[FFmpeg-devel] [PATCH] WIP: lavc/flacdsp: R-V Zvl256b lpc33
Rémi Denis-Courmont
remi at remlab.net
Fri May 24 23:04:40 EEST 2024
flac_lpc_33_13_c: 499.7
flac_lpc_33_13_rvv_i64: 197.7
flac_lpc_33_16_c: 601.5
flac_lpc_33_16_rvv_i64: 195.2
flac_lpc_33_29_c: 1011.5
flac_lpc_33_29_rvv_i64: 300.7
flac_lpc_33_32_c: 1099.0
flac_lpc_33_32_rvv_i64: 296.7
---
libavcodec/riscv/flacdsp_init.c | 7 +++++--
libavcodec/riscv/flacdsp_rvv.S | 27 +++++++++++++++++++++++++++
2 files changed, 32 insertions(+), 2 deletions(-)
diff --git a/libavcodec/riscv/flacdsp_init.c b/libavcodec/riscv/flacdsp_init.c
index 830ae36534..861276be28 100644
--- a/libavcodec/riscv/flacdsp_init.c
+++ b/libavcodec/riscv/flacdsp_init.c
@@ -31,6 +31,8 @@ void ff_flac_lpc32_rvv(int32_t *decoded, const int coeffs[32],
int pred_order, int qlevel, int len);
void ff_flac_lpc32_rvv_simple(int32_t *decoded, const int coeffs[32],
int pred_order, int qlevel, int len);
+void ff_flac_lpc33_rvv(int64_t *, const int32_t *, const int coeffs[32],
+ int pred_order, int qlevel, int len);
void ff_flac_wasted32_rvv(int32_t *, int shift, int len);
void ff_flac_wasted33_rvv(int64_t *, const int32_t *, int shift, int len);
void ff_flac_decorrelate_indep2_16_rvv(uint8_t **out, int32_t **in,
@@ -76,9 +78,10 @@ av_cold void ff_flacdsp_init_riscv(FLACDSPContext *c, enum AVSampleFormat fmt,
# if (__riscv_xlen >= 64)
if (flags & AV_CPU_FLAG_RVV_I64) {
- if (vlenb > 16)
+ if (vlenb > 16) {
c->lpc32 = ff_flac_lpc32_rvv_simple;
- else
+ c->lpc33 = ff_flac_lpc33_rvv;
+ } else
c->lpc32 = ff_flac_lpc32_rvv;
}
# endif
diff --git a/libavcodec/riscv/flacdsp_rvv.S b/libavcodec/riscv/flacdsp_rvv.S
index 0e10940f74..ea42f8bea4 100644
--- a/libavcodec/riscv/flacdsp_rvv.S
+++ b/libavcodec/riscv/flacdsp_rvv.S
@@ -103,6 +103,33 @@ func ff_flac_lpc32_rvv_simple, zve64x, zbb
ret
endfunc
+
+func ff_flac_lpc33_rvv, zve64x, zbb
+ vtype_vli t0, a3, t1, e64, ta, ma
+ vsetvl zero, a3, t0
+ vmv.s.x v0, zero
+ sh2add a1, a3, a1
+ vle32.v v24, (a2)
+ sub a5, a5, a3
+ vle64.v v16, (a0)
+ sh3add a0, a3, a0
+ vsext.vf2 v8, v24
+1:
+ vmul.vv v24, v8, v16
+ lw t0, (a1)
+ addi a1, a1, 4
+ vredsum.vs v24, v24, v0
+ addi a5, a5, -1
+ vmv.x.s t1, v24
+ sra t1, t1, a4
+ add t0, t0, t1
+ vslide1down.vx v16, v16, t0
+ sd t0, (a0)
+ addi a0, a0, 8
+ bnez a5, 1b
+
+ ret
+endfunc
#endif
func ff_flac_wasted32_rvv, zve32x
--
2.45.1
More information about the ffmpeg-devel
mailing list