lavc/vc1dsp: use saturating arithmetic for RVV inv_trans_dc

T-Head C908 (cycles):
vc1dsp.vc1_inv_trans_4x4_dc_c:      113.7
vc1dsp.vc1_inv_trans_4x4_dc_rvv_i32: 46.5 (before)
vc1dsp.vc1_inv_trans_4x4_dc_rvv_i32: 45.5 (after)
vc1dsp.vc1_inv_trans_4x8_dc_c:      230.7
vc1dsp.vc1_inv_trans_4x8_dc_rvv_i32: 65.7 (before)
vc1dsp.vc1_inv_trans_4x8_dc_rvv_i32: 52.5 (after)
vc1dsp.vc1_inv_trans_8x4_dc_c:      246.7
vc1dsp.vc1_inv_trans_8x4_dc_rvv_i64: 56.7 (before)
vc1dsp.vc1_inv_trans_8x4_dc_rvv_i64: 45.5 (after)
vc1dsp.vc1_inv_trans_8x8_dc_c:      419.7
vc1dsp.vc1_inv_trans_8x8_dc_rvv_i64: 81.2 (before)
vc1dsp.vc1_inv_trans_8x8_dc_rvv_i64: 53.5 (after)
This commit is contained in:
Rémi Denis-Courmont 2024-07-27 18:31:07 +03:00
parent 784a72a116
commit 324eba69f7

View File

@ -38,18 +38,20 @@ func ff_vc1_inv_trans_\cols\()x\rows\()_dc_rvv, zve64x, zba
.endif
addi t2, t2, 64
srai t2, t2, 7
.if \rows * \cols == 64
vsetvli zero, t0, e16, m8, ta, ma
.elseif \rows * \cols == 32
vsetvli zero, t0, e16, m4, ta, ma
.if \rows * \cols >= 32
vsetvli zero, t0, e8, m\mat_lmul, ta, ma
.else
vsetivli zero, \rows * \cols, e16, m2, ta, ma
vsetivli zero, \rows * \cols, e8, m\mat_lmul, ta, ma
.endif
vzext.vf2 v8, v0
vadd.vx v8, v8, t2
vmax.vx v8, v8, zero
vsetvli zero, zero, e8, m\mat_lmul, ta, ma
vnclipu.wi v0, v8, 0
bgez t2, 1f
neg t2, t2
vssubu.vx v0, v0, t2
vsetivli zero, \rows, e8, m\row_lmul, ta, ma
vsse\w\().v v0, (a0), a1
ret
1:
vsaddu.vx v0, v0, t2
vsetivli zero, \rows, e8, m\row_lmul, ta, ma
vsse\w\().v v0, (a0), a1
ret