Home
last modified time | relevance | path

Searched refs:r0_tm_7 (Results 1 – 25 of 53) sorted by relevance

123

/dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/
H A Dconvolution_3x3_pack8to1_fp16s.h281 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
343 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
352 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
H A Dconvolution_3x3_pack4to1_bf16s.h165 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
227 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_bf16s_neon()
236 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
H A Dconvolution_3x3_pack8to4_fp16s.h287 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
349 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
358 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
H A Dconvolution_3x3_pack4_fp16s.h387 __fp16* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
449 vst1_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
458 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
H A Dconvolution_3x3_pack4to1.h422 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_neon() local
484 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_neon()
493 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_neon()
/dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/
H A Dconvolution_3x3_pack8to1_fp16s.h281 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
343 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
352 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
H A Dconvolution_3x3_pack4to1_bf16s.h165 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
227 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_bf16s_neon()
236 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
H A Dconvolution_3x3_pack8to4_fp16s.h287 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
349 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
358 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
H A Dconvolution_3x3_pack4_fp16s.h387 __fp16* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
449 vst1_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
458 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
H A Dconvolution_3x3_pack4to1.h422 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_neon() local
484 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_neon()
493 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_neon()
/dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/
H A Dconvolution_3x3_pack8to1_fp16s.h281 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
343 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
352 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
H A Dconvolution_3x3_pack4to1_bf16s.h165 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
227 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_bf16s_neon()
236 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
H A Dconvolution_3x3_pack8to4_fp16s.h287 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
349 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
358 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
H A Dconvolution_3x3_pack4_fp16s.h387 __fp16* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
449 vst1_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
458 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
H A Dconvolution_3x3_pack4to1.h422 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_neon() local
484 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_neon()
493 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_neon()
/dports/misc/ncnn/ncnn-20211208/src/layer/arm/
H A Dconvolution_3x3_pack8to1_fp16s.h281 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
343 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
352 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
H A Dconvolution_3x3_pack4to1_bf16s.h165 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
227 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_bf16s_neon()
236 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
H A Dconvolution_3x3_pack8to4_fp16s.h287 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
349 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
358 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
H A Dconvolution_3x3_pack4_fp16s.h387 __fp16* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
449 vst1_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
458 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
H A Dconvolution_3x3_pack4to1.h422 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_neon() local
484 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_neon()
493 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_neon()
/dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/
H A Dconvolution_3x3_pack8to1_fp16s.h281 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
343 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
352 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
H A Dconvolution_3x3_pack4to1_bf16s.h165 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
227 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_bf16s_neon()
236 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
H A Dconvolution_3x3_pack8to4_fp16s.h287 __fp16* r0_tm_7 = r0_tm_0 + tiles * 56; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
349 vst1q_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
358 r0_tm_7 += tiles * 64; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
H A Dconvolution_3x3_pack4_fp16s.h387 __fp16* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
449 vst1_f16(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
458 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
H A Dconvolution_3x3_pack4to1.h422 float* r0_tm_7 = r0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4to1_neon() local
484 vst1q_f32(r0_tm_7, _r0tm7); in conv3x3s1_winograd64_pack4to1_neon()
493 r0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4to1_neon()

123