/external/XNNPACK/src/qs8-gavgpool/gen/ |
D | 7x-minmax-neon-c16-acc2.c | 106 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() local 124 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 125 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 142 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2() 143 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c16_acc2()
|
D | 7x-minmax-sse2-c16-acc2.c | 124 const __m128i vsgnaccCDEF = _mm_cmpgt_epi32(_mm_setzero_si128(), vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c16_acc2() local 129 const __m128i vabsaccCDEF = _mm_sub_epi32(_mm_xor_si128(vaccCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c16_acc2() 171 const __m128i voutCDEF = _mm_sub_epi32(_mm_xor_si128(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c16_acc2()
|
D | 7x-minmax-sse2-c24-acc2.c | 147 const __m128i vsgnaccCDEF = _mm_cmpgt_epi32(_mm_setzero_si128(), vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c24_acc2() local 154 const __m128i vabsaccCDEF = _mm_sub_epi32(_mm_xor_si128(vaccCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c24_acc2() 214 const __m128i voutCDEF = _mm_sub_epi32(_mm_xor_si128(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__sse2_c24_acc2()
|
D | 7x-minmax-neon-c24-acc2.c | 121 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() local 145 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 146 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 171 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2() 172 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c24_acc2()
|
D | 7p7x-minmax-neon-c16-acc2.c | 226 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() local 244 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 245 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 262 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2() 263 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c16_acc2()
|
D | 7x-minmax-wasmsimd-c16-acc2.c | 113 const v128_t vsgnaccCDEF = wasm_i32x4_gt(vabsaccCDEF, vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__wasmsimd_c16_acc2() local 150 const v128_t voutCDEF = wasm_i32x4_sub(wasm_v128_xor(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__wasmsimd_c16_acc2()
|
D | 7x-minmax-neon-c32-acc2.c | 136 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() local 166 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 167 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 200 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2() 201 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7x__neon_c32_acc2()
|
D | 7x-minmax-wasmsimd-c24-acc2.c | 130 const v128_t vsgnaccCDEF = wasm_i32x4_gt(vabsaccCDEF, vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__wasmsimd_c24_acc2() local 183 const v128_t voutCDEF = wasm_i32x4_sub(wasm_v128_xor(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7x__wasmsimd_c24_acc2()
|
D | 7p7x-minmax-sse2-c16-acc2.c | 285 const __m128i vsgnaccCDEF = _mm_cmpgt_epi32(_mm_setzero_si128(), vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c16_acc2() local 290 const __m128i vabsaccCDEF = _mm_sub_epi32(_mm_xor_si128(vaccCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c16_acc2() 332 const __m128i voutCDEF = _mm_sub_epi32(_mm_xor_si128(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c16_acc2()
|
D | 7p7x-minmax-neon-c24-acc2.c | 340 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() local 364 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 365 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 390 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2() 391 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c24_acc2()
|
D | 7p7x-minmax-wasmsimd-c16-acc2.c | 240 const v128_t vsgnaccCDEF = wasm_i32x4_gt(vabsaccCDEF, vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__wasmsimd_c16_acc2() local 277 const v128_t voutCDEF = wasm_i32x4_sub(wasm_v128_xor(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__wasmsimd_c16_acc2()
|
D | 7p7x-minmax-sse2-c24-acc2.c | 450 const __m128i vsgnaccCDEF = _mm_cmpgt_epi32(_mm_setzero_si128(), vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c24_acc2() local 457 const __m128i vabsaccCDEF = _mm_sub_epi32(_mm_xor_si128(vaccCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c24_acc2() 517 const __m128i voutCDEF = _mm_sub_epi32(_mm_xor_si128(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__sse2_c24_acc2()
|
D | 7p7x-minmax-neon-c32-acc2.c | 393 const int32x4_t vsgnaccCDEF = vreinterpretq_s32_u32(vcltq_s32(vaccCDEF, vmovq_n_s32(0))); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() local 423 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 424 const int64x2_t vadjprodEF = vaddw_high_s32(vprodEF, vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 457 const int64x2_t vadjprodCD = vaddw_s32(vprodCD, vget_low_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2() 458 const int64x2_t vadjprodEF = vaddw_s32(vprodEF, vget_high_s32(vsgnaccCDEF)); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__neon_c32_acc2()
|
D | 7p7x-minmax-wasmsimd-c24-acc2.c | 365 const v128_t vsgnaccCDEF = wasm_i32x4_gt(vabsaccCDEF, vaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__wasmsimd_c24_acc2() local 418 const v128_t voutCDEF = wasm_i32x4_sub(wasm_v128_xor(vabsoutCDEF, vsgnaccCDEF), vsgnaccCDEF); in xnn_qs8_gavgpool_minmax_ukernel_7p7x__wasmsimd_c24_acc2()
|