Home
last modified time | relevance | path

Searched refs:_sum47 (Results 1 – 5 of 5) sorted by relevance

/dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/
H A Dconvolution_3x3_int8.h4139 int32x4_t _sum47 = {}; in conv3x3s2_packed_int8_neon() local
4145 _sum47 = vld1q_lane_s32(outptr4, _sum47, 0); // out4 in conv3x3s2_packed_int8_neon()
4146 _sum47 = vld1q_lane_s32(outptr5, _sum47, 1); // out5 in conv3x3s2_packed_int8_neon()
4147 _sum47 = vld1q_lane_s32(outptr6, _sum47, 2); // out6 in conv3x3s2_packed_int8_neon()
4148 _sum47 = vld1q_lane_s32(outptr7, _sum47, 3); // out7 in conv3x3s2_packed_int8_neon()
4164 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r0, 2); in conv3x3s2_packed_int8_neon()
4180 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r1, 2); in conv3x3s2_packed_int8_neon()
4196 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r2, 2); in conv3x3s2_packed_int8_neon()
4201 _sum47 = vaddq_s32(_sum47, _sum0n); in conv3x3s2_packed_int8_neon()
4207 vst1q_lane_s32(outptr4, _sum47, 0); in conv3x3s2_packed_int8_neon()
[all …]
/dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/
H A Dconvolution_3x3_int8.h4139 int32x4_t _sum47 = {}; in conv3x3s2_packed_int8_neon() local
4145 _sum47 = vld1q_lane_s32(outptr4, _sum47, 0); // out4 in conv3x3s2_packed_int8_neon()
4146 _sum47 = vld1q_lane_s32(outptr5, _sum47, 1); // out5 in conv3x3s2_packed_int8_neon()
4147 _sum47 = vld1q_lane_s32(outptr6, _sum47, 2); // out6 in conv3x3s2_packed_int8_neon()
4148 _sum47 = vld1q_lane_s32(outptr7, _sum47, 3); // out7 in conv3x3s2_packed_int8_neon()
4164 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r0, 2); in conv3x3s2_packed_int8_neon()
4180 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r1, 2); in conv3x3s2_packed_int8_neon()
4196 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r2, 2); in conv3x3s2_packed_int8_neon()
4201 _sum47 = vaddq_s32(_sum47, _sum0n); in conv3x3s2_packed_int8_neon()
4207 vst1q_lane_s32(outptr4, _sum47, 0); in conv3x3s2_packed_int8_neon()
[all …]
/dports/misc/ncnn/ncnn-20211208/src/layer/arm/
H A Dconvolution_3x3_int8.h4139 int32x4_t _sum47 = {}; in conv3x3s2_packed_int8_neon() local
4145 _sum47 = vld1q_lane_s32(outptr4, _sum47, 0); // out4 in conv3x3s2_packed_int8_neon()
4146 _sum47 = vld1q_lane_s32(outptr5, _sum47, 1); // out5 in conv3x3s2_packed_int8_neon()
4147 _sum47 = vld1q_lane_s32(outptr6, _sum47, 2); // out6 in conv3x3s2_packed_int8_neon()
4148 _sum47 = vld1q_lane_s32(outptr7, _sum47, 3); // out7 in conv3x3s2_packed_int8_neon()
4164 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r0, 2); in conv3x3s2_packed_int8_neon()
4180 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r1, 2); in conv3x3s2_packed_int8_neon()
4196 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r2, 2); in conv3x3s2_packed_int8_neon()
4201 _sum47 = vaddq_s32(_sum47, _sum0n); in conv3x3s2_packed_int8_neon()
4207 vst1q_lane_s32(outptr4, _sum47, 0); in conv3x3s2_packed_int8_neon()
[all …]
/dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/
H A Dconvolution_3x3_int8.h4139 int32x4_t _sum47 = {}; in conv3x3s2_packed_int8_neon() local
4145 _sum47 = vld1q_lane_s32(outptr4, _sum47, 0); // out4 in conv3x3s2_packed_int8_neon()
4146 _sum47 = vld1q_lane_s32(outptr5, _sum47, 1); // out5 in conv3x3s2_packed_int8_neon()
4147 _sum47 = vld1q_lane_s32(outptr6, _sum47, 2); // out6 in conv3x3s2_packed_int8_neon()
4148 _sum47 = vld1q_lane_s32(outptr7, _sum47, 3); // out7 in conv3x3s2_packed_int8_neon()
4164 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r0, 2); in conv3x3s2_packed_int8_neon()
4180 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r1, 2); in conv3x3s2_packed_int8_neon()
4196 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r2, 2); in conv3x3s2_packed_int8_neon()
4201 _sum47 = vaddq_s32(_sum47, _sum0n); in conv3x3s2_packed_int8_neon()
4207 vst1q_lane_s32(outptr4, _sum47, 0); in conv3x3s2_packed_int8_neon()
[all …]
/dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/
H A Dconvolution_3x3_int8.h4139 int32x4_t _sum47 = {}; in conv3x3s2_packed_int8_neon() local
4145 _sum47 = vld1q_lane_s32(outptr4, _sum47, 0); // out4 in conv3x3s2_packed_int8_neon()
4146 _sum47 = vld1q_lane_s32(outptr5, _sum47, 1); // out5 in conv3x3s2_packed_int8_neon()
4147 _sum47 = vld1q_lane_s32(outptr6, _sum47, 2); // out6 in conv3x3s2_packed_int8_neon()
4148 _sum47 = vld1q_lane_s32(outptr7, _sum47, 3); // out7 in conv3x3s2_packed_int8_neon()
4164 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r0, 2); in conv3x3s2_packed_int8_neon()
4180 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r1, 2); in conv3x3s2_packed_int8_neon()
4196 _sum47 = vmlal_laneq_s16(_sum47, vget_high_s16(_k2), _r2, 2); in conv3x3s2_packed_int8_neon()
4201 _sum47 = vaddq_s32(_sum47, _sum0n); in conv3x3s2_packed_int8_neon()
4207 vst1q_lane_s32(outptr4, _sum47, 0); in conv3x3s2_packed_int8_neon()
[all …]