riscv64/itx: Add 8x16 8bpc eob test

Kendryte K230                                        Before          After

inv_txfm_add_8x16_adst_adst_0_8bpc_rvv:           853.9 ( 9.00x)  698.3 (11.03x)
inv_txfm_add_8x16_adst_adst_1_8bpc_rvv:           853.8 ( 9.00x)  698.3 (11.03x)
inv_txfm_add_8x16_adst_dct_0_8bpc_rvv:            763.0 ( 9.55x)  609.2 (12.00x)
inv_txfm_add_8x16_adst_dct_1_8bpc_rvv:            763.1 ( 9.55x)  609.3 (11.94x)
inv_txfm_add_8x16_adst_flipadst_0_8bpc_rvv:       857.1 ( 8.99x)  701.6 (11.00x)
inv_txfm_add_8x16_adst_flipadst_1_8bpc_rvv:       856.8 ( 8.98x)  701.3 (10.97x)
inv_txfm_add_8x16_adst_identity_0_8bpc_rvv:       622.9 ( 9.22x)  468.5 (12.36x)
inv_txfm_add_8x16_adst_identity_1_8bpc_rvv:       622.9 ( 9.23x)  468.6 (12.37x)
inv_txfm_add_8x16_dct_adst_0_8bpc_rvv:            770.1 ( 9.32x)  655.1 (10.93x)
inv_txfm_add_8x16_dct_adst_1_8bpc_rvv:            770.1 ( 9.34x)  655.4 (10.93x)
inv_txfm_add_8x16_dct_dct_0_8bpc_rvv:             679.8 ( 1.23x)  566.1 ( 1.48x)
inv_txfm_add_8x16_dct_dct_1_8bpc_rvv:             679.8 ( 9.98x)  566.5 (11.89x)
inv_txfm_add_8x16_dct_flipadst_0_8bpc_rvv:        771.1 ( 9.34x)  667.4 (10.75x)
inv_txfm_add_8x16_dct_flipadst_1_8bpc_rvv:        771.1 ( 9.34x)  667.3 (10.76x)
inv_txfm_add_8x16_dct_identity_0_8bpc_rvv:        532.3 ( 9.84x)  422.1 (12.42x)
inv_txfm_add_8x16_dct_identity_1_8bpc_rvv:        532.4 ( 9.85x)  422.2 (12.40x)
inv_txfm_add_8x16_flipadst_adst_0_8bpc_rvv:       858.4 ( 8.98x)  699.2 (11.03x)
inv_txfm_add_8x16_flipadst_adst_1_8bpc_rvv:       858.5 ( 8.98x)  699.3 (11.03x)
inv_txfm_add_8x16_flipadst_dct_0_8bpc_rvv:        768.6 ( 9.52x)  609.7 (11.97x)
inv_txfm_add_8x16_flipadst_dct_1_8bpc_rvv:        768.4 ( 9.52x)  609.6 (11.97x)
inv_txfm_add_8x16_flipadst_flipadst_0_8bpc_rvv:   866.5 ( 8.91x)  706.5 (10.92x)
inv_txfm_add_8x16_flipadst_flipadst_1_8bpc_rvv:   866.4 ( 8.92x)  706.6 (10.95x)
inv_txfm_add_8x16_flipadst_identity_0_8bpc_rvv:   621.9 ( 9.28x)  464.6 (12.46x)
inv_txfm_add_8x16_flipadst_identity_1_8bpc_rvv:   621.8 ( 9.28x)  464.6 (12.46x)
inv_txfm_add_8x16_identity_adst_0_8bpc_rvv:       584.9 ( 9.78x)  564.1 (10.12x)
inv_txfm_add_8x16_identity_adst_1_8bpc_rvv:       584.8 ( 9.78x)  563.9 (10.12x)
inv_txfm_add_8x16_identity_dct_0_8bpc_rvv:        495.0 (10.75x)  474.6 (11.13x)
inv_txfm_add_8x16_identity_dct_1_8bpc_rvv:        494.3 (10.75x)  474.7 (11.12x)
inv_txfm_add_8x16_identity_flipadst_0_8bpc_rvv:   588.1 ( 9.76x)  568.1 (10.07x)
inv_txfm_add_8x16_identity_flipadst_1_8bpc_rvv:   588.7 ( 9.74x)  568.0 (10.07x)
inv_txfm_add_8x16_identity_identity_0_8bpc_rvv:   349.5 (10.78x)  328.8 (11.46x)
inv_txfm_add_8x16_identity_identity_1_8bpc_rvv:   349.4 (10.79x)  328.7 (11.46x)

1 files changed, 30 insertions, 17 deletions

diff --git a/src/riscv/64/itx.S b/src/riscv/64/itx.S
index 5ed3c90..c446a90 100644
--- a/src/riscv/64/itx.S
+++ b/src/riscv/64/itx.S
@@ -1866,6 +1866,8 @@ function inv_txfm_\variant\()add_8x16_rvv, export=1, ext=v
 
   vsetivli zero, 8, e16, m1, ta, ma
 
+  blt a3, a6, 1f
+
   vmv.v.x v16, zero
   addi t0, a2, 16
   vle16.v v0, (t0)
@@ -1903,6 +1905,14 @@ function inv_txfm_\variant\()add_8x16_rvv, export=1, ext=v
   vssra.vi v15, v7, 1
 .endif
 
+  j 2f
+
+1:
+.irp i, 8, 9, 10, 11, 12, 13, 14, 15
+  vmv.v.x v\i, zero
+.endr
+
+2:
   vmv.v.x v16, zero
   vle16.v v0, (a2)
   vse16.v v16, (a2)
@@ -2050,12 +2060,15 @@ endfunc
 def_fn_816_base identity_
 def_fn_816_base
 
-.macro def_fn_816 w, h, txfm1, txfm2
+.macro def_fn_816 w, h, txfm1, txfm2, eob_half
 function inv_txfm_add_\txfm1\()_\txfm2\()_\w\()x\h\()_8bpc_rvv, export=1
 .ifnc \txfm1, identity
   la a4, inv_\txfm1\()_e16_x\w\()_rvv
 .endif
   la a5, inv_\txfm2\()_e16_x\h\()_rvv
+.if \w == 8
+  li a6, \eob_half
+.endif
 .ifc \txfm1, identity
   j inv_txfm_identity_add_\w\()x\h\()_rvv
 .else
@@ -2065,22 +2078,22 @@ endfunc
 .endm
 
 .macro def_fns_816 w, h
-def_fn_816 \w, \h, dct, dct
-def_fn_816 \w, \h, identity, identity
-def_fn_816 \w, \h, dct, adst
-def_fn_816 \w, \h, dct, flipadst
-def_fn_816 \w, \h, dct, identity
-def_fn_816 \w, \h, adst, dct
-def_fn_816 \w, \h, adst, adst
-def_fn_816 \w, \h, adst, flipadst
-def_fn_816 \w, \h, flipadst, dct
-def_fn_816 \w, \h, flipadst, adst
-def_fn_816 \w, \h, flipadst, flipadst
-def_fn_816 \w, \h, identity, dct
-def_fn_816 \w, \h, adst, identity
-def_fn_816 \w, \h, flipadst, identity
-def_fn_816 \w, \h, identity, adst
-def_fn_816 \w, \h, identity, flipadst
+def_fn_816 \w, \h, dct, dct, 43
+def_fn_816 \w, \h, identity, identity, 43
+def_fn_816 \w, \h, dct, adst, 43
+def_fn_816 \w, \h, dct, flipadst, 43
+def_fn_816 \w, \h, dct, identity, 8
+def_fn_816 \w, \h, adst, dct, 43
+def_fn_816 \w, \h, adst, adst, 43
+def_fn_816 \w, \h, adst, flipadst, 43
+def_fn_816 \w, \h, flipadst, dct, 43
+def_fn_816 \w, \h, flipadst, adst, 43
+def_fn_816 \w, \h, flipadst, flipadst, 43
+def_fn_816 \w, \h, identity, dct, 64
+def_fn_816 \w, \h, adst, identity, 8
+def_fn_816 \w, \h, flipadst, identity, 8
+def_fn_816 \w, \h, identity, adst, 64
+def_fn_816 \w, \h, identity, flipadst, 64
 .endm
 
 def_fns_816 8, 16