| /external/XNNPACK/src/f32-velu/gen/ |
| D | velu-wasmsimd-x86-rr2-p6-x8.c | 56 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x8() local
68 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x8()
89 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x8()
90 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x8()
97 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x8()
|
| D | velu-wasmsimd-x86-rr2-lut16-p3-x8.c | 77 v128_t vs0123 = wasm_i32x4_add(vl0123, ven0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x8() local
87 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x8()
99 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x8()
100 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x8()
107 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x8()
|
| D | velu-wasmsimd-x86-rr2-p6-x12.c | 59 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x12() local
75 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x12()
104 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x12()
105 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x12()
115 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x12()
|
| D | velu-neonfma-rr1-p6-x8.c | 54 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x8() local
77 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x8()
78 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x8()
85 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x8()
|
| D | velu-neon-rr2-p6-x8.c | 55 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neon_rr2_p6_x8() local
81 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neon_rr2_p6_x8()
82 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neon_rr2_p6_x8()
89 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neon_rr2_p6_x8()
|
| D | velu-wasmsimd-x86-rr2-lut16-p3-x12.c | 89 v128_t vs0123 = wasm_i32x4_add(vl0123, ven0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x12() local
103 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x12()
120 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x12()
121 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x12()
131 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x12()
|
| D | velu-wasmsimd-x86-rr2-p6-x16.c | 62 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x16() local
82 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x16()
119 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x16()
120 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x16()
133 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x16()
|
| D | velu-sse41-rr2-p6-x8.c | 56 __m128 vs0123 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(vn0123), 23)); in xnn_f32_velu_ukernel__sse41_rr2_p6_x8() local
83 vt0123 = _mm_mul_ps(vt0123, vs0123); in xnn_f32_velu_ukernel__sse41_rr2_p6_x8()
84 vs0123 = _mm_sub_ps(vs0123, vone); in xnn_f32_velu_ukernel__sse41_rr2_p6_x8()
91 const __m128 ve0123 = _mm_mul_ps(_mm_add_ps(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__sse41_rr2_p6_x8()
|
| D | velu-wasmsimd-arm-rr2-p6-x8.c | 56 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x8() local
83 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x8()
84 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x8()
91 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x8()
|
| D | velu-neonfma-rr1-p6-x12.c | 57 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x12() local
88 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x12()
89 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x12()
99 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x12()
|
| D | velu-sse2-rr2-p6-x8.c | 56 __m128 vs0123 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(vn0123), 23)); in xnn_f32_velu_ukernel__sse2_rr2_p6_x8() local
83 vt0123 = _mm_mul_ps(vt0123, vs0123); in xnn_f32_velu_ukernel__sse2_rr2_p6_x8()
84 vs0123 = _mm_sub_ps(vs0123, vone); in xnn_f32_velu_ukernel__sse2_rr2_p6_x8()
91 const __m128 ve0123 = _mm_mul_ps(_mm_add_ps(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__sse2_rr2_p6_x8()
|
| D | velu-wasmsimd-x86-rr2-p6-x20.c | 65 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x20() local
89 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x20()
134 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x20()
135 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x20()
151 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x20()
|
| D | velu-wasmsimd-x86-rr2-lut16-p3-x16.c | 101 v128_t vs0123 = wasm_i32x4_add(vl0123, ven0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x16() local
119 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x16()
141 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x16()
142 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x16()
155 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x16()
|
| D | velu-neon-rr2-p6-x12.c | 58 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neon_rr2_p6_x12() local
93 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neon_rr2_p6_x12()
94 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neon_rr2_p6_x12()
104 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neon_rr2_p6_x12()
|
| D | velu-wasmsimd-arm-rr2-lut16-p3-x8.c | 77 v128_t vs0123 = wasm_i32x4_add(vl0123, ven0123); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x8() local
93 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x8()
94 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x8()
101 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x8()
|
| D | velu-wasmsimd-arm-rr2-p6-x12.c | 59 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x12() local
95 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x12()
96 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x12()
106 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x12()
|
| D | velu-sse2-rr2-p6-x12.c | 59 __m128 vs0123 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(vn0123), 23)); in xnn_f32_velu_ukernel__sse2_rr2_p6_x12() local
95 vt0123 = _mm_mul_ps(vt0123, vs0123); in xnn_f32_velu_ukernel__sse2_rr2_p6_x12()
96 vs0123 = _mm_sub_ps(vs0123, vone); in xnn_f32_velu_ukernel__sse2_rr2_p6_x12()
106 const __m128 ve0123 = _mm_mul_ps(_mm_add_ps(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__sse2_rr2_p6_x12()
|
| D | velu-sse41-rr2-p6-x12.c | 59 __m128 vs0123 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(vn0123), 23)); in xnn_f32_velu_ukernel__sse41_rr2_p6_x12() local
95 vt0123 = _mm_mul_ps(vt0123, vs0123); in xnn_f32_velu_ukernel__sse41_rr2_p6_x12()
96 vs0123 = _mm_sub_ps(vs0123, vone); in xnn_f32_velu_ukernel__sse41_rr2_p6_x12()
106 const __m128 ve0123 = _mm_mul_ps(_mm_add_ps(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__sse41_rr2_p6_x12()
|
| D | velu-neonfma-rr1-lut16-p3-x8.c | 75 float32x4_t vs0123 = vreinterpretq_f32_s32(vaddq_s32(vl0123, ven0123)); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8() local
88 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8()
89 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8()
96 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8()
|
| D | velu-wasmsimd-x86-rr2-p6-x24.c | 68 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x24() local
96 vs0123 = wasm_v128_andnot(vs0123, vsatm0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x24()
149 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x24()
150 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x24()
169 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_p6_x24()
|
| D | velu-neonfma-rr1-p6-x16.c | 60 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x16() local
99 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x16()
100 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x16()
113 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neonfma_rr1_p6_x16()
|
| D | velu-wasmsimd-arm-rr2-p6-x16.c | 62 v128_t vs0123 = wasm_i32x4_shl(vn0123, 23); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x16() local
107 vt0123 = wasm_f32x4_mul(vt0123, vs0123); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x16()
108 vs0123 = wasm_f32x4_sub(vs0123, vone); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x16()
121 const v128_t ve0123 = wasm_f32x4_mul(wasm_f32x4_add(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_p6_x16()
|
| D | velu-sse41-rr2-p6-x16.c | 62 __m128 vs0123 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(vn0123), 23)); in xnn_f32_velu_ukernel__sse41_rr2_p6_x16() local
107 vt0123 = _mm_mul_ps(vt0123, vs0123); in xnn_f32_velu_ukernel__sse41_rr2_p6_x16()
108 vs0123 = _mm_sub_ps(vs0123, vone); in xnn_f32_velu_ukernel__sse41_rr2_p6_x16()
121 const __m128 ve0123 = _mm_mul_ps(_mm_add_ps(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__sse41_rr2_p6_x16()
|
| D | velu-neon-rr2-lut16-p3-x8.c | 76 float32x4_t vs0123 = vreinterpretq_f32_s32(vaddq_s32(vl0123, ven0123)); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8() local
92 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8()
93 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8()
100 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8()
|
| D | velu-neon-rr2-p6-x16.c | 61 float32x4_t vs0123 = vreinterpretq_f32_s32(vshlq_n_s32(vreinterpretq_s32_f32(vn0123), 23)); in xnn_f32_velu_ukernel__neon_rr2_p6_x16() local
105 vt0123 = vmulq_f32(vt0123, vs0123); in xnn_f32_velu_ukernel__neon_rr2_p6_x16()
106 vs0123 = vsubq_f32(vs0123, vone); in xnn_f32_velu_ukernel__neon_rr2_p6_x16()
119 const float32x4_t ve0123 = vmulq_f32(vaddq_f32(vp0123, vs0123), valpha); in xnn_f32_velu_ukernel__neon_rr2_p6_x16()
|