[FFmpeg-devel] [PATCH 3/3] lavc/flacdsp: optimise RVV vector type for lpc32

[Rémi Denis-Courmont]

This is pretty much the same as for lpc16, though it only improves half
as large prediction orders. With 128-bit vectors, this gives:

   C      V old  V new
1   69.2  181.5   95.5
2  107.7  180.7   95.2
3  145.5  180.0  103.5
4  183.0  179.2  102.7
5  220.7  178.5  128.0
6  257.7  194.0  127.5
7  294.5  193.7  126.7
8  331.0  193.0  126.5

Larger prediction orders see no significant changes at that size.

The code is pretty ugly, so clean-up suggestions are most welcome.
---
 libavcodec/riscv/flacdsp_init.c | 15 ++++++++-------
 libavcodec/riscv/flacdsp_rvv.S  | 25 ++++++++++++++++++++-----
 2 files changed, 28 insertions(+), 12 deletions(-)

diff --git a/libavcodec/riscv/flacdsp_init.c b/libavcodec/riscv/flacdsp_init.c
index 735aec0691..830ae36534 100644
--- a/libavcodec/riscv/flacdsp_init.c
+++ b/libavcodec/riscv/flacdsp_init.c
@@ -71,17 +71,18 @@ av_cold void ff_flacdsp_init_riscv(FLACDSPContext *c, enum AVSampleFormat fmt,
     if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB_ADDR)) {
         int vlenb = ff_get_rv_vlenb();
 
-        if ((flags & AV_CPU_FLAG_RVB_BASIC) && vlenb >= 16)
+        if ((flags & AV_CPU_FLAG_RVB_BASIC) && vlenb >= 16) {
             c->lpc16 = ff_flac_lpc16_rvv;
 
 # if (__riscv_xlen >= 64)
-        if (flags & AV_CPU_FLAG_RVV_I64) {
-            if (vlenb > 16)
-                c->lpc32 = ff_flac_lpc32_rvv_simple;
-            else
-                c->lpc32 = ff_flac_lpc32_rvv;
-        }
+            if (flags & AV_CPU_FLAG_RVV_I64) {
+                if (vlenb > 16)
+                    c->lpc32 = ff_flac_lpc32_rvv_simple;
+                else
+                    c->lpc32 = ff_flac_lpc32_rvv;
+            }
 # endif
+        }
 
         c->wasted32 = ff_flac_wasted32_rvv;
 
diff --git a/libavcodec/riscv/flacdsp_rvv.S b/libavcodec/riscv/flacdsp_rvv.S
index 7d83909335..b292c15c8c 100644
--- a/libavcodec/riscv/flacdsp_rvv.S
+++ b/libavcodec/riscv/flacdsp_rvv.S
@@ -20,6 +20,12 @@
 
 #include "libavutil/riscv/asm.S"
 
+        .macro  vnarrow rd, rs
+        xori    \rd, \rs, 4
+        addi    \rd, \rd, -9
+        xori    \rd, \rd, 4
+        .endm
+
 func ff_flac_lpc16_rvv, zve32x, zbb
         csrr    t0, vlenb
         addi    t2, a2, -1
@@ -83,22 +89,31 @@ func ff_flac_lpc32_rvv, zve64x
         ret
 endfunc
 
-func ff_flac_lpc32_rvv_simple, zve64x
-        vsetivli zero, 1, e64, m1, ta, ma
+func ff_flac_lpc32_rvv_simple, zve64x, zbb
+        csrr    t0, vlenb
+        addi    t2, a2, -1
+        clz     t0, t0
+        clz     t2, t2
+        addi    t0, t0, (VTYPE_E64 | VTYPE_M8 | VTYPE_TA | VTYPE_MA) + 1
+        li      t1, VTYPE_E64 | VTYPE_M1 | VTYPE_TA | VTYPE_MA
+        sub     t0, t0, t2 // t0 += log2(next_power_of_two(len) / vlenb) - 1
+        max     t3, t0, t1
+        vnarrow t2, t3
+        vsetvl  zero, a2, t3 // e64
         vmv.s.x v0, zero
-        vsetvli zero, a2, e32, m4, ta, ma
+        vsetvl  zero, zero, t2 // e32
         vle32.v v8, (a1)
         sub     a4, a4, a2
         vle32.v v16, (a0)
         sh2add  a0, a2, a0
 1:
         vwmul.vv v24, v8, v16
-        vsetvli zero, zero, e64, m8, ta, ma
+        vsetvl  zero, zero, t3 // e64
         vredsum.vs v24, v24, v0
         lw      t0, (a0)
         addi    a4, a4, -1
         vmv.x.s t1, v24
-        vsetvli zero, zero, e32, m4, ta, ma
+        vsetvl  zero, zero, t2 // e32
         sra     t1, t1, a3
         add     t0, t0, t1
         vslide1down.vx v16, v16, t0
-- 
2.43.0