/* * Copyright (c) 2025, Alliance for Open Media. All rights reserved. * * This source code is subject to the terms of the BSD 2 Clause License and * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License * was not distributed with this source code in the LICENSE file, you can * obtain it at www.aomedia.org/license/software. If the Alliance for Open * Media Patent License 1.0 was not distributed with this source code in the * PATENTS file, you can obtain it at www.aomedia.org/license/patent. */ #include #include "config/aom_config.h" #include "config/av1_rtcd.h" #include "aom_dsp/riscv/mem_rvv.h" #include "aom_ports/mem.h" #include "av1/common/filter.h" #include "av1/common/riscv/convolve_rvv.h" static inline vuint16mf2_t highbd_convolve6_4_y_rvv( const vint16mf2_t s0, const vint16mf2_t s1, const vint16mf2_t s2, const vint16mf2_t s3, const vint16mf2_t s4, const vint16mf2_t s5, const int16_t *filter, const uint16_t max, size_t vl) { // Values at indices 0 and 7 of y_filter are zero. vint32m1_t sum = __riscv_vwmul_vx_i32m1(s0, filter[1], vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[2], s1, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[3], s2, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[4], s3, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[5], s4, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[6], s5, vl); // Add rounding constant and shift sum = __riscv_vadd_vx_i32m1(sum, 1 << (COMPOUND_ROUND1_BITS - 1), vl); // Narrow result to 16-bit with rounding and saturation vint16mf2_t res = __riscv_vnsra_wx_i16mf2(sum, COMPOUND_ROUND1_BITS, vl); // Clamp result to max value vuint16mf2_t d0 = __riscv_vreinterpret_v_i16mf2_u16mf2(__riscv_vmax_vx_i16mf2(res, 0, vl)); return __riscv_vminu_vx_u16mf2(d0, max, vl); } static inline vuint16m1_t highbd_convolve6_8_y_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const int16_t *filter, const uint16_t max, size_t vl) { // Values at indices 0 and 7 of y_filter are zero. vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[1], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s5, vl); // Add rounding constant and shift sum = __riscv_vadd_vx_i32m2(sum, 1 << (COMPOUND_ROUND1_BITS - 1), vl); // Narrow result to 16-bit with rounding and saturation vint16m1_t res = __riscv_vnsra_wx_i16m1(sum, COMPOUND_ROUND1_BITS, vl); // Clamp result to max value vuint16m1_t d0 = __riscv_vreinterpret_v_i16m1_u16m1(__riscv_vmax_vx_i16m1(res, 0, vl)); return __riscv_vminu_vx_u16m1(d0, max, vl); } static inline void highbd_convolve_y_sr_6tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter, int bd) { const uint16_t max = (1 << bd) - 1; size_t vl = __riscv_vsetvl_e16m1(w); if (w == 4) { const int16_t *s = (const int16_t *)(src_ptr + src_stride); uint16_t *d = dst_ptr; // Load initial 5 rows of data vint16mf2_t s0, s1, s2, s3, s4; load_s16_4x5(s, src_stride, &s0, &s1, &s2, &s3, &s4, vl); s += 5 * src_stride; do { // Load next 4 rows of data vint16mf2_t s5, s6, s7, s8; load_s16_4x4(s, src_stride, &s5, &s6, &s7, &s8, vl); // Perform 6-tap convolution for 4 rows vuint16mf2_t d0 = highbd_convolve6_4_y_rvv(s0, s1, s2, s3, s4, s5, y_filter, max, vl); vuint16mf2_t d1 = highbd_convolve6_4_y_rvv(s1, s2, s3, s4, s5, s6, y_filter, max, vl); vuint16mf2_t d2 = highbd_convolve6_4_y_rvv(s2, s3, s4, s5, s6, s7, y_filter, max, vl); vuint16mf2_t d3 = highbd_convolve6_4_y_rvv(s3, s4, s5, s6, s7, s8, y_filter, max, vl); // Store results store_u16_4x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16mf2(s4, vl); s1 = __riscv_vmv_v_v_i16mf2(s5, vl); s2 = __riscv_vmv_v_v_i16mf2(s6, vl); s3 = __riscv_vmv_v_v_i16mf2(s7, vl); s4 = __riscv_vmv_v_v_i16mf2(s8, vl); s += 4 * src_stride; d += 4 * dst_stride; h -= 4; } while (h != 0); } else { do { int height = h; const int16_t *s = (const int16_t *)(src_ptr + src_stride); uint16_t *d = dst_ptr; // Load initial 5 rows of data vint16m1_t s0, s1, s2, s3, s4; load_s16_8x5(s, src_stride, &s0, &s1, &s2, &s3, &s4, vl); s += 5 * src_stride; do { // Load next 4 rows of data vint16m1_t s5, s6, s7, s8; load_s16_8x4(s, src_stride, &s5, &s6, &s7, &s8, vl); // Perform 6-tap convolution for 4 rows vuint16m1_t d0 = highbd_convolve6_8_y_rvv(s0, s1, s2, s3, s4, s5, y_filter, max, vl); vuint16m1_t d1 = highbd_convolve6_8_y_rvv(s1, s2, s3, s4, s5, s6, y_filter, max, vl); vuint16m1_t d2 = highbd_convolve6_8_y_rvv(s2, s3, s4, s5, s6, s7, y_filter, max, vl); vuint16m1_t d3 = highbd_convolve6_8_y_rvv(s3, s4, s5, s6, s7, s8, y_filter, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16m1(s4, vl); s1 = __riscv_vmv_v_v_i16m1(s5, vl); s2 = __riscv_vmv_v_v_i16m1(s6, vl); s3 = __riscv_vmv_v_v_i16m1(s7, vl); s4 = __riscv_vmv_v_v_i16m1(s8, vl); s += 4 * src_stride; d += 4 * dst_stride; height -= 4; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w > 0); } } static inline vuint16mf2_t highbd_convolve8_4_y_rvv( const vint16mf2_t s0, const vint16mf2_t s1, const vint16mf2_t s2, const vint16mf2_t s3, const vint16mf2_t s4, const vint16mf2_t s5, const vint16mf2_t s6, const vint16mf2_t s7, const int16_t *filter, const uint16_t max, size_t vl) { vint32m1_t sum = __riscv_vwmul_vx_i32m1(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[7], s7, vl); // Convert to unsigned 16-bit with saturation vuint32m1_t d0 = __riscv_vreinterpret_v_i32m1_u32m1(__riscv_vmax_vx_i32m1(sum, 0, vl)); vuint16mf2_t res = __riscv_vnclipu_wx_u16mf2(d0, COMPOUND_ROUND1_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16mf2(res, max, vl); } static inline vuint16m1_t highbd_convolve8_8_y_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const int16_t *filter, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[7], s7, vl); // Convert to unsigned 16-bit with saturation vuint32m2_t d0 = __riscv_vreinterpret_v_i32m2_u32m2(__riscv_vmax_vx_i32m2(sum, 0, vl)); vuint16m1_t res = __riscv_vnclipu_wx_u16m1(d0, COMPOUND_ROUND1_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16m1(res, max, vl); } static inline void highbd_convolve_y_sr_8tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter, int bd) { const uint16_t max = (1 << bd) - 1; size_t vl = __riscv_vsetvl_e16m1(w); if (w == 4) { const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; // Load initial 7 rows of data vint16mf2_t s0, s1, s2, s3, s4, s5, s6; load_s16_4x7(s, src_stride, &s0, &s1, &s2, &s3, &s4, &s5, &s6, vl); s += 7 * src_stride; do { // Load next 4 rows of data vint16mf2_t s7, s8, s9, s10; load_s16_4x4(s, src_stride, &s7, &s8, &s9, &s10, vl); // Perform 8-tap convolution for 4 rows vuint16mf2_t d0 = highbd_convolve8_4_y_rvv(s0, s1, s2, s3, s4, s5, s6, s7, y_filter, max, vl); vuint16mf2_t d1 = highbd_convolve8_4_y_rvv(s1, s2, s3, s4, s5, s6, s7, s8, y_filter, max, vl); vuint16mf2_t d2 = highbd_convolve8_4_y_rvv(s2, s3, s4, s5, s6, s7, s8, s9, y_filter, max, vl); vuint16mf2_t d3 = highbd_convolve8_4_y_rvv(s3, s4, s5, s6, s7, s8, s9, s10, y_filter, max, vl); // Store results store_u16_4x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16mf2(s4, vl); s1 = __riscv_vmv_v_v_i16mf2(s5, vl); s2 = __riscv_vmv_v_v_i16mf2(s6, vl); s3 = __riscv_vmv_v_v_i16mf2(s7, vl); s4 = __riscv_vmv_v_v_i16mf2(s8, vl); s5 = __riscv_vmv_v_v_i16mf2(s9, vl); s6 = __riscv_vmv_v_v_i16mf2(s10, vl); s += 4 * src_stride; d += 4 * dst_stride; h -= 4; } while (h != 0); } else { do { int height = h; const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; // Load initial 7 rows of data vint16m1_t s0, s1, s2, s3, s4, s5, s6; load_s16_8x7(s, src_stride, &s0, &s1, &s2, &s3, &s4, &s5, &s6, vl); s += 7 * src_stride; do { // Load next 4 rows of data vint16m1_t s7, s8, s9, s10; load_s16_8x4(s, src_stride, &s7, &s8, &s9, &s10, vl); // Perform 8-tap convolution for 4 rows vuint16m1_t d0 = highbd_convolve8_8_y_rvv(s0, s1, s2, s3, s4, s5, s6, s7, y_filter, max, vl); vuint16m1_t d1 = highbd_convolve8_8_y_rvv(s1, s2, s3, s4, s5, s6, s7, s8, y_filter, max, vl); vuint16m1_t d2 = highbd_convolve8_8_y_rvv(s2, s3, s4, s5, s6, s7, s8, s9, y_filter, max, vl); vuint16m1_t d3 = highbd_convolve8_8_y_rvv(s3, s4, s5, s6, s7, s8, s9, s10, y_filter, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16m1(s4, vl); s1 = __riscv_vmv_v_v_i16m1(s5, vl); s2 = __riscv_vmv_v_v_i16m1(s6, vl); s3 = __riscv_vmv_v_v_i16m1(s7, vl); s4 = __riscv_vmv_v_v_i16m1(s8, vl); s5 = __riscv_vmv_v_v_i16m1(s9, vl); s6 = __riscv_vmv_v_v_i16m1(s10, vl); s += 4 * src_stride; d += 4 * dst_stride; height -= 4; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w > 0); } } static inline vuint16mf2_t highbd_convolve12_4_y_rvv( const vint16mf2_t s0, const vint16mf2_t s1, const vint16mf2_t s2, const vint16mf2_t s3, const vint16mf2_t s4, const vint16mf2_t s5, const vint16mf2_t s6, const vint16mf2_t s7, const vint16mf2_t s8, const vint16mf2_t s9, const vint16mf2_t s10, const vint16mf2_t s11, const int16_t *filter, const uint16_t max, size_t vl) { vint32m1_t sum = __riscv_vwmul_vx_i32m1(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[11], s11, vl); // Convert to unsigned 16-bit with saturation vuint32m1_t d0 = __riscv_vreinterpret_v_i32m1_u32m1(__riscv_vmax_vx_i32m1(sum, 0, vl)); vuint16mf2_t res = __riscv_vnclipu_wx_u16mf2(d0, COMPOUND_ROUND1_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16mf2(res, max, vl); } static inline vuint16m1_t highbd_convolve12_8_y_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const vint16m1_t s8, const vint16m1_t s9, const vint16m1_t s10, const vint16m1_t s11, const int16_t *filter, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[11], s11, vl); // Convert to unsigned 16-bit with saturation vuint32m2_t d0 = __riscv_vreinterpret_v_i32m2_u32m2(__riscv_vmax_vx_i32m2(sum, 0, vl)); vuint16m1_t res = __riscv_vnclipu_wx_u16m1(d0, COMPOUND_ROUND1_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16m1(res, max, vl); } static inline void highbd_convolve_y_sr_12tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter, int bd) { const uint16_t max = (1 << bd) - 1; size_t vl = __riscv_vsetvl_e16m1(w); if (w == 4) { const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; // Load initial 11 rows of data vint16mf2_t s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10; load_s16_4x11(s, src_stride, &s0, &s1, &s2, &s3, &s4, &s5, &s6, &s7, &s8, &s9, &s10, vl); s += 11 * src_stride; do { // Load next 4 rows of data vint16mf2_t s11, s12, s13, s14; load_s16_4x4(s, src_stride, &s11, &s12, &s13, &s14, vl); // Perform 12-tap convolution for 4 rows vuint16mf2_t d0 = highbd_convolve12_4_y_rvv( s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, y_filter, max, vl); vuint16mf2_t d1 = highbd_convolve12_4_y_rvv( s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, y_filter, max, vl); vuint16mf2_t d2 = highbd_convolve12_4_y_rvv(s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, y_filter, max, vl); vuint16mf2_t d3 = highbd_convolve12_4_y_rvv(s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, y_filter, max, vl); // Store results store_u16_4x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16mf2(s4, vl); s1 = __riscv_vmv_v_v_i16mf2(s5, vl); s2 = __riscv_vmv_v_v_i16mf2(s6, vl); s3 = __riscv_vmv_v_v_i16mf2(s7, vl); s4 = __riscv_vmv_v_v_i16mf2(s8, vl); s5 = __riscv_vmv_v_v_i16mf2(s9, vl); s6 = __riscv_vmv_v_v_i16mf2(s10, vl); s7 = __riscv_vmv_v_v_i16mf2(s11, vl); s8 = __riscv_vmv_v_v_i16mf2(s12, vl); s9 = __riscv_vmv_v_v_i16mf2(s13, vl); s10 = __riscv_vmv_v_v_i16mf2(s14, vl); s += 4 * src_stride; d += 4 * dst_stride; h -= 4; } while (h != 0); } else { do { int height = h; const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; // Load initial 11 rows of data vint16m1_t s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10; load_s16_8x11(s, src_stride, &s0, &s1, &s2, &s3, &s4, &s5, &s6, &s7, &s8, &s9, &s10, vl); s += 11 * src_stride; do { // Load next 4 rows of data vint16m1_t s11, s12, s13, s14; load_s16_8x4(s, src_stride, &s11, &s12, &s13, &s14, vl); // Perform 12-tap convolution for 4 rows vuint16m1_t d0 = highbd_convolve12_8_y_rvv(s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, y_filter, max, vl); vuint16m1_t d1 = highbd_convolve12_8_y_rvv(s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, y_filter, max, vl); vuint16m1_t d2 = highbd_convolve12_8_y_rvv(s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, y_filter, max, vl); vuint16m1_t d3 = highbd_convolve12_8_y_rvv(s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, y_filter, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); // Update source pointers for next iteration s0 = __riscv_vmv_v_v_i16m1(s4, vl); s1 = __riscv_vmv_v_v_i16m1(s5, vl); s2 = __riscv_vmv_v_v_i16m1(s6, vl); s3 = __riscv_vmv_v_v_i16m1(s7, vl); s4 = __riscv_vmv_v_v_i16m1(s8, vl); s5 = __riscv_vmv_v_v_i16m1(s9, vl); s6 = __riscv_vmv_v_v_i16m1(s10, vl); s7 = __riscv_vmv_v_v_i16m1(s11, vl); s8 = __riscv_vmv_v_v_i16m1(s12, vl); s9 = __riscv_vmv_v_v_i16m1(s13, vl); s10 = __riscv_vmv_v_v_i16m1(s14, vl); s += 4 * src_stride; d += 4 * dst_stride; height -= 4; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w > 0); } } void av1_highbd_convolve_y_sr_rvv(const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_y, const int subpel_y_qn, int bd) { if (w == 2 || h == 2) { av1_highbd_convolve_y_sr_c(src, src_stride, dst, dst_stride, w, h, filter_params_y, subpel_y_qn, bd); return; } const int y_filter_taps = get_filter_tap(filter_params_y, subpel_y_qn); const int vert_offset = filter_params_y->taps / 2 - 1; const int16_t *y_filter_ptr = av1_get_interp_filter_subpel_kernel( filter_params_y, subpel_y_qn & SUBPEL_MASK); src -= vert_offset * src_stride; if (y_filter_taps > 8) { highbd_convolve_y_sr_12tap_rvv(src, src_stride, dst, dst_stride, w, h, y_filter_ptr, bd); return; } if (y_filter_taps < 8) { highbd_convolve_y_sr_6tap_rvv(src, src_stride, dst, dst_stride, w, h, y_filter_ptr, bd); return; } highbd_convolve_y_sr_8tap_rvv(src, src_stride, dst, dst_stride, w, h, y_filter_ptr, bd); } static inline vuint16m1_t highbd_convolve6_8_x_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const int16_t *filter, const int32_t offset, const uint16_t max, size_t vl) { // Values at indices 0 and 7 of y_filter are zero. vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[1], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s5, vl); // Add rounding constant and offset sum = __riscv_vadd_vx_i32m2(sum, (1 << (FILTER_BITS - 1)) + offset, vl); // Narrow result to 16-bit with rounding and saturation vint16m1_t res = __riscv_vnsra_wx_i16m1(sum, FILTER_BITS, vl); // Clamp result to max value vuint16m1_t d0 = __riscv_vreinterpret_v_i16m1_u16m1(__riscv_vmax_vx_i16m1(res, 0, vl)); return __riscv_vminu_vx_u16m1(d0, max, vl); } static inline void highbd_convolve_x_sr_6tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter, ConvolveParams *conv_params, int bd) { const uint16_t max = (1 << bd) - 1; // This shim allows to do only one rounding shift instead of two. const int32_t offset = 1 << (conv_params->round_0 - 1); int height = h; size_t vl = __riscv_vsetvl_e16m1(w); do { int width = w; const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; do { vint16m1_t s00, s01, s02, s03, s04, s05; vint16m1_t s10, s11, s12, s13, s14, s15; vint16m1_t s20, s21, s22, s23, s24, s25; vint16m1_t s30, s31, s32, s33, s34, s35; // Load 6 elements for each of 4 rows load_s16_8x6(s + 0 * src_stride, 1, &s00, &s01, &s02, &s03, &s04, &s05, vl); load_s16_8x6(s + 1 * src_stride, 1, &s10, &s11, &s12, &s13, &s14, &s15, vl); load_s16_8x6(s + 2 * src_stride, 1, &s20, &s21, &s22, &s23, &s24, &s25, vl); load_s16_8x6(s + 3 * src_stride, 1, &s30, &s31, &s32, &s33, &s34, &s35, vl); // Perform convolution vuint16m1_t d0 = highbd_convolve6_8_x_rvv(s00, s01, s02, s03, s04, s05, x_filter, offset, max, vl); vuint16m1_t d1 = highbd_convolve6_8_x_rvv(s10, s11, s12, s13, s14, s15, x_filter, offset, max, vl); vuint16m1_t d2 = highbd_convolve6_8_x_rvv(s20, s21, s22, s23, s24, s25, x_filter, offset, max, vl); vuint16m1_t d3 = highbd_convolve6_8_x_rvv(s30, s31, s32, s33, s34, s35, x_filter, offset, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); s += vl; d += vl; width -= vl; } while (width > 0); src_ptr += 4 * src_stride; dst_ptr += 4 * dst_stride; height -= 4; } while (height != 0); } static inline vuint16mf2_t highbd_convolve4_4_x_rvv( const vint16mf2_t s0, const vint16mf2_t s1, const vint16mf2_t s2, const vint16mf2_t s3, const int16_t *filter, const int32_t offset, const uint16_t max, size_t vl) { vint32m1_t sum = __riscv_vwmul_vx_i32m1(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[3], s3, vl); // Add rounding constant and offset sum = __riscv_vadd_vx_i32m1(sum, (1 << (FILTER_BITS - 1)) + offset, vl); // Narrow result to 16-bit with rounding and saturation vint16mf2_t res = __riscv_vnsra_wx_i16mf2(sum, FILTER_BITS, vl); // Clamp result to max value vuint16mf2_t d0 = __riscv_vreinterpret_v_i16mf2_u16mf2(__riscv_vmax_vx_i16mf2(res, 0, vl)); return __riscv_vminu_vx_u16mf2(d0, max, vl); } static inline vuint16m1_t highbd_convolve8_8_x_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const int16_t *filter, const int32_t offset, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[7], s7, vl); sum = __riscv_vwadd_wx_i32m2(sum, offset, vl); // Convert to unsigned 16-bit with saturation vuint32m2_t d0 = __riscv_vreinterpret_v_i32m2_u32m2(__riscv_vmax_vx_i32m2(sum, 0, vl)); vuint16m1_t res = __riscv_vnclipu_wx_u16m1(d0, FILTER_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16m1(res, max, vl); } static inline void highbd_convolve_x_sr_rvv(const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter, ConvolveParams *conv_params, int bd) { // This shim allows to do only one rounding shift instead of two. const int32_t offset = 1 << (conv_params->round_0 - 1); const uint16_t max = (1 << bd) - 1; size_t vl = __riscv_vsetvl_e16m1(w); if (w == 4) { // 4-tap filters are used for blocks having width == 4. const int16_t *s = (const int16_t *)(src_ptr + 2); uint16_t *d = dst_ptr; const int16_t *x_filter_ptr = x_filter + 2; do { vint16mf2_t s00, s01, s02, s03; vint16mf2_t s10, s11, s12, s13; vint16mf2_t s20, s21, s22, s23; vint16mf2_t s30, s31, s32, s33; // Load pixels from each of 4 rows load_s16_4x4(s + 0 * src_stride, 1, &s00, &s01, &s02, &s03, vl); load_s16_4x4(s + 1 * src_stride, 1, &s10, &s11, &s12, &s13, vl); load_s16_4x4(s + 2 * src_stride, 1, &s20, &s21, &s22, &s23, vl); load_s16_4x4(s + 3 * src_stride, 1, &s30, &s31, &s32, &s33, vl); // Perform convolution for 4 rows vuint16mf2_t d0 = highbd_convolve4_4_x_rvv(s00, s01, s02, s03, x_filter_ptr, offset, max, vl); vuint16mf2_t d1 = highbd_convolve4_4_x_rvv(s10, s11, s12, s13, x_filter_ptr, offset, max, vl); vuint16mf2_t d2 = highbd_convolve4_4_x_rvv(s20, s21, s22, s23, x_filter_ptr, offset, max, vl); vuint16mf2_t d3 = highbd_convolve4_4_x_rvv(s30, s31, s32, s33, x_filter_ptr, offset, max, vl); // Store results store_u16_4x4(d, dst_stride, d0, d1, d2, d3, vl); s += 4 * src_stride; d += 4 * dst_stride; h -= 4; } while (h != 0); } else { int height = h; do { int width = w; const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; do { vint16m1_t s00, s01, s02, s03, s04, s05, s06, s07; vint16m1_t s10, s11, s12, s13, s14, s15, s16, s17; vint16m1_t s20, s21, s22, s23, s24, s25, s26, s27; vint16m1_t s30, s31, s32, s33, s34, s35, s36, s37; // Load elements for each of 4 rows load_s16_8x8(s + 0 * src_stride, 1, &s00, &s01, &s02, &s03, &s04, &s05, &s06, &s07, vl); load_s16_8x8(s + 1 * src_stride, 1, &s10, &s11, &s12, &s13, &s14, &s15, &s16, &s17, vl); load_s16_8x8(s + 2 * src_stride, 1, &s20, &s21, &s22, &s23, &s24, &s25, &s26, &s27, vl); load_s16_8x8(s + 3 * src_stride, 1, &s30, &s31, &s32, &s33, &s34, &s35, &s36, &s37, vl); // Perform convolution vuint16m1_t d0 = highbd_convolve8_8_x_rvv( s00, s01, s02, s03, s04, s05, s06, s07, x_filter, offset, max, vl); vuint16m1_t d1 = highbd_convolve8_8_x_rvv( s10, s11, s12, s13, s14, s15, s16, s17, x_filter, offset, max, vl); vuint16m1_t d2 = highbd_convolve8_8_x_rvv( s20, s21, s22, s23, s24, s25, s26, s27, x_filter, offset, max, vl); vuint16m1_t d3 = highbd_convolve8_8_x_rvv( s30, s31, s32, s33, s34, s35, s36, s37, x_filter, offset, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); s += vl; d += vl; width -= vl; } while (width > 0); src_ptr += 4 * src_stride; dst_ptr += 4 * dst_stride; height -= 4; } while (height != 0); } } static inline vuint16mf2_t highbd_convolve12_4_x_rvv( const vint16mf2_t s0, const vint16mf2_t s1, const vint16mf2_t s2, const vint16mf2_t s3, const vint16mf2_t s4, const vint16mf2_t s5, const vint16mf2_t s6, const vint16mf2_t s7, const vint16mf2_t s8, const vint16mf2_t s9, const vint16mf2_t s10, const vint16mf2_t s11, const int16_t *filter, const int32_t offset, const uint16_t max, size_t vl) { vint32m1_t sum = __riscv_vwmul_vx_i32m1(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m1(sum, filter[11], s11, vl); sum = __riscv_vwadd_wx_i32m1(sum, offset, vl); // Convert to unsigned 16-bit with saturation vuint32m1_t d0 = __riscv_vreinterpret_v_i32m1_u32m1(__riscv_vmax_vx_i32m1(sum, 0, vl)); vuint16mf2_t res = __riscv_vnclipu_wx_u16mf2(d0, FILTER_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16mf2(res, max, vl); } static inline vuint16m1_t highbd_convolve12_8_x_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const vint16m1_t s8, const vint16m1_t s9, const vint16m1_t s10, const vint16m1_t s11, const int16_t *filter, const int32_t offset, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m2(sum, filter[11], s11, vl); sum = __riscv_vwadd_wx_i32m2(sum, offset, vl); // Convert to unsigned 16-bit with saturation vuint32m2_t d0 = __riscv_vreinterpret_v_i32m2_u32m2(__riscv_vmax_vx_i32m2(sum, 0, vl)); vuint16m1_t res = __riscv_vnclipu_wx_u16m1(d0, FILTER_BITS, __RISCV_VXRM_RNU, vl); // Clamp to max return __riscv_vminu_vx_u16m1(res, max, vl); } static inline void highbd_convolve_x_sr_12tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter, ConvolveParams *conv_params, int bd) { // This shim allows to do only one rounding shift instead of two. const int32_t offset = 1 << (conv_params->round_0 - 1); const uint16_t max = (1 << bd) - 1; size_t vl = __riscv_vsetvl_e16m1(w); if (w == 4) { const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; do { vint16mf2_t s00, s01, s02, s03, s04, s05, s06, s07, s08, s09, s010, s011; vint16mf2_t s10, s11, s12, s13, s14, s15, s16, s17, s18, s19, s110, s111; vint16mf2_t s20, s21, s22, s23, s24, s25, s26, s27, s28, s29, s210, s211; vint16mf2_t s30, s31, s32, s33, s34, s35, s36, s37, s38, s39, s310, s311; // Load elements for each of 4 rows load_s16_4x12(s + 0 * src_stride, 1, &s00, &s01, &s02, &s03, &s04, &s05, &s06, &s07, &s08, &s09, &s010, &s011, vl); load_s16_4x12(s + 1 * src_stride, 1, &s10, &s11, &s12, &s13, &s14, &s15, &s16, &s17, &s18, &s19, &s110, &s111, vl); load_s16_4x12(s + 2 * src_stride, 1, &s20, &s21, &s22, &s23, &s24, &s25, &s26, &s27, &s28, &s29, &s210, &s211, vl); load_s16_4x12(s + 3 * src_stride, 1, &s30, &s31, &s32, &s33, &s34, &s35, &s36, &s37, &s38, &s39, &s310, &s311, vl); // Perform convolution vuint16mf2_t d0 = highbd_convolve12_4_x_rvv(s00, s01, s02, s03, s04, s05, s06, s07, s08, s09, s010, s011, x_filter, offset, max, vl); vuint16mf2_t d1 = highbd_convolve12_4_x_rvv(s10, s11, s12, s13, s14, s15, s16, s17, s18, s19, s110, s111, x_filter, offset, max, vl); vuint16mf2_t d2 = highbd_convolve12_4_x_rvv(s20, s21, s22, s23, s24, s25, s26, s27, s28, s29, s210, s211, x_filter, offset, max, vl); vuint16mf2_t d3 = highbd_convolve12_4_x_rvv(s30, s31, s32, s33, s34, s35, s36, s37, s38, s39, s310, s311, x_filter, offset, max, vl); // Store results store_u16_4x4(d, dst_stride, d0, d1, d2, d3, vl); s += 4 * src_stride; d += 4 * dst_stride; h -= 4; } while (h != 0); } else { int height = h; do { const int16_t *s = (const int16_t *)src_ptr; uint16_t *d = dst_ptr; int width = w; do { vint16m1_t s00, s01, s02, s03, s04, s05, s06, s07, s08, s09, s010, s011; vint16m1_t s10, s11, s12, s13, s14, s15, s16, s17, s18, s19, s110, s111; vint16m1_t s20, s21, s22, s23, s24, s25, s26, s27, s28, s29, s210, s211; vint16m1_t s30, s31, s32, s33, s34, s35, s36, s37, s38, s39, s310, s311; // Load elements for each of 4 rows load_s16_8x12(s + 0 * src_stride, 1, &s00, &s01, &s02, &s03, &s04, &s05, &s06, &s07, &s08, &s09, &s010, &s011, vl); load_s16_8x12(s + 1 * src_stride, 1, &s10, &s11, &s12, &s13, &s14, &s15, &s16, &s17, &s18, &s19, &s110, &s111, vl); load_s16_8x12(s + 2 * src_stride, 1, &s20, &s21, &s22, &s23, &s24, &s25, &s26, &s27, &s28, &s29, &s210, &s211, vl); load_s16_8x12(s + 3 * src_stride, 1, &s30, &s31, &s32, &s33, &s34, &s35, &s36, &s37, &s38, &s39, &s310, &s311, vl); // Perform convolution vuint16m1_t d0 = highbd_convolve12_8_x_rvv( s00, s01, s02, s03, s04, s05, s06, s07, s08, s09, s010, s011, x_filter, offset, max, vl); vuint16m1_t d1 = highbd_convolve12_8_x_rvv( s10, s11, s12, s13, s14, s15, s16, s17, s18, s19, s110, s111, x_filter, offset, max, vl); vuint16m1_t d2 = highbd_convolve12_8_x_rvv( s20, s21, s22, s23, s24, s25, s26, s27, s28, s29, s210, s211, x_filter, offset, max, vl); vuint16m1_t d3 = highbd_convolve12_8_x_rvv( s30, s31, s32, s33, s34, s35, s36, s37, s38, s39, s310, s311, x_filter, offset, max, vl); // Store results store_u16_8x4(d, dst_stride, d0, d1, d2, d3, vl); s += vl; d += vl; width -= vl; } while (width > 0); src_ptr += 4 * src_stride; dst_ptr += 4 * dst_stride; height -= 4; } while (height != 0); } } void av1_highbd_convolve_x_sr_rvv(const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_x, const int subpel_x_qn, ConvolveParams *conv_params, int bd) { if (w == 2 || h == 2) { av1_highbd_convolve_x_sr_c(src, src_stride, dst, dst_stride, w, h, filter_params_x, subpel_x_qn, conv_params, bd); return; } const int x_filter_taps = get_filter_tap(filter_params_x, subpel_x_qn); const int horiz_offset = filter_params_x->taps / 2 - 1; const int16_t *x_filter_ptr = av1_get_interp_filter_subpel_kernel( filter_params_x, subpel_x_qn & SUBPEL_MASK); src -= horiz_offset; if (x_filter_taps > 8) { highbd_convolve_x_sr_12tap_rvv(src, src_stride, dst, dst_stride, w, h, x_filter_ptr, conv_params, bd); return; } if (x_filter_taps <= 6 && w != 4) { highbd_convolve_x_sr_6tap_rvv(src + 1, src_stride, dst, dst_stride, w, h, x_filter_ptr, conv_params, bd); return; } highbd_convolve_x_sr_rvv(src, src_stride, dst, dst_stride, w, h, x_filter_ptr, conv_params, bd); } // store_strided_u16_4xN static inline void store_strided_u16_4xN(uint16_t *addr, vuint16m1_t vdst, ptrdiff_t stride, size_t vl) { __riscv_vse16_v_u16m1(addr, vdst, vl >> 1); vdst = __riscv_vslidedown_vx_u16m1(vdst, vl >> 1, vl); __riscv_vse16_v_u16m1(addr + stride, vdst, vl >> 1); } static inline vuint16m1_t highbd_convolve12_2d_v_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const vint16m1_t s8, const vint16m1_t s9, const vint16m1_t s10, const vint16m1_t s11, const int16_t *y_filter, const int32_t offset, const int32_t shift, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, y_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[11], s11, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); vint16m1_t i16_sum = __riscv_vnsra_wx_i16m1(sum, shift, vl); vint16m1_t iclip_sum = __riscv_vmin_vx_i16m1(__riscv_vmax_vx_i16m1(i16_sum, 0, vl), max, vl); return __riscv_vreinterpret_v_i16m1_u16m1(iclip_sum); } static inline void highbd_convolve_2d_sr_vert_12tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter_ptr, ConvolveParams *conv_params, const int bd, const int offset, size_t vl) { const int32_t shift_s32 = conv_params->round_1; const int32_t offset_s32 = offset; const uint16_t max_u16 = (1 << bd) - 1; if (w == 4) { int16_t *s = (int16_t *)src_ptr; vl = vl << 1; vint16m1_t s0 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s1 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s2 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s3 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s4 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s5 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s6 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s7 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s8 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s9 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; do { vint16m1_t s10 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s11 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s12 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s13 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vuint16m1_t d0 = highbd_convolve12_2d_v_rvv( s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d1 = highbd_convolve12_2d_v_rvv( s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); store_strided_u16_4xN(dst_ptr, d0, dst_stride, vl); dst_ptr += dst_stride << 1; store_strided_u16_4xN(dst_ptr, d1, dst_stride, vl); dst_ptr += dst_stride << 1; s0 = s4; s1 = s5; s2 = s6; s3 = s7; s4 = s8; s5 = s9; s6 = s10; s7 = s11; s8 = s12; s9 = s13; h -= 4; } while (h != 0); } else { do { int height = h; int16_t *s = (int16_t *)src_ptr; uint16_t *d = dst_ptr; vint16m1_t s0 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s1 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s2 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s3 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s4 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s5 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s6 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s7 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s8 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s9 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s10 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; do { vint16m1_t s11 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s12 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s13 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s14 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vuint16m1_t d0 = highbd_convolve12_2d_v_rvv( s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d1 = highbd_convolve12_2d_v_rvv( s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d2 = highbd_convolve12_2d_v_rvv( s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d3 = highbd_convolve12_2d_v_rvv( s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); __riscv_vse16_v_u16m1(d, d0, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d1, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d2, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d3, vl); d += dst_stride; s0 = s4; s1 = s5; s2 = s6; s3 = s7; s4 = s8; s5 = s9; s6 = s10; s7 = s11; s8 = s12; s9 = s13; s10 = s14; height -= 4; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w != 0); } } static inline vuint16m1_t highbd_convolve8_2d_v_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const int16_t *y_filter, const int32_t offset, const int32_t shift, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, y_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[7], s7, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); vint16m1_t i16_sum = __riscv_vnsra_wx_i16m1(sum, shift, vl); vint16m1_t iclip_sum = __riscv_vmin_vx_i16m1(__riscv_vmax_vx_i16m1(i16_sum, 0, vl), max, vl); return __riscv_vreinterpret_v_i16m1_u16m1(iclip_sum); } static inline void highbd_convolve_2d_sr_vert_8tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter_ptr, ConvolveParams *conv_params, int bd, const int offset, size_t vl) { const int32_t shift_s32 = conv_params->round_1; const int32_t offset_s32 = offset; const uint16_t max_u16 = (1 << bd) - 1; if (w <= 4) { int16_t *s = (int16_t *)src_ptr; vl = vl << 1; vint16m1_t s0 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s1 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s2 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s3 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s4 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s5 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; do { vint16m1_t s6 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s7 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vuint16m1_t d0 = highbd_convolve8_2d_v_rvv(s0, s1, s2, s3, s4, s5, s6, s7, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); store_strided_u16_4xN(dst_ptr, d0, dst_stride, vl); dst_ptr += dst_stride << 1; s0 = s2; s1 = s3; s2 = s4; s3 = s5; s4 = s6; s5 = s7; h -= 2; } while (h != 0); } else { do { int height = h; int16_t *s = (int16_t *)src_ptr; uint16_t *d = dst_ptr; vint16m1_t s0 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s1 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s2 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s3 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s4 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s5 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s6 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; do { vint16m1_t s7 = __riscv_vle16_v_i16m1(s, vl); vuint16m1_t d0 = highbd_convolve8_2d_v_rvv(s0, s1, s2, s3, s4, s5, s6, s7, y_filter_ptr, offset_s32, shift_s32, max_u16, vl); __riscv_vse16_v_u16m1(d, d0, vl); s0 = s1; s1 = s2; s2 = s3; s3 = s4; s4 = s5; s5 = s6; s6 = s7; s += src_stride; d += dst_stride; height--; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w != 0); } } static inline vuint16m1_t highbd_convolve6_2d_v_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const int16_t *y_filter, const int32_t offset, const int32_t shift, const uint16_t max, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, y_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, y_filter[5], s5, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); vint16m1_t i16_sum = __riscv_vnsra_wx_i16m1(sum, shift, vl); vint16m1_t iclip_sum = __riscv_vmin_vx_i16m1(__riscv_vmax_vx_i16m1(i16_sum, 0, vl), max, vl); return __riscv_vreinterpret_v_i16m1_u16m1(iclip_sum); } static inline void highbd_convolve_2d_sr_vert_6tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *y_filter_ptr, ConvolveParams *conv_params, int bd, const int offset, size_t vl) { const int32_t shift_s32 = conv_params->round_1; const int32_t offset_s32 = offset; const uint16_t max_u16 = (1 << bd) - 1; const int16_t *yfilter_6tap = y_filter_ptr + 1; if (w == 4) { int16_t *s = (int16_t *)src_ptr; vl = vl << 1; vint16m1_t s0 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s1 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s2 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s3 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; do { vint16m1_t s4 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s5 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s6 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vint16m1_t s7 = load_strided_i16_4xN(s, src_stride, vl); s += src_stride; vuint16m1_t d0 = highbd_convolve6_2d_v_rvv(s0, s1, s2, s3, s4, s5, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d1 = highbd_convolve6_2d_v_rvv(s2, s3, s4, s5, s6, s7, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); store_strided_u16_4xN(dst_ptr, d0, dst_stride, vl); dst_ptr += dst_stride << 1; store_strided_u16_4xN(dst_ptr, d1, dst_stride, vl); dst_ptr += dst_stride << 1; s0 = s4; s1 = s5; s2 = s6; s3 = s7; h -= 4; } while (h != 0); } else { do { int height = h; int16_t *s = (int16_t *)src_ptr; uint16_t *d = dst_ptr; vint16m1_t s0 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s1 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s2 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s3 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s4 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; do { vint16m1_t s5 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s6 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s7 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vint16m1_t s8 = __riscv_vle16_v_i16m1(s, vl); s += src_stride; vuint16m1_t d0 = highbd_convolve6_2d_v_rvv(s0, s1, s2, s3, s4, s5, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d1 = highbd_convolve6_2d_v_rvv(s1, s2, s3, s4, s5, s6, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d2 = highbd_convolve6_2d_v_rvv(s2, s3, s4, s5, s6, s7, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); vuint16m1_t d3 = highbd_convolve6_2d_v_rvv(s3, s4, s5, s6, s7, s8, yfilter_6tap, offset_s32, shift_s32, max_u16, vl); __riscv_vse16_v_u16m1(d, d0, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d1, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d2, vl); d += dst_stride; __riscv_vse16_v_u16m1(d, d3, vl); d += dst_stride; s0 = s4; s1 = s5; s2 = s6; s3 = s7; s4 = s8; height -= 4; } while (height != 0); src_ptr += vl; dst_ptr += vl; w -= vl; } while (w != 0); } } static inline vint16m1_t highbd_convolve12_8_2d_h_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const vint16m1_t s8, const vint16m1_t s9, const vint16m1_t s10, const vint16m1_t s11, const int16_t *x_filter, const int32_t offset, const int32_t shift, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, x_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[7], s7, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[8], s8, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[9], s9, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[10], s10, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[11], s11, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); return __riscv_vnclip_wx_i16m1(sum, shift, __RISCV_VXRM_RNU, vl); } static inline void highbd_convolve_2d_sr_horiz_12tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter_ptr, ConvolveParams *conv_params, const int offset, size_t vl) { assert(h >= 5); const int32_t shift_s32 = conv_params->round_0; const int32_t offset_s32 = offset; if (w == 4) { const int16_t *s = (int16_t *)src_ptr; int16_t *d = (int16_t *)dst_ptr; do { vint16m1_t t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11; load_s16_8x12(s, 1, &t0, &t1, &t2, &t3, &t4, &t5, &t6, &t7, &t8, &t9, &t10, &t11, vl); vint16m1_t d0 = highbd_convolve12_8_2d_h_rvv( t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, x_filter_ptr, offset_s32, shift_s32, vl); __riscv_vse16_v_i16m1(d, d0, vl); s += src_stride; d += dst_stride; } while (--h != 0); } else { int height = h; do { const int16_t *s = (int16_t *)src_ptr; int16_t *d = (int16_t *)dst_ptr; int width = w; do { vint16m1_t t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11; load_s16_8x12(s, 1, &t0, &t1, &t2, &t3, &t4, &t5, &t6, &t7, &t8, &t9, &t10, &t11, vl); vint16m1_t d0 = highbd_convolve12_8_2d_h_rvv( t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, x_filter_ptr, offset_s32, shift_s32, vl); __riscv_vse16_v_i16m1(d, d0, vl); s += vl; d += vl; width -= vl; } while (width != 0); src_ptr += src_stride; dst_ptr += dst_stride; } while (--height != 0); } } static inline vint16m1_t highbd_convolve8_4_2d_h_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const int16_t *x_filter, const int32_t offset, const int32_t shift, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, x_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[3], s3, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); return __riscv_vnclip_wx_i16m1(sum, shift, __RISCV_VXRM_RNU, vl); } static inline vint16m1_t highbd_convolve8_8_2d_h_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const vint16m1_t s6, const vint16m1_t s7, const int16_t *x_filter, const int32_t offset, const int32_t shift, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, x_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[5], s5, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[6], s6, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[7], s7, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); return __riscv_vnclip_wx_i16m1(sum, shift, __RISCV_VXRM_RNU, vl); } static inline void highbd_convolve_2d_sr_horiz_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter_ptr, ConvolveParams *conv_params, const int offset, size_t vl) { assert(h >= 5); const int32_t shift_s32 = conv_params->round_0; const int32_t offset_s32 = offset; if (w == 4) { const int16_t *x_filter = (x_filter_ptr + 2); const int16_t *s = (int16_t *)(src_ptr + 1); int16_t *d = (int16_t *)dst_ptr; do { vint16m1_t t0, t1, t2, t3; load_s16_8x4(s, 1, &t0, &t1, &t2, &t3, vl); vint16m1_t d0 = highbd_convolve8_4_2d_h_rvv(t0, t1, t2, t3, x_filter, offset_s32, shift_s32, vl); __riscv_vse16_v_i16m1(d, d0, vl); s += src_stride; d += dst_stride; } while (--h != 0); } else { do { const int16_t *s = (int16_t *)src_ptr; int16_t *d = (int16_t *)dst_ptr; int width = w; do { vint16m1_t t0, t1, t2, t3, t4, t5, t6, t7; load_s16_8x8(s, 1, &t0, &t1, &t2, &t3, &t4, &t5, &t6, &t7, vl); vint16m1_t d0 = highbd_convolve8_8_2d_h_rvv(t0, t1, t2, t3, t4, t5, t6, t7, x_filter_ptr, offset_s32, shift_s32, vl); __riscv_vse16_v_i16m1(d, d0, vl); s += vl; d += vl; width -= vl; } while (width != 0); src_ptr += src_stride; dst_ptr += dst_stride; } while (--h != 0); } } static inline vint16m1_t highbd_convolve6_8_2d_h_rvv( const vint16m1_t s0, const vint16m1_t s1, const vint16m1_t s2, const vint16m1_t s3, const vint16m1_t s4, const vint16m1_t s5, const int16_t *x_filter, const int32_t offset, const int32_t shift, size_t vl) { vint32m2_t sum = __riscv_vwmul_vx_i32m2(s0, x_filter[0], vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[1], s1, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[2], s2, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[3], s3, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[4], s4, vl); sum = __riscv_vwmacc_vx_i32m2(sum, x_filter[5], s5, vl); sum = __riscv_vadd_vx_i32m2(sum, offset, vl); return __riscv_vnclip_wx_i16m1(sum, shift, __RISCV_VXRM_RNU, vl); } static inline void highbd_convolve_2d_sr_horiz_6tap_rvv( const uint16_t *src_ptr, int src_stride, uint16_t *dst_ptr, int dst_stride, int w, int h, const int16_t *x_filter_ptr, ConvolveParams *conv_params, const int offset, size_t vl) { assert(h >= 5); const int32_t shift_s32 = conv_params->round_0; const int32_t offset_s32 = offset; const int16_t *x_filter = (x_filter_ptr + 1); do { const int16_t *s = (int16_t *)src_ptr; int16_t *d = (int16_t *)dst_ptr; int width = w; do { vint16m1_t t0, t1, t2, t3, t4, t5; load_s16_8x6(s, 1, &t0, &t1, &t2, &t3, &t4, &t5, vl); vint16m1_t d0 = highbd_convolve6_8_2d_h_rvv( t0, t1, t2, t3, t4, t5, x_filter, offset_s32, shift_s32, vl); __riscv_vse16_v_i16m1(d, d0, vl); s += vl; d += vl; width -= vl; } while (width != 0); src_ptr += src_stride; dst_ptr += dst_stride; } while (--h != 0); } void av1_highbd_convolve_2d_sr_rvv(const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_x, const InterpFilterParams *filter_params_y, const int subpel_x_qn, const int subpel_y_qn, ConvolveParams *conv_params, int bd) { if (w == 2 || h == 2) { av1_highbd_convolve_2d_sr_c(src, src_stride, dst, dst_stride, w, h, filter_params_x, filter_params_y, subpel_x_qn, subpel_y_qn, conv_params, bd); return; } DECLARE_ALIGNED(16, uint16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * MAX_SB_SIZE]); const int x_filter_taps = get_filter_tap(filter_params_x, subpel_x_qn); const int clamped_x_taps = x_filter_taps < 6 ? 6 : x_filter_taps; const int y_filter_taps = get_filter_tap(filter_params_y, subpel_y_qn); const int clamped_y_taps = y_filter_taps < 6 ? 6 : y_filter_taps; const int im_h = h + clamped_y_taps - 1; const int im_stride = MAX_SB_SIZE; const int vert_offset = clamped_y_taps / 2 - 1; const int horiz_offset = clamped_x_taps / 2 - 1; const int x_offset_initial = (1 << (bd + FILTER_BITS - 1)); const int y_offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; // The extra shim of (1 << (conv_params->round_1 - 1)) allows us to do a // simple shift left instead of a rounding saturating shift left. const int y_offset = (1 << (conv_params->round_1 - 1)) - (1 << (y_offset_bits - 1)); const uint16_t *src_ptr = src - vert_offset * src_stride - horiz_offset; const int16_t *x_filter_ptr = av1_get_interp_filter_subpel_kernel( filter_params_x, subpel_x_qn & SUBPEL_MASK); const int16_t *y_filter_ptr = av1_get_interp_filter_subpel_kernel( filter_params_y, subpel_y_qn & SUBPEL_MASK); size_t vl = __riscv_vsetvl_e16m1(w); if (x_filter_taps > 8) { highbd_convolve_2d_sr_horiz_12tap_rvv(src_ptr, src_stride, im_block, im_stride, w, im_h, x_filter_ptr, conv_params, x_offset_initial, vl); highbd_convolve_2d_sr_vert_12tap_rvv(im_block, im_stride, dst, dst_stride, w, h, y_filter_ptr, conv_params, bd, y_offset, vl); return; } if (x_filter_taps <= 6 && w != 4) { highbd_convolve_2d_sr_horiz_6tap_rvv(src_ptr, src_stride, im_block, im_stride, w, im_h, x_filter_ptr, conv_params, x_offset_initial, vl); } else { highbd_convolve_2d_sr_horiz_rvv(src_ptr, src_stride, im_block, im_stride, w, im_h, x_filter_ptr, conv_params, x_offset_initial, vl); } if (y_filter_taps <= 6) { highbd_convolve_2d_sr_vert_6tap_rvv(im_block, im_stride, dst, dst_stride, w, h, y_filter_ptr, conv_params, bd, y_offset, vl); } else { highbd_convolve_2d_sr_vert_8tap_rvv(im_block, im_stride, dst, dst_stride, w, h, y_filter_ptr, conv_params, bd, y_offset, vl); } } // Filter used is [64, 64]. void av1_highbd_convolve_x_sr_intrabc_rvv( const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_x, const int subpel_x_qn, ConvolveParams *conv_params, int bd) { assert(subpel_x_qn == 8); assert(filter_params_x->taps == 2); assert((conv_params->round_0 + conv_params->round_1) == 2 * FILTER_BITS); (void)filter_params_x; (void)subpel_x_qn; (void)conv_params; (void)bd; size_t vl = __riscv_vsetvl_e16m1(w); if (w <= 4) { do { // Load vuint16mf2_t s0_0 = __riscv_vle16_v_u16mf2(src, vl); vuint16mf2_t s0_1 = __riscv_vle16_v_u16mf2(src + 1, vl); vuint16mf2_t s1_0 = __riscv_vle16_v_u16mf2(src + src_stride, vl); vuint16mf2_t s1_1 = __riscv_vle16_v_u16mf2(src + src_stride + 1, vl); // Average the values vuint16mf2_t d0 = __riscv_vaaddu_vv_u16mf2(s0_0, s0_1, __RISCV_VXRM_RNU, vl); vuint16mf2_t d1 = __riscv_vaaddu_vv_u16mf2(s1_0, s1_1, __RISCV_VXRM_RNU, vl); // Store __riscv_vse16_v_u16mf2(dst, d0, vl); __riscv_vse16_v_u16mf2(dst + dst_stride, d1, vl); src += src_stride << 1; dst += dst_stride << 1; h -= 2; } while (h > 0); } else { do { const uint16_t *src_ptr = src; uint16_t *dst_ptr = dst; int width = w; do { // Load vuint16m1_t s0 = __riscv_vle16_v_u16m1(src_ptr, vl); vuint16m1_t s1 = __riscv_vle16_v_u16m1(src_ptr + 1, vl); vuint16m1_t s2 = __riscv_vle16_v_u16m1(src_ptr + src_stride, vl); vuint16m1_t s3 = __riscv_vle16_v_u16m1(src_ptr + src_stride + 1, vl); // Average the values vuint16m1_t d0 = __riscv_vaaddu_vv_u16m1(s0, s1, __RISCV_VXRM_RNU, vl); vuint16m1_t d1 = __riscv_vaaddu_vv_u16m1(s2, s3, __RISCV_VXRM_RNU, vl); // Store __riscv_vse16_v_u16m1(dst_ptr, d0, vl); __riscv_vse16_v_u16m1(dst_ptr + dst_stride, d1, vl); src_ptr += vl; dst_ptr += vl; width -= vl; } while (width > 0); src += src_stride << 1; dst += dst_stride << 1; h -= 2; } while (h > 0); } } // Filter used is [64, 64]. void av1_highbd_convolve_y_sr_intrabc_rvv( const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_y, const int subpel_y_qn, int bd) { assert(subpel_y_qn == 8); assert(filter_params_y->taps == 2); (void)filter_params_y; (void)subpel_y_qn; (void)bd; size_t vl = __riscv_vsetvl_e16m1(w); if (w <= 4) { vuint16mf2_t s0 = __riscv_vle16_v_u16mf2(src, vl); do { vuint16mf2_t s1 = __riscv_vle16_v_u16mf2(src + src_stride, vl); vuint16mf2_t s2 = __riscv_vle16_v_u16mf2(src + 2 * src_stride, vl); // Average the values vuint16mf2_t d0 = __riscv_vaaddu_vv_u16mf2(s0, s1, __RISCV_VXRM_RNU, vl); vuint16mf2_t d1 = __riscv_vaaddu_vv_u16mf2(s1, s2, __RISCV_VXRM_RNU, vl); // Store __riscv_vse16_v_u16mf2(dst, d0, vl); __riscv_vse16_v_u16mf2(dst + dst_stride, d1, vl); s0 = s2; src += src_stride << 1; dst += dst_stride << 1; h -= 2; } while (h > 0); } else { do { const uint16_t *src_ptr = src; uint16_t *dst_ptr = dst; int height = h; vuint16m1_t s0 = __riscv_vle16_v_u16m1(src_ptr, vl); do { vuint16m1_t s1 = __riscv_vle16_v_u16m1(src_ptr + src_stride, vl); vuint16m1_t s2 = __riscv_vle16_v_u16m1(src_ptr + 2 * src_stride, vl); // Average the values vuint16m1_t d0 = __riscv_vaaddu_vv_u16m1(s0, s1, __RISCV_VXRM_RNU, vl); vuint16m1_t d1 = __riscv_vaaddu_vv_u16m1(s1, s2, __RISCV_VXRM_RNU, vl); // Store __riscv_vse16_v_u16m1(dst_ptr, d0, vl); __riscv_vse16_v_u16m1(dst_ptr + dst_stride, d1, vl); s0 = s2; src_ptr += src_stride << 1; dst_ptr += dst_stride << 1; height -= 2; } while (height > 0); src += vl; dst += vl; w -= vl; } while (w > 0); } } // Both horizontal and vertical passes use the same 2-tap filter: [64, 64]. void av1_highbd_convolve_2d_sr_intrabc_rvv( const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams *filter_params_x, const InterpFilterParams *filter_params_y, const int subpel_x_qn, const int subpel_y_qn, ConvolveParams *conv_params, int bd) { assert(subpel_x_qn == 8); assert(subpel_y_qn == 8); assert(filter_params_x->taps == 2 && filter_params_y->taps == 2); assert((conv_params->round_0 + conv_params->round_1) == 2 * FILTER_BITS); assert(w <= MAX_SB_SIZE && h <= MAX_SB_SIZE); (void)filter_params_x; (void)subpel_x_qn; (void)filter_params_y; (void)subpel_y_qn; (void)conv_params; (void)bd; size_t vl = __riscv_vsetvl_e16m1(w); if (w <= 8) { // Horizontal filter. vuint16m1_t s0 = __riscv_vle16_v_u16m1(src, vl); vuint16m1_t s1 = __riscv_vle16_v_u16m1(src + 1, vl); src += src_stride; vuint16m1_t sum0 = __riscv_vadd_vv_u16m1(s0, s1, vl); do { vuint16m1_t s2 = __riscv_vle16_v_u16m1(src, vl); vuint16m1_t s3 = __riscv_vle16_v_u16m1(src + 1, vl); src += src_stride; vuint16m1_t s4 = __riscv_vle16_v_u16m1(src, vl); vuint16m1_t s5 = __riscv_vle16_v_u16m1(src + 1, vl); src += src_stride; vuint16m1_t sum1 = __riscv_vadd_vv_u16m1(s2, s3, vl); vuint16m1_t sum2 = __riscv_vadd_vv_u16m1(s4, s5, vl); // Vertical filter. vuint16m1_t d0 = __riscv_vadd_vx_u16m1(__riscv_vadd_vv_u16m1(sum0, sum1, vl), 2, vl); vuint16m1_t d1 = __riscv_vadd_vx_u16m1(__riscv_vadd_vv_u16m1(sum1, sum2, vl), 2, vl); d0 = __riscv_vsrl_vx_u16m1(d0, 2, vl); d1 = __riscv_vsrl_vx_u16m1(d1, 2, vl); __riscv_vse16_v_u16m1(dst, d0, vl); dst += dst_stride; __riscv_vse16_v_u16m1(dst, d1, vl); dst += dst_stride; sum0 = sum2; h -= 2; } while (h != 0); } else { do { uint16_t *src_ptr = (uint16_t *)src; uint16_t *dst_ptr = dst; int height = h; // Horizontal filter. vuint16m1_t s0 = __riscv_vle16_v_u16m1(src_ptr, vl); vuint16m1_t s1 = __riscv_vle16_v_u16m1(src_ptr + 1, vl); src_ptr += src_stride; vuint16m1_t sum0 = __riscv_vadd_vv_u16m1(s0, s1, vl); do { vuint16m1_t s2 = __riscv_vle16_v_u16m1(src_ptr, vl); vuint16m1_t s3 = __riscv_vle16_v_u16m1(src_ptr + 1, vl); src_ptr += src_stride; vuint16m1_t s4 = __riscv_vle16_v_u16m1(src_ptr, vl); vuint16m1_t s5 = __riscv_vle16_v_u16m1(src_ptr + 1, vl); src_ptr += src_stride; vuint16m1_t sum1 = __riscv_vadd_vv_u16m1(s2, s3, vl); vuint16m1_t sum2 = __riscv_vadd_vv_u16m1(s4, s5, vl); // Vertical filter. vuint16m1_t d0 = __riscv_vadd_vx_u16m1(__riscv_vadd_vv_u16m1(sum0, sum1, vl), 2, vl); vuint16m1_t d1 = __riscv_vadd_vx_u16m1(__riscv_vadd_vv_u16m1(sum1, sum2, vl), 2, vl); d0 = __riscv_vsrl_vx_u16m1(d0, 2, vl); d1 = __riscv_vsrl_vx_u16m1(d1, 2, vl); __riscv_vse16_v_u16m1(dst_ptr, d0, vl); dst_ptr += dst_stride; __riscv_vse16_v_u16m1(dst_ptr, d1, vl); dst_ptr += dst_stride; sum0 = __riscv_vmv_v_v_u16m1(sum2, vl); height -= 2; } while (height != 0); src += vl; dst += vl; w -= vl; } while (w != 0); } }