Searched refs:MS_MAXQ_F32 (Results 1 – 9 of 9) sorted by relevance
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
| D | winograd_utils.c | 407 m[l] = MS_MAXQ_F32(zero, m[l]); in OutputTransform4x2ReluUnit()
408 m[l + 2] = MS_MAXQ_F32(zero, m[l + 2]); in OutputTransform4x2ReluUnit()
475 m[l] = MS_MAXQ_F32(zero, m[l]); in OutputTransform4x2Relu6Unit()
477 m[l + 2] = MS_MAXQ_F32(zero, m[l + 2]); in OutputTransform4x2Relu6Unit()
617 m[l] = MS_MAXQ_F32(zero, m[l]); in OutputTransform4x3ReluUnit()
618 m[l + 3] = MS_MAXQ_F32(zero, m[l + 3]); in OutputTransform4x3ReluUnit()
619 m[l + 6] = MS_MAXQ_F32(zero, m[l + 6]); in OutputTransform4x3ReluUnit()
692 m[l] = MS_MAXQ_F32(zero, m[l]); in OutputTransform4x3Relu6Unit()
694 m[l + 3] = MS_MAXQ_F32(zero, m[l + 3]); in OutputTransform4x3Relu6Unit()
696 m[l + 6] = MS_MAXQ_F32(zero, m[l + 6]); in OutputTransform4x3Relu6Unit()
[all …]
|
| D | add_fp32.c | 143 MS_FLOAT32X4 vout = MS_MAXQ_F32(MS_ADDQ_F32(vin0_opt, vin1), zeros); in ElementOptAddRelu()
161 MS_FLOAT32X4 vout = MS_MAXQ_F32(MS_ADDQ_F32(vin0, vin1_opt), zeros); in ElementOptAddRelu()
197 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_ADDQ_F32(vin0_opt, vin1), zeros), bounds); in ElementOptAddRelu6()
215 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_ADDQ_F32(vin0, vin1_opt), zeros), bounds); in ElementOptAddRelu6()
304 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_ADDQ_F32(vin0, vin1), zeros), bounds); in ElementAddRelu6()
|
| D | scale_fp32.c | 136 MS_FLOAT32X4 result = MS_MAXQ_F32(tmp, zeros); in ScaleInnerRelu()
178 MS_FLOAT32X4 result = MS_MAXQ_F32(tmp, zeros); in ScaleAxisRelu()
240 MS_FLOAT32X4 result = MS_MINQ_F32(MS_MAXQ_F32(tmp, zeros), bounds); in ScaleInnerRelu6()
284 MS_FLOAT32X4 result = MS_MINQ_F32(MS_MAXQ_F32(tmp, zeros), bounds); in ScaleAxisRelu6()
|
| D | pooling_fp32.c | 105 tmp_avg = MS_MAXQ_F32(tmp_avg, min_value); in AvgPooling()
205 tmp_max = MS_MAXQ_F32(tmp_max, MS_LDQ_F32(src_win_ptr)); in MaxPooling()
208 tmp_max = MS_MAXQ_F32(tmp_max, min_value); in MaxPooling()
|
| D | mul_fp32.c | 96 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_MULQ_F32(vin0, vin1), zeros), bounds); in ElementMulRelu6()
250 MS_FLOAT32X4 vout = MS_MAXQ_F32(MS_MULQ_F32(vin0_opt, vin1), zeros); in ElementOptMulRelu()
272 MS_FLOAT32X4 vout = MS_MAXQ_F32(MS_MULQ_F32(vin0, vin1_opt), zeros); in ElementOptMulRelu()
302 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_MULQ_F32(vin0_opt, vin1), zeros), bounds); in ElementOptMulRelu6()
326 MS_FLOAT32X4 vout = MS_MINQ_F32(MS_MAXQ_F32(MS_MULQ_F32(vin0, vin1_opt), zeros), bounds); in ElementOptMulRelu6()
|
| D | exp_fp32.h | 53 input = MS_MAXQ_F32(minv, MS_MINQ_F32(input, maxv)); in simd_exp()
|
| D | activation_fp32.c | 33 MS_STQ_F32(dst + i, MS_MAXQ_F32(MS_LDQ_F32(src + i), zero)); in Fp32Relu()
59 MS_FLOAT32X4 dst_tmp = MS_MAXQ_F32(MS_LDQ_F32(src + i), zero); in Fp32Relu6()
|
| D | conv_depthwise_fp32.c | 621 res0 = MS_MAXQ_F32(res0, MS_MOVQ_F32(0.0f)); in ConvDw3x3Line()
622 res1 = MS_MAXQ_F32(res1, MS_MOVQ_F32(0.0f)); in ConvDw3x3Line()
655 res0 = MS_MAXQ_F32(res0, MS_MOVQ_F32(0.0f)); in ConvDw3x3Line()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/intrinsics/ |
| D | ms_simd_instructions.h | 59 #define MS_MAXQ_F32 vmaxq_f32 macro
136 #define MS_MAXQ_F32 _mm_max_ps macro
201 return MS_MINQ_F32(MS_MAXQ_F32(MS_DIVQ_F32(a, b), neg), pos); in MS_TANHX4_F32()
|