[FFmpeg-cvslog] lavc/vp7dsp: revector ff_vp7_dc_wht_rvv

[Rémi Denis-Courmont]

ffmpeg | branch: master | Rémi Denis-Courmont <remi at remlab.net> | Sun May 26 13:19:15 2024 +0300| [4a0e629b6f4b6fb612ce08a471fcf3c249d35eaa] | committer: Rémi Denis-Courmont

lavc/vp7dsp: revector ff_vp7_dc_wht_rvv

This prepares for some code reuse.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4a0e629b6f4b6fb612ce08a471fcf3c249d35eaa
---

 libavcodec/riscv/vp7dsp_init.c |  3 ++-
 libavcodec/riscv/vp7dsp_rvv.S  | 23 ++++++++++++++---------
 2 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/libavcodec/riscv/vp7dsp_init.c b/libavcodec/riscv/vp7dsp_init.c
index 5a57a39d0b..1406eb1c6a 100644
--- a/libavcodec/riscv/vp7dsp_init.c
+++ b/libavcodec/riscv/vp7dsp_init.c
@@ -32,7 +32,8 @@ av_cold void ff_vp7dsp_init_riscv(VP8DSPContext *c)
 #if HAVE_RVV
     int flags = av_get_cpu_flags();
 
-    if ((flags & AV_CPU_FLAG_RVV_I32) && ff_rv_vlen_least(128)) {
+    if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB_ADDR) &&
+        ff_rv_vlen_least(128)) {
 #if __riscv_xlen >= 64
         c->vp8_luma_dc_wht = ff_vp7_luma_dc_wht_rvv;
 #endif
diff --git a/libavcodec/riscv/vp7dsp_rvv.S b/libavcodec/riscv/vp7dsp_rvv.S
index 819d2056dc..06e251f5ce 100644
--- a/libavcodec/riscv/vp7dsp_rvv.S
+++ b/libavcodec/riscv/vp7dsp_rvv.S
@@ -22,6 +22,15 @@
 
 #if __riscv_xlen >= 64
 func ff_vp7_luma_dc_wht_rvv, zve32x
+        li          a2, 4 * 16 * 2
+        li          a7, 16 * 2
+        jal         t0, 1f
+        vsse16.v    v4, (a0), a7
+        vsse16.v    v5, (t1), a7
+        vsse16.v    v6, (t2), a7
+        vsse16.v    v7, (t3), a7
+        ret
+1:
         csrwi       vxrm, 0
         li          t4, 12540
         vsetivli    zero, 4, e16, mf2, ta, ma
@@ -58,14 +67,14 @@ func ff_vp7_luma_dc_wht_rvv, zve32x
         vle16.v     v2, (t2)
         vle16.v     v3, (t3)
         vwmul.vx    v8, v1, t4
-        li          t0, 16 * 2
         vwmul.vx    v9, v3, t6
-        addi        t1, a0, 1 * 4 * 16 * 2
+        add         t1, a2, a0
         vwmul.vx    v10, v1, t6
-        addi        t2, a0, 2 * 4 * 16 * 2
+        sh1add      t2, a2, a0
         vwmul.vx    v11, v3, t4
-        addi        t3, a0, 3 * 4 * 16 * 2
+        sh1add      a2, a2, a2 # a2 *= 3
         vwadd.vv    v4, v0, v2
+        add         t3, a2, a0
         vwsub.vv    v5, v0, v2
         vsetvli     zero, zero, e32, m1, ta, ma
         vmul.vx     v4, v4, t5
@@ -86,10 +95,6 @@ func ff_vp7_luma_dc_wht_rvv, zve32x
         vnclip.wi   v5, v1, 18
         vnclip.wi   v6, v2, 18
         vnclip.wi   v7, v3, 18
-        vsse16.v    v4, (a0), t0
-        vsse16.v    v5, (t1), t0
-        vsse16.v    v6, (t2), t0
-        vsse16.v    v7, (t3), t0
-        ret
+        jr          t0
 endfunc
 #endif