Functions — github.com/Maratyszcza/NNPACK

CodeHub MCPcopy Create free account

hub / github.com/Maratyszcza/NNPACK / functions

Functions873 in github.com/Maratyszcza/NNPACK

⨍ Functions 873◇ Types & classes 270

↓ 900 callersMethoderrorLimit

test/testers/relu.h:70

↓ 452 callersMethodtestInference

test/testers/convolution.h:372

↓ 441 callersFunctioncimagf

include/nnpack/complex.h:16

↓ 441 callersFunctioncrealf

include/nnpack/complex.h:12

↓ 321 callersMethodinputSize

test/testers/pooling.h:118

↓ 321 callersMethoditerations

test/testers/relu.h:61

↓ 288 callersMethodbatchSize

test/testers/relu.h:94

↓ 258 callersFunctionbutterflyfc

src/ref/fft/complex.h:18

↓ 246 callersFunctionvst1q_f32_aligned

include/nnpack/arm_neon.h:20

↓ 244 callersFunctionvld1q_f32_aligned

include/nnpack/arm_neon.h:16

↓ 228 callersFunctionvmuladdq_f32

include/nnpack/arm_neon.h:65

↓ 226 callersMethodtestOutput

test/testers/relu.h:121

↓ 198 callersFunctionscalar_butterfly

src/scalar/butterfly.h:11

↓ 101 callersMethodkernelSize

test/testers/convolution.h:149

↓ 98 callersFunctionpsimd_butterfly_f32

src/psimd/butterfly.h:6

↓ 98 callersFunctionvmuladd_f32

include/nnpack/arm_neon.h:113

↓ 96 callersFunctionconv3

* VGG model A conv3 layer: * input channels = 128 * output channels = 256 * input size = 56x56 * implicit padding = 1 * k

test/models/vgg-a.h:80

↓ 96 callersFunctionconv4

* VGG model A conv4 layer: * input channels = 256 * output channels = 256 * input size = 56x56 * implicit padding = 1 * k

test/models/vgg-a.h:110

↓ 96 callersFunctionconv5

* VGG model A conv5 layer: * input channels = 256 * output channels = 512 * input size = 28x28 * implicit padding = 1 * k

test/models/vgg-a.h:128

↓ 89 callersMethodtestInputGradient

test/testers/relu.h:180

↓ 84 callersMethodsimdWidth

test/testers/fourier.h:41

↓ 81 callersFunctionmin

include/nnpack/utils.h:21

↓ 81 callersMethodoutputChannels

test/testers/convolution.h:121

↓ 79 callersMethodfftSize

test/testers/fourier.h:32

↓ 74 callersMethodaccumulateC

test/testers/gemm-ukernel.h:87

↓ 72 callersFunctionconv2

* VGG model A conv2 layer: * input channels = 64 * output channels = 128 * input size = 112x112 * implicit padding = 1 *

test/models/vgg-a.h:50

↓ 72 callersMethodinputChannels

test/testers/convolution.h:112

↓ 72 callersFunctionscalar_swap

src/scalar/butterfly.h:4

↓ 72 callersFunctionvld1q_f32_f16

include/nnpack/arm_neon.h:26

↓ 69 callersMethodtestKernelGradient

test/testers/convolution.h:319

↓ 64 callersMethodkernelHeight

test/testers/convolution.h:153

↓ 64 callersMethodkernelWidth

test/testers/convolution.h:157

↓ 63 callersFunctionvst1q_f16_f32

include/nnpack/arm_neon.h:35

↓ 59 callersFunctionbutterfly

(a, b, negate_a=False, negate_b=False, scale_a=None, scale_b=None, negate_out_b=False, writeback=True)

src/x86_64-fma/common.py:65

↓ 58 callersMethodmultithreading

test/testers/relu.h:85

↓ 54 callersMethodtestSGEMM

test/testers/gemm-ukernel.h:290

↓ 51 callersMethodinputPadding

test/testers/pooling.h:185

↓ 48 callersMethodchannels

test/testers/relu.h:103

↓ 42 callersFunctionscalar_butterfly_with_negated_b

src/scalar/butterfly.h:25

↓ 42 callersFunctionswapfc

src/ref/fft/complex.h:25

↓ 40 callersFunctionconv1

* VGG model A conv1 layer: * input channels = 3 * output channels = 64 * input size = 224x224 * implicit padding = 1 * ke

test/models/vgg-a.h:20

↓ 40 callersMethodmr

test/testers/gemm-ukernel.h:37

↓ 40 callersMethodnr

test/testers/gemm-ukernel.h:46

↓ 40 callersMethodpoolingSize

test/testers/pooling.h:136

↓ 40 callersMethodpoolingStride

test/testers/pooling.h:154

↓ 39 callersFunctionnnp_initialize

↓ 37 callersMethodkc

test/testers/gemm-ukernel.h:55

↓ 36 callersFunctionvmuladd_lane0_f32

include/nnpack/arm_neon.h:129

↓ 36 callersFunctionvmuladd_lane1_f32

include/nnpack/arm_neon.h:137

↓ 32 callersFunctionconv6

* VGG model A conv6 layer: * input channels = 512 * output channels = 512 * input size = 28x28 * implicit padding = 1 * k

test/models/vgg-a.h:158

↓ 32 callersFunctionconv8

* VGG model A conv8 layer: * input channels = 512 * output channels = 512 * input size = 14x14 * implicit padding = 1 * k

test/models/vgg-a.h:176

↓ 31 callersFunctionmax

include/nnpack/utils.h:17

↓ 31 callersFunctionvmuladdq_lane0_f32

include/nnpack/arm_neon.h:81

↓ 30 callersFunctiondoz

include/nnpack/utils.h:13

↓ 30 callersMethodoutputSubsampling

test/testers/convolution.h:182

↓ 29 callersFunctionmedian

bench/median.c:43

↓ 29 callersFunctionneon_reluq_f32

include/nnpack/activations.h:27

↓ 28 callersMethodtestOutputInplace

test/testers/relu.h:151

↓ 27 callersFunctionneon_relu_f32

include/nnpack/activations.h:37

↓ 27 callersFunctionrelu

include/nnpack/activations.h:6

↓ 27 callersFunctionround_down

include/nnpack/utils.h:33

↓ 27 callersFunctionvmuladdq_lane1_f32

include/nnpack/arm_neon.h:89

↓ 26 callersFunctiondivide_round_up

include/nnpack/utils.h:37

↓ 26 callersFunctionvmulsubq_f32

include/nnpack/arm_neon.h:73

↓ 24 callersMethodoutputSize

test/testers/pooling.h:158

↓ 24 callersFunctionvmulsub_f32

include/nnpack/arm_neon.h:121

↓ 23 callersMethodtestOptimizedComplex

* Validates that optimized complex 1D FFT produces the same output as reference implementation. */

test/testers/fourier.h:379

↓ 23 callersFunctionwinograd_f6k3_output_transformq

src/neon/winograd/f6x6k3x3.h:222

↓ 20 callersFunctionpsimd_exp_f32

src/psimd/exp.h:6

↓ 16 callersFunctionstore_ymm_result

(variable, result)

src/x86_64-fma/fft16x16.py:26

↓ 15 callersFunction_MM_SHUFFLE

src/x86_64-fma/common.py:51

↓ 14 callersFunctionwinograd_f6k3_output_transform

src/neon/winograd/f6x6k3x3.h:160

↓ 13 callersFunctionpsimd_ifft16_real_f32

src/psimd/fft/real.h:161

↓ 13 callersFunctiontranspose2x2x2x64

(ymm_a, ymm_b, use_blend=True)

src/x86_64-fma/common.py:178

↓ 12 callersFunctionnnp_convolution_inference

include/nnpack.h:635

↓ 12 callersFunctionpsimd_transpose4x4_f32

src/psimd/transpose.h:6

↓ 12 callersMethodtestHXGEMM

test/testers/gemm-ukernel.h:194

↓ 12 callersMethodtestOptimizedDualReal

* Validates that optimized dual-sequence real 1D FFT produces the same output as reference implementation. */

test/testers/fourier.h:393

↓ 12 callersMethodtestOptimizedReal

* Validates that optimized real 1D FFT produces the same output as reference implementation. */

test/testers/fourier.h:386

↓ 12 callersFunctiontranspose2x2x128

(ymm_a, ymm_b, use_blend=True)

src/x86_64-fma/common.py:152

↓ 11 callersFunctionTEST

test/sgemm/neon.cc:8

↓ 11 callersFunctionextract_time

↓ 10 callersFunctionpsimd_butterfly_with_negated_b_f32

src/psimd/butterfly.h:26

↓ 10 callersFunctionread_timer

bench/perf_counter.h:74

↓ 10 callersMethodtestSoa

* Validates that complex 1D FFT with structure-of-arrays layout produces the same output * as FFT with array-of-structures layout. This function wor

test/testers/fourier.h:117

↓ 10 callersFunctionwinograd_f6k3_kernel_transform

src/neon/winograd/f6x6k3x3.h:85

↓ 9 callersFunctionconv1_relu

* VGG model A conv1 ReLU layer: * channels = 64 * image size = 224x224 */

test/models/vgg-a.h:35

↓ 9 callersFunctionconv2_relu

* VGG model A conv2 ReLU layer: * channels = 128 * image size = 224x224 */

test/models/vgg-a.h:65

↓ 9 callersFunctionconv3_relu

* VGG model A conv3 ReLU layer: * channels = 256 * image size = 56x56 */

test/models/vgg-a.h:95

↓ 9 callersFunctionfc6

* VGG model A fc6 layer: * input channels = 25088 * output channels = 4096 */

test/models/vgg-a.h:203

↓ 9 callersFunctionfc6_relu

* VGG model A fc6 ReLU layer: * channels = 4096 */

test/models/vgg-a.h:214

↓ 9 callersFunctionfc7

* VGG model A fc7 layer: * input channels = 4096 * output channels = 4096 */

test/models/vgg-a.h:225

↓ 9 callersFunctionfc8

* VGG model A fc8 layer: * input channels = 4096 * output channels = 1000 */

test/models/vgg-a.h:237

↓ 9 callersFunctionfc8_relu

* VGG model A fc8 ReLU layer: * channels = 1000 */

test/models/vgg-a.h:248

↓ 9 callersFunctiongrad_relu

include/nnpack/activations.h:10

↓ 9 callersMethodtestInferenceF16F32

test/testers/fully-connected.h:181

↓ 9 callersMethodtestInferenceF32

test/testers/fully-connected.h:147

↓ 8 callersFunctionpsimd_cmul_soa_f32

src/psimd/fft/soa.h:8

↓ 8 callersFunctionpsimd_cmulc_soa_f32

src/psimd/fft/soa.h:23

↓ 8 callersFunctionpsimd_relu_f32

include/nnpack/activations.h:15

next →1–100 of 873, ranked by callers