| /external/XNNPACK/src/f32-hswish/gen/ |
| D | hswish-neon-x16.c | 47 vacc0123 = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc0123), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
48 vacc4567 = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc4567), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
49 vacc89AB = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc89AB), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
50 vaccCDEF = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vaccCDEF), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
71 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
80 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x16()
|
| D | hswish-neon-x8.c | 41 vacc0123 = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc0123), vzero)); in xnn_f32_hswish_ukernel__neon_x8()
42 vacc4567 = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc4567), vzero)); in xnn_f32_hswish_ukernel__neon_x8()
57 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x8()
66 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x8()
|
| D | hswish-neon-x4.c | 36 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x4()
45 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero)); in xnn_f32_hswish_ukernel__neon_x4()
|
| /external/XNNPACK/src/f32-hswish/ |
| D | neon.c.in | 41 …vacc${ABC[N:N+4]} = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc${ABC[N:N+4]}), vzer…
56 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero));
65 vacc = vreinterpretq_f32_s32(vmaxq_s32(vreinterpretq_s32_f32(vacc), vzero));
|
| /external/tensorflow/tensorflow/lite/kernels/internal/optimized/integer_ops/ |
| D | mean.h | 100 temp_sum.val[0] = vminq_s32(vmaxq_s32(temp_sum.val[0], min_dup), max_dup); in MeanImpl()
101 temp_sum.val[1] = vminq_s32(vmaxq_s32(temp_sum.val[1], min_dup), max_dup); in MeanImpl()
102 temp_sum.val[2] = vminq_s32(vmaxq_s32(temp_sum.val[2], min_dup), max_dup); in MeanImpl()
103 temp_sum.val[3] = vminq_s32(vmaxq_s32(temp_sum.val[3], min_dup), max_dup); in MeanImpl()
|
| /external/libaom/libaom/av1/common/arm/ |
| D | convolve_neon.h | 55 sum0 = vmaxq_s32(sum0, zero); in wiener_convolve8_vert_4x8()
56 sum1 = vmaxq_s32(sum1, zero); in wiener_convolve8_vert_4x8()
152 sum_0 = vmaxq_s32(sum_0, zero); in wiener_convolve8_horiz_4x8()
222 sum0 = vmaxq_s32(sum0, zero); in convolve8_4x4_s32()
|
| D | convolve_neon.c | 143 sum0 = vmaxq_s32(sum0, zero); in convolve8_vert_4x4_s32()
184 sum0 = vmaxq_s32(sum0, zero); in convolve8_vert_8x4_s32()
185 sum1 = vmaxq_s32(sum1, zero); in convolve8_vert_8x4_s32()
|
| /external/webrtc/common_audio/signal_processing/ |
| D | min_max_operations_neon.c | 181 max32x4_0 = vmaxq_s32(max32x4_0, in32x4_0); in WebRtcSpl_MaxValueW32Neon()
182 max32x4_1 = vmaxq_s32(max32x4_1, in32x4_1); in WebRtcSpl_MaxValueW32Neon()
185 int32x4_t max32x4 = vmaxq_s32(max32x4_0, max32x4_1); in WebRtcSpl_MaxValueW32Neon()
|
| /external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
| D | neon_tensor_utils.cc | 884 vmaxq_s32(vminq_s32(temp_val.val[0], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
886 vmaxq_s32(vminq_s32(temp_val.val[1], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
949 vmaxq_s32(vminq_s32(temp_val.val[0], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
951 vmaxq_s32(vminq_s32(temp_val.val[1], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
953 vmaxq_s32(vminq_s32(temp_val.val[2], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
955 vmaxq_s32(vminq_s32(temp_val.val[3], max_val_dup), min_val_dup); in NeonMatrixBatchVectorAccumulateImpl()
1787 vmaxq_s32(vminq_s32(temp_val.val[0], max_val_dup), min_val_dup); in NeonCwiseMul()
1789 vmaxq_s32(vminq_s32(temp_val.val[1], max_val_dup), min_val_dup); in NeonCwiseMul()
2338 int32x4_t max0_i32x4 = vmaxq_s32(f2i0_i32x4, neg_scale_i32x4); in NeonSymmetricQuantizeFloats()
2339 int32x4_t max1_i32x4 = vmaxq_s32(f2i1_i32x4, neg_scale_i32x4); in NeonSymmetricQuantizeFloats()
[all …]
|
| D | optimized_ops.h | 1046 temp_sum.val[0] = vminq_s32(vmaxq_s32(temp_sum.val[0], min_dup), max_dup); in MeanImpl()
1047 temp_sum.val[1] = vminq_s32(vmaxq_s32(temp_sum.val[1], min_dup), max_dup); in MeanImpl()
1048 temp_sum.val[2] = vminq_s32(vmaxq_s32(temp_sum.val[2], min_dup), max_dup); in MeanImpl()
1049 temp_sum.val[3] = vminq_s32(vmaxq_s32(temp_sum.val[3], min_dup), max_dup); in MeanImpl()
4041 vmaxq_s32(vminq_s32(temp_val.val[0], max_val_dup), min_val_dup); in SoftmaxInt8LUT()
4043 vmaxq_s32(vminq_s32(temp_val.val[1], max_val_dup), min_val_dup); in SoftmaxInt8LUT()
4045 vmaxq_s32(vminq_s32(temp_val.val[2], max_val_dup), min_val_dup); in SoftmaxInt8LUT()
4047 vmaxq_s32(vminq_s32(temp_val.val[3], max_val_dup), min_val_dup); in SoftmaxInt8LUT()
5813 vmaxq_s32(vminq_s32(temp_val.val[0], max_val_dup), min_val_dup); in Quantize()
5815 vmaxq_s32(vminq_s32(temp_val.val[1], max_val_dup), min_val_dup); in Quantize()
[all …]
|
| D | depthwiseconv_uint8.h | 1885 acc[j] = vmaxq_s32(acc[j], output_activation_min_vec);
1928 acc0 = vmaxq_s32(acc0, output_activation_min_vec);
1929 acc1 = vmaxq_s32(acc1, output_activation_min_vec);
1960 acc = vmaxq_s32(acc, output_activation_min_vec);
|
| /external/llvm-project/clang/test/CodeGen/arm-mve-intrinsics/ |
| D | vmaxq.c | 48 return vmaxq_s32(a, b); in test_vmaxq_s32()
|
| /external/XNNPACK/src/math/ |
| D | exp-neonfma-rr2-p5.c | 61 int32x4_t ven = vmaxq_s32(veo, vmin_exponent); in xnn_math_f32_exp__neonfma_rr2_p5()
|
| D | exp-neonfma-rr2-lut64-p2.c | 63 int32x4_t ven = vmaxq_s32(veo, vmin_exponent); in xnn_math_f32_exp__neonfma_rr2_lut64_p2()
|
| /external/libopus/silk/arm/ |
| D | LPC_inv_pred_gain_neon_intr.c | 128 max_s32x4 = vmaxq_s32( max_s32x4, s0_s32x4 ); in LPC_inverse_pred_gain_QA_neon()
130 max_s32x4 = vmaxq_s32( max_s32x4, s1_s32x4 ); in LPC_inverse_pred_gain_QA_neon()
|
| D | NSQ_del_dec_neon_intr.c | 697 tmp1_s32x4 = vmaxq_s32( tmp1_s32x4, vdupq_n_s32( -( 31 << 10 ) ) ); in silk_noise_shape_quantizer_del_dec_neon()
755 tmp2_s32x4 = vaddq_s32( tmp2_s32x4, vmaxq_s32( rd1_Q10_s32x4, rd2_Q10_s32x4 ) ); in silk_noise_shape_quantizer_del_dec_neon()
|
| /external/gemmlowp/internal/ |
| D | simd_wrappers_neon.h | 102 inline Int32x4 Max(Int32x4 a, Int32x4 b) { return vmaxq_s32(a, b); }
105 return vmaxq_s32(a, vdupq_n_s32(b));
|
| /external/skqp/src/core/ |
| D | SkBitmapProcState_matrixProcs.cpp | 325 res = vmaxq_s32(res, vdupq_n_s32(0)); in clamp4()
|
| /external/tensorflow/tensorflow/lite/kernels/ |
| D | cpu_backend_gemm_custom_gemv.h | 571 vmaxq_s32(multiplier_exponent, vdupq_n_s32(0));
|
| /external/eigen/Eigen/src/Core/arch/NEON/ |
| D | PacketMath.h | 219 …ONG_INLINE Packet4i pmax<Packet4i>(const Packet4i& a, const Packet4i& b) { return vmaxq_s32(a,b); }
|
| /external/skia/include/private/ |
| D | SkNx_neon.h | 541 AI static SkNx Max(const SkNx& a, const SkNx& b) { return vmaxq_s32(a.fVec, b.fVec); } in Max()
|
| /external/skqp/include/private/ |
| D | SkNx_neon.h | 563 AI static SkNx Max(const SkNx& a, const SkNx& b) { return vmaxq_s32(a.fVec, b.fVec); } in Max()
|
| /external/libaom/libaom/aom_dsp/simd/ |
| D | v128_intrinsics_arm.h | 412 vmaxq_s32(vreinterpretq_s32_s64(x), vreinterpretq_s32_s64(y))); in v128_max_s32()
|
| /external/libgav1/libgav1/src/dsp/arm/ |
| D | inverse_transform_10bit_neon.cc | 181 *a = vmaxq_s32(vminq_s32(x, *max), *min); in HadamardRotation()
182 *b = vmaxq_s32(vminq_s32(y, *max), *min); in HadamardRotation()
|
| /external/psimd/include/ |
| D | psimd.h | 839 return (psimd_s32) vmaxq_s32((int32x4_t) a, (int32x4_t) b); in psimd_max_s32()
|