[FFmpeg-devel] [PATCHv2 2/2] lavc/flacdsp: R-V V flac_wasted33

[Rémi Denis-Courmont]

T-Head C908:
flac_wasted_33_c:       786.2
flac_wasted_33_rvv_i64: 486.5
---
 libavcodec/riscv/flacdsp_init.c |  4 ++++
 libavcodec/riscv/flacdsp_rvv.S  | 16 ++++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/libavcodec/riscv/flacdsp_init.c b/libavcodec/riscv/flacdsp_init.c
index 4043715a3b..0d7cff8bfe 100644
--- a/libavcodec/riscv/flacdsp_init.c
+++ b/libavcodec/riscv/flacdsp_init.c
@@ -32,6 +32,7 @@ void ff_flac_lpc32_rvv(int32_t *decoded, const int coeffs[32],
 void ff_flac_lpc32_rvv_simple(int32_t *decoded, 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,
                                        int channels, int len, int shift);
 void ff_flac_decorrelate_indep4_16_rvv(uint8_t **out, int32_t **in,
@@ -83,6 +84,9 @@ av_cold void ff_flacdsp_init_riscv(FLACDSPContext *c, enum AVSampleFormat fmt,
         c->wasted32 = ff_flac_wasted32_rvv;
 
 # if (__riscv_xlen >= 64)
+        if (flags & AV_CPU_FLAG_RVV_I64)
+            c->wasted33 = ff_flac_wasted33_rvv;
+
         switch (fmt) {
         case AV_SAMPLE_FMT_S16:
             switch (channels) {
diff --git a/libavcodec/riscv/flacdsp_rvv.S b/libavcodec/riscv/flacdsp_rvv.S
index d576a0cc21..5e5f4e7b64 100644
--- a/libavcodec/riscv/flacdsp_rvv.S
+++ b/libavcodec/riscv/flacdsp_rvv.S
@@ -116,6 +116,22 @@ func ff_flac_wasted32_rvv, zve32x
 endfunc
 
 #if (__riscv_xlen == 64)
+func ff_flac_wasted33_rvv, zve64x
+        li          t0, 1
+        sll         a2, t0, a2
+1:
+        vsetvli     t0, a3, e32, m4, ta, ma
+        vle32.v     v8, (a1)
+        sub         a3, a3, t0
+        vwmul.vx    v16, v8, a2
+        sh2add      a1, t0, a1
+        vse64.v     v16, (a0)
+        sh3add      a0, t0, a0
+        bnez        a3, 1b
+
+        ret
+endfunc
+
 func ff_flac_decorrelate_indep2_16_rvv, zve32x
         ld      a0,  (a0)
         ld      a2, 8(a1)
-- 
2.43.0