Searched refs:MS_DIVQ_F32 (Results 1 – 10 of 10) sorted by relevance
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
D | instance_norm_fp32.h | 33 in1 = MS_DIVQ_F32(in1, v); \ 34 in2 = MS_DIVQ_F32(in2, v); \ 35 in3 = MS_DIVQ_F32(in3, v); \ 36 in4 = MS_DIVQ_F32(in4, v);
|
D | instance_norm_fp32.c | 144 deno = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno)); in InstanceNormC4HW4ArmSse() 145 deno1 = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno1)); in InstanceNormC4HW4ArmSse() 146 deno2 = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno2)); in InstanceNormC4HW4ArmSse() 147 deno3 = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno3)); in InstanceNormC4HW4ArmSse() 184 deno = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno)); in InstanceNormC4HW4ArmSse() 185 deno1 = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno1)); in InstanceNormC4HW4ArmSse() 207 mean = MS_DIVQ_F32(mean, hw_planev), squ_m = MS_DIVQ_F32(squ_m, hw_planev); in InstanceNormC4HW4ArmSse() 209 deno = MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_SQRTFX4_F32(deno)); in InstanceNormC4HW4ArmSse()
|
D | exp_fp32.h | 41 MS_INT32X4 integer = MS_CVTQPS_EPI32(MS_DIVQ_F32(input, param[0])); in VexpFp32()
|
D | power_fp32.c | 34 return MS_DIVQ_F32(MS_MOVQ_F32(1), result); in OptimizedPowerSimd()
|
D | activation_fp32.c | 116 …MS_STQ_F32(dst + i, MS_DIVQ_F32(MS_MOVQ_F32(1.0f), MS_ADDQ_F32(MS_MOVQ_F32(1.0f), MS_LDQ_F32(dst +… in Sigmoid() 275 …es = MS_MULQ_F32(MS_MULQ_F32(para3, in), MS_ADDQ_F32(para2, MS_ERFX4_F32(MS_DIVQ_F32(in, para1)))); in Gelu()
|
D | pooling_fp32.c | 104 tmp_avg = MS_DIVQ_F32(tmp_avg, MS_MOVQ_F32(real_count)); in AvgPooling()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16/ |
D | instance_norm_fp16.c | 131 …vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(mean1, hw_plane_4)), vcvt_f16_f32(MS_DIVQ_F32(mean2, hw_plan… in InstanceNormNC8HW8Fp16() 133 …vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(mean3, hw_plane_4)), vcvt_f16_f32(MS_DIVQ_F32(mean4, hw_plan… in InstanceNormNC8HW8Fp16() 134 float16x8_t square_mean = vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(square_mean1, hw_plane_4)), in InstanceNormNC8HW8Fp16() 135 vcvt_f16_f32(MS_DIVQ_F32(square_mean2, hw_plane_4))); in InstanceNormNC8HW8Fp16() 136 float16x8_t square_mean_1 = vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(square_mean3, hw_plane_4)), in InstanceNormNC8HW8Fp16() 137 vcvt_f16_f32(MS_DIVQ_F32(square_mean4, hw_plane_4))); in InstanceNormNC8HW8Fp16() 177 …vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(mean1, hw_plane_4)), vcvt_f16_f32(MS_DIVQ_F32(mean2, hw_plan… in InstanceNormNC8HW8Fp16() 178 float16x8_t square_mean = vcombine_f16(vcvt_f16_f32(MS_DIVQ_F32(square_mean1, hw_plane_4)), in InstanceNormNC8HW8Fp16() 179 vcvt_f16_f32(MS_DIVQ_F32(square_mean2, hw_plane_4))); in InstanceNormNC8HW8Fp16()
|
D | activation_fp16.c | 81 …vst1_f16(dst + i, vcvt_f16_f32(MS_DIVQ_F32(vdupq_n_f32(1.0f), vaddq_f32(vdupq_n_f32(1.0f), tmp)))); in SigmoidFp16() 126 vst1_f16(dst + i, vcvt_f16_f32(vminq_f32(vmaxq_f32(MS_DIVQ_F32(a, b), neg_one), pos_one))); in TanhFp16()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/intrinsics/ |
D | ms_simd_instructions.h | 66 #define MS_DIVQ_F32(src1, src2) vdivq_f32(src1, src2) macro 74 #define MS_DIVQ_F32(src1, src2) vmulq_f32(src1, vrecp(src2)) macro 142 #define MS_DIVQ_F32(src1, src2) _mm_div_ps(src1, src2) macro 201 return MS_MINQ_F32(MS_MAXQ_F32(MS_DIVQ_F32(a, b), neg), pos); in MS_TANHX4_F32()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32_grad/ |
D | activation_grad.c | 128 …MS_STQ_F32(dst + i, MS_DIVQ_F32(MS_LDQ_F32(src0 + i), MS_ADDQ_F32(MS_MOVQ_F32(1.0f), MS_LDQ_F32(ds… in SoftplusGrad()
|