lavc/me_cmp: R-V V vsse vsad intra

C908:
vsad_4_c: 681.0
vsad_4_rvv_i32: 182.2
vsad_5_c: 278.0
vsad_5_rvv_i32: 145.2
vsse_4_c: 595.0
vsse_4_rvv_i32: 125.2
vsse_5_c: 281.0
vsse_5_rvv_i32: 101.2

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

libavcodec/riscv/me_cmp_init.c

@@ -48,8 +48,12 @@ int ff_sse4_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
 
 int ff_vsse16_rvv(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h);
 int ff_vsse8_rvv(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h);
+int ff_vsse_intra16_rvv(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, ptrdiff_t stride, int h);
+int ff_vsse_intra8_rvv(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, ptrdiff_t stride, int h);
 int ff_vsad16_rvv(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h);
 int ff_vsad8_rvv(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h);
+int ff_vsad_intra16_rvv(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, ptrdiff_t stride, int h);
+int ff_vsad_intra8_rvv(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, ptrdiff_t stride, int h);
 
 av_cold void ff_me_cmp_init_riscv(MECmpContext *c, AVCodecContext *avctx)
 {
@@ -72,8 +76,12 @@ av_cold void ff_me_cmp_init_riscv(MECmpContext *c, AVCodecContext *avctx)
 
         c->vsse[0] = ff_vsse16_rvv;
         c->vsse[1] = ff_vsse8_rvv;
+        c->vsse[4] = ff_vsse_intra16_rvv;
+        c->vsse[5] = ff_vsse_intra8_rvv;
         c->vsad[0] = ff_vsad16_rvv;
         c->vsad[1] = ff_vsad8_rvv;
+        c->vsad[4] = ff_vsad_intra16_rvv;
+        c->vsad[5] = ff_vsad_intra8_rvv;
     }
 #endif
 }

libavcodec/riscv/me_cmp_rvv.S

@@ -314,6 +314,68 @@ endfunc
         ret
 .endm
 
+.macro  vsad_vsse_intra16 type
+        vsetivli        t0, 16, e32, m4, ta, ma
+        addi            a4, a4, -1
+        add             t1, a1, a3
+        vmv.v.x         v24, zero
+        vmv.s.x         v0, zero
+1:
+        vsetvli         zero, zero, e8, m1, tu, ma
+        vle8.v          v4, (a1)
+        vle8.v          v12, (t1)
+        addi            a4, a4, -1
+        vwsubu.vv       v16, v4, v12
+        vsetvli         zero, zero, e16, m2, tu, ma
+
+.ifc \type,abs
+        vabsaddu        v24, v16, v12
+.endif
+.ifc \type,square
+        vwmacc.vv       v24, v16, v16
+.endif
+
+        add             a1, a1, a3
+        add             t1, t1, a3
+        bnez            a4, 1b
+
+        vsetvli         zero, zero, e32, m4, tu, ma
+        vredsum.vs      v0, v24, v0
+        vmv.x.s         a0, v0
+        ret
+.endm
+
+.macro  vsad_vsse_intra8 type
+        vsetivli        t0, 8, e32, m2, ta, ma
+        addi            a4, a4, -1
+        add             t1, a1, a3
+        vmv.v.x         v24, zero
+        vmv.s.x         v0, zero
+1:
+        vsetvli         zero, zero, e8, mf2, tu, ma
+        vle8.v          v4, (a1)
+        vle8.v          v12, (t1)
+        addi            a4, a4, -1
+        vwsubu.vv       v16, v4, v12
+        vsetvli         zero, zero, e16, m1, tu, ma
+
+.ifc \type,abs
+        vabsaddu        v24, v16, v12
+.endif
+.ifc \type,square
+        vwmacc.vv       v24, v16, v16
+.endif
+
+        add             a1, a1, a3
+        add             t1, t1, a3
+        bnez            a4, 1b
+
+        vsetvli         zero, zero, e32, m2, tu, ma
+        vredsum.vs      v0, v24, v0
+        vmv.x.s         a0, v0
+        ret
+.endm
+
 func ff_vsse16_rvv, zve32x
         vsad_vsse16 square
 endfunc
@@ -322,6 +384,14 @@ func ff_vsse8_rvv, zve32x
         vsad_vsse8 square
 endfunc
 
+func ff_vsse_intra16_rvv, zve32x
+        vsad_vsse_intra16 square
+endfunc
+
+func ff_vsse_intra8_rvv, zve32x
+        vsad_vsse_intra8 square
+endfunc
+
 func ff_vsad16_rvv, zve32x
         vsad_vsse16 abs
 endfunc
@@ -329,3 +399,11 @@ endfunc
 func ff_vsad8_rvv, zve32x
         vsad_vsse8 abs
 endfunc
+
+func ff_vsad_intra16_rvv, zve32x
+        vsad_vsse_intra16 abs
+endfunc
+
+func ff_vsad_intra8_rvv, zve32x
+        vsad_vsse_intra8 abs
+endfunc