++ b/libavcodec/riscv/vp9dsp.h
@@ -72,6 +72,14 @@ void ff_h_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
void ff_h_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
const uint8_t *a);
+void ff_tm_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
+void ff_tm_4x4_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l,
+ const uint8_t *a);
#define VP9_8TAP_RISCV_RVV_FUNC(SIZE, type, type_idx) \
void ff_put_8tap_##type##_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \
diff --git a/libavcodec/riscv/vp9dsp_init.c b/libavcodec/riscv/vp9dsp_init.c
index 16aeeb260a..f08c8f6a42 100644
--- a/libavcodec/riscv/vp9dsp_init.c
+++ b/libavcodec/riscv/vp9dsp_init.c
@@ -57,6 +57,10 @@ static av_cold void vp9dsp_intrapred_init_rvv(VP9DSPContext *dsp, int bpp)
dsp->intra_pred[TX_32X32][HOR_PRED] = ff_h_32x32_rvv;
dsp->intra_pred[TX_16X16][HOR_PRED] = ff_h_16x16_rvv;
dsp->intra_pred[TX_8X8][HOR_PRED] = ff_h_8x8_rvv;
+ dsp->intra_pred[TX_32X32][TM_VP8_PRED] = ff_tm_32x32_rvv;
+ dsp->intra_pred[TX_16X16][TM_VP8_PRED] = ff_tm_16x16_rvv;
+ dsp->intra_pred[TX_8X8][TM_VP8_PRED] = ff_tm_8x8_rvv;
+ dsp->intra_pred[TX_4X4][TM_VP8_PRED] = ff_tm_4x4_rvv;
}
#endif
}
--
2.44.0
[-- Attachment #3: Type: text/plain, Size: 251 bytes --]
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-03-22 6:03 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-03-02 7:42 [FFmpeg-devel] [PATCH 4/4] lavc/vp9dsp: R-V V ipred tm flow gg
2024-03-03 2:01 ` flow gg
2024-03-22 6:03 ` flow gg
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git