3 Commits
Author SHA1 Message Date
Najmus Sakib AfsanandRonald S. Bultje 1718ff9ade riscv64/ipred_h: Implement ipred_h in RISC-V asm 2026-05-15 15:40:14 +00:00
Najmus Sakib Afsan 1cfad6dbca riscv64/ipred_v: Remove redundent vxrm set instr
In function ipred_v_8bpc_rvv, rvv instructions vsetvli, vle8.v,
vse8.v do not use vxrm.

Kendryte K230                     Before            After         Delta

intra_pred_v_w4_8bpc_c:       419.2 ( 1.00x)    405.2 ( 1.00x)   -3.34%
intra_pred_v_w4_8bpc_rvv:      56.7 ( 6.88x)     48.5 ( 7.73x)  -14.46%
intra_pred_v_w8_8bpc_c:       772.9 ( 1.00x)    753.3 ( 1.00x)   -2.54%
intra_pred_v_w8_8bpc_rvv:      69.9 (10.54x)     61.5 (11.67x)  -12.02%
intra_pred_v_w16_8bpc_c:     1209.7 ( 1.00x)   1221.9 ( 1.00x)    1.01%
intra_pred_v_w16_8bpc_rvv:     88.5 (13.25x)     79.4 (14.93x)  -10.28%
intra_pred_v_w32_8bpc_c:     1898.5 ( 1.00x)   1888.9 ( 1.00x)   -0.51%
intra_pred_v_w32_8bpc_rvv:    104.9 (17.49x)     95.3 (19.18x)   -9.15%
intra_pred_v_w64_8bpc_c:     3266.0 ( 1.00x)   3138.6 ( 1.00x)   -3.90%
intra_pred_v_w64_8bpc_rvv:    196.1 (16.24x)    184.6 (16.59x)   -5.86%

SpacemiT K1                       Before            After         Delta

intra_pred_v_w4_8bpc_c:       419.2 ( 1.00x)    403.5 ( 1.00x)   -3.75%
intra_pred_v_w4_8bpc_rvv:      56.7 ( 6.88x)     31.9 (11.57x)  -43.74%
intra_pred_v_w8_8bpc_c:       772.9 ( 1.00x)    756.8 ( 1.00x)   -2.08%
intra_pred_v_w8_8bpc_rvv:      69.9 (10.54x)     43.9 (16.39x)  -37.20%
intra_pred_v_w16_8bpc_c:     1209.7 ( 1.00x)   1136.5 ( 1.00x)   -6.05%
intra_pred_v_w16_8bpc_rvv:     88.5 (13.25x)     61.1 (18.00x)  -30.96%
intra_pred_v_w32_8bpc_c:     1898.5 ( 1.00x)   1837.0 ( 1.00x)   -3.24%
intra_pred_v_w32_8bpc_rvv:    104.9 (17.49x)     77.5 (22.93x)  -26.12%
intra_pred_v_w64_8bpc_c:     3266.0 ( 1.00x)   3110.6 ( 1.00x)   -4.76%
intra_pred_v_w64_8bpc_rvv:    196.1 (16.24x)    166.2 (18.28x)  -15.25%

Blackhole p100a                  Before             After         Delta

intra_pred_v_w4_8bpc_c:       368.5 ( 1.00x)    370.1 ( 1.00x)    0.43%
intra_pred_v_w4_8bpc_rvv:      36.7 ( 9.37x)     23.7 (13.99x)  -35.42%
intra_pred_v_w8_8bpc_c:       666.6 ( 1.00x)    670.2 ( 1.00x)    0.54%
intra_pred_v_w8_8bpc_rvv:      44.4 (14.34x)     33.2 (18.92x)  -25.23%
intra_pred_v_w16_8bpc_c:      970.4 ( 1.00x)    971.9 ( 1.00x)    0.15%
intra_pred_v_w16_8bpc_rvv:     58.5 (16.07x)     48.5 (19.28x)  -17.09%
intra_pred_v_w32_8bpc_c:     1577.3 ( 1.00x)   1575.8 ( 1.00x)   -0.10%
intra_pred_v_w32_8bpc_rvv:     81.5 (18.79x)     65.9 (23.11x)  -19.14%
intra_pred_v_w64_8bpc_c:     2720.1 ( 1.00x)   2724.9 ( 1.00x)    0.18%
intra_pred_v_w64_8bpc_rvv:    134.9 (19.65x)     91.6 (28.67x)  -32.10%

Benchmark results provided by Sungjoon Moon.
2026-05-12 11:26:56 +00:00
Najmus Sakib Afsan de223ad6ab riscv64/cdef: Fix up code style
The functions of cdef_filter did not use the conventional names and
the macros for declarations.

This commit matches the style used for other archs and adjusts the
following:

 - decl_cdef_fn() macro for declaration
 - dav1d_cdef_filter_wxh as the name
2026-05-10 19:58:55 +06:00