/external/XNNPACK/src/f32-gemm/gen-inc/ |
D | 4x4inc-minmax-wasm.c | 74 float vacc23 = acc[11]; in xnn_f32_gemminc_minmax_ukernel_4x4__wasm() local 105 vacc23 += va2 * vb3; in xnn_f32_gemminc_minmax_ukernel_4x4__wasm() 125 vacc23 = __builtin_wasm_max_f32(vacc23, vmin); in xnn_f32_gemminc_minmax_ukernel_4x4__wasm() 142 vacc23 = __builtin_wasm_min_f32(vacc23, vmax); in xnn_f32_gemminc_minmax_ukernel_4x4__wasm() 157 c2[3] = vacc23; in xnn_f32_gemminc_minmax_ukernel_4x4__wasm()
|
D | 4x4inc-minmax-scalar.c | 74 float vacc23 = acc[11]; in xnn_f32_gemminc_minmax_ukernel_4x4__scalar() local 105 vacc23 += va2 * vb3; in xnn_f32_gemminc_minmax_ukernel_4x4__scalar() 125 vacc23 = math_max_f32(vacc23, vmin); in xnn_f32_gemminc_minmax_ukernel_4x4__scalar() 142 vacc23 = math_min_f32(vacc23, vmax); in xnn_f32_gemminc_minmax_ukernel_4x4__scalar() 157 c2[3] = vacc23; in xnn_f32_gemminc_minmax_ukernel_4x4__scalar()
|
/external/XNNPACK/src/f32-gemm/gen/ |
D | 4x4-minmax-scalar.c | 73 float vacc23 = vacc03; in xnn_f32_gemm_minmax_ukernel_4x4__scalar() local 103 vacc23 += va2 * vb3; in xnn_f32_gemm_minmax_ukernel_4x4__scalar() 123 vacc23 = math_max_f32(vacc23, vmin); in xnn_f32_gemm_minmax_ukernel_4x4__scalar() 140 vacc23 = math_min_f32(vacc23, vmax); in xnn_f32_gemm_minmax_ukernel_4x4__scalar() 155 c2[3] = vacc23; in xnn_f32_gemm_minmax_ukernel_4x4__scalar()
|
D | 4x4-minmax-wasm.c | 73 float vacc23 = vacc03; in xnn_f32_gemm_minmax_ukernel_4x4__wasm() local 103 vacc23 += va2 * vb3; in xnn_f32_gemm_minmax_ukernel_4x4__wasm() 123 vacc23 = __builtin_wasm_max_f32(vacc23, vmin); in xnn_f32_gemm_minmax_ukernel_4x4__wasm() 140 vacc23 = __builtin_wasm_min_f32(vacc23, vmax); in xnn_f32_gemm_minmax_ukernel_4x4__wasm() 155 c2[3] = vacc23; in xnn_f32_gemm_minmax_ukernel_4x4__wasm()
|
D | 4x4-relu-wasm.c | 71 float vacc23 = vacc03; in xnn_f32_gemm_relu_ukernel_4x4__wasm() local 101 vacc23 += va2 * vb3; in xnn_f32_gemm_relu_ukernel_4x4__wasm() 121 vacc23 = __builtin_wasm_max_f32(vacc23, 0.0f); in xnn_f32_gemm_relu_ukernel_4x4__wasm() 136 c2[3] = vacc23; in xnn_f32_gemm_relu_ukernel_4x4__wasm()
|
D | 4x4-relu-scalar.c | 71 float vacc23 = vacc03; in xnn_f32_gemm_relu_ukernel_4x4__scalar() local 101 vacc23 += va2 * vb3; in xnn_f32_gemm_relu_ukernel_4x4__scalar() 121 vacc23 = math_max_f32(vacc23, 0.0f); in xnn_f32_gemm_relu_ukernel_4x4__scalar() 136 c2[3] = vacc23; in xnn_f32_gemm_relu_ukernel_4x4__scalar()
|
D | 4x4-wasm.c | 71 float vacc23 = vacc03; in xnn_f32_gemm_ukernel_4x4__wasm() local 101 vacc23 += va2 * vb3; in xnn_f32_gemm_ukernel_4x4__wasm() 120 c2[3] = vacc23; in xnn_f32_gemm_ukernel_4x4__wasm()
|
D | 4x4-scalar.c | 71 float vacc23 = vacc03; in xnn_f32_gemm_ukernel_4x4__scalar() local 101 vacc23 += va2 * vb3; in xnn_f32_gemm_ukernel_4x4__scalar() 120 c2[3] = vacc23; in xnn_f32_gemm_ukernel_4x4__scalar()
|
/external/XNNPACK/src/f32-igemm/gen/ |
D | 4x4-minmax-wasm.c | 70 float vacc23 = vacc03; in xnn_f32_igemm_minmax_ukernel_4x4__wasm() local 125 vacc23 += va2 * vb3; in xnn_f32_igemm_minmax_ukernel_4x4__wasm() 147 vacc23 = __builtin_wasm_max_f32(vacc23, vmin); in xnn_f32_igemm_minmax_ukernel_4x4__wasm() 164 vacc23 = __builtin_wasm_min_f32(vacc23, vmax); in xnn_f32_igemm_minmax_ukernel_4x4__wasm() 179 c2[3] = vacc23; in xnn_f32_igemm_minmax_ukernel_4x4__wasm()
|
D | 4x4-minmax-scalar.c | 70 float vacc23 = vacc03; in xnn_f32_igemm_minmax_ukernel_4x4__scalar() local 125 vacc23 += va2 * vb3; in xnn_f32_igemm_minmax_ukernel_4x4__scalar() 147 vacc23 = math_max_f32(vacc23, vmin); in xnn_f32_igemm_minmax_ukernel_4x4__scalar() 164 vacc23 = math_min_f32(vacc23, vmax); in xnn_f32_igemm_minmax_ukernel_4x4__scalar() 179 c2[3] = vacc23; in xnn_f32_igemm_minmax_ukernel_4x4__scalar()
|
D | 4x4-relu-scalar.c | 68 float vacc23 = vacc03; in xnn_f32_igemm_relu_ukernel_4x4__scalar() local 123 vacc23 += va2 * vb3; in xnn_f32_igemm_relu_ukernel_4x4__scalar() 145 vacc23 = math_max_f32(vacc23, 0.0f); in xnn_f32_igemm_relu_ukernel_4x4__scalar() 160 c2[3] = vacc23; in xnn_f32_igemm_relu_ukernel_4x4__scalar()
|
D | 4x4-relu-wasm.c | 68 float vacc23 = vacc03; in xnn_f32_igemm_relu_ukernel_4x4__wasm() local 123 vacc23 += va2 * vb3; in xnn_f32_igemm_relu_ukernel_4x4__wasm() 145 vacc23 = __builtin_wasm_max_f32(vacc23, 0.0f); in xnn_f32_igemm_relu_ukernel_4x4__wasm() 160 c2[3] = vacc23; in xnn_f32_igemm_relu_ukernel_4x4__wasm()
|
D | 4x4-scalar.c | 68 float vacc23 = vacc03; in xnn_f32_igemm_ukernel_4x4__scalar() local 123 vacc23 += va2 * vb3; in xnn_f32_igemm_ukernel_4x4__scalar() 144 c2[3] = vacc23; in xnn_f32_igemm_ukernel_4x4__scalar()
|
D | 4x4-wasm.c | 68 float vacc23 = vacc03; in xnn_f32_igemm_ukernel_4x4__wasm() local 123 vacc23 += va2 * vb3; in xnn_f32_igemm_ukernel_4x4__wasm() 144 c2[3] = vacc23; in xnn_f32_igemm_ukernel_4x4__wasm()
|
/external/XNNPACK/src/qs8-gavgpool/gen/ |
D | 7x-minmax-neon-c8-acc2.c | 114 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2() local 119 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2() 126 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2() 191 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2() local 196 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2() 203 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c8_acc2()
|
D | 7x-minmax-neon-c16-acc2.c | 147 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() local 156 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 166 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 234 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() local 239 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 246 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2()
|
D | 7p7x-minmax-neon-c8-acc2.c | 196 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2() local 201 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2() 208 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2() 276 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2() local 281 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2() 288 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c8_acc2()
|
D | 7x-minmax-neon-c24-acc2.c | 180 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() local 193 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 207 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 282 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() local 287 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 294 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2()
|
D | 7p7x-minmax-neon-c16-acc2.c | 267 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() local 276 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 286 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 357 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() local 362 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 369 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2()
|
D | 7x-minmax-neon-c32-acc2.c | 213 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() local 230 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 247 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 325 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() local 330 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 337 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2()
|
D | 7p7x-minmax-neon-c24-acc2.c | 399 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() local 412 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 426 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 504 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() local 509 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 516 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2()
|
D | 7p7x-minmax-neon-c32-acc2.c | 470 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() local 487 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 504 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 585 const int64x2_t vacc23 = vrshlq_s64(vadjprod23, vleft_shift); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() local 590 vacc0123 = vuzp1q_s32(vreinterpretq_s32_s64(vacc01), vreinterpretq_s32_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 597 vacc0123 = vcombine_s32(vmovn_s64(vacc01), vmovn_s64(vacc23)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2()
|
/external/XNNPACK/src/qs8-dwconv/gen/ |
D | up8x9-minmax-wasmsimd-mul16.c | 174 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up8x9__wasmsimd_mul16() local 181 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up8x9__wasmsimd_mul16() 290 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up8x9__wasmsimd_mul16() local 297 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up8x9__wasmsimd_mul16()
|
D | up16x9-minmax-wasmsimd-mul16.c | 223 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16() local 234 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16() 366 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16() local 373 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16()
|
D | up24x9-minmax-wasmsimd-mul16.c | 272 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16() local 287 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16() 432 const v128_t vacc23 = wasm_v32x4_shuffle(vacc0123, vsign0123, 2, 6, 3, 7); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16() local 439 const v128_t vprod23 = wasm_i64x2_add(wasm_i64x2_mul(vacc23, vmultiplier), vrounding); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16()
|