/external/XNNPACK/src/qs8-gemm/gen/ |
D | 1x4c8-xw-minmax-sse2.c | 81 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() local 83 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() 89 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() 97 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
|
D | 1x4c8-minmax-sse2-ld128.c | 83 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() local 85 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() 91 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() 99 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
|
D | 1x4c8-minmax-sse2-ld64.c | 85 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() local 87 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() 93 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() 101 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
|
D | 1x4c2-xw-minmax-sse2.c | 104 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2() local 106 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2() 112 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2() 120 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2()
|
D | 1x4c2-minmax-sse2-ld64.c | 111 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64() local 113 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64() 119 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64() 127 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64()
|
D | 1x4c2-minmax-sse2-ld128.c | 111 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128() local 113 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128() 119 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128() 127 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128()
|
D | 2x4c8-xw-minmax-sse2.c | 101 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2() local 104 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2() 113 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2() 125 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2()
|
D | 2x4c8-minmax-sse2-ld128.c | 103 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld128() local 106 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld128() 115 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld128() 127 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld128()
|
D | 2x4c8-minmax-sse2-ld64.c | 105 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld64() local 108 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld64() 117 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld64() 129 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__sse2_ld64()
|
D | 1x4c8-xw-minmax-ssse3.c | 81 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3() local 89 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3() 97 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3()
|
D | 1x4c8-minmax-ssse3-ld128.c | 83 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128() local 91 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128() 99 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128()
|
D | 1x4c8-minmax-ssse3-ld64.c | 85 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64() local 93 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64() 101 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64()
|
D | 3x4c8-minmax-sse2-ld64.c | 125 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld64() local 129 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld64() 141 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld64() 157 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld64()
|
D | 3x4c8-minmax-sse2-ld128.c | 123 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld128() local 127 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld128() 139 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld128() 155 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__sse2_ld128()
|
D | 3x4c8-xw-minmax-sse2.c | 121 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__sse2() local 125 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__sse2() 137 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__sse2() 153 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__sse2()
|
D | 1x4c2-xw-minmax-ssse3.c | 104 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3() local 112 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3() 120 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 1x4c8-minmax-sse2-ld64.c | 98 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64() local 100 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64() 106 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64() 114 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64()
|
D | 1x4c8-minmax-sse2-ld128.c | 96 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128() local 98 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128() 104 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128() 112 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128()
|
D | 1x4c2-minmax-sse2-ld128.c | 124 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128() local 126 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128() 132 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128() 140 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128()
|
D | 1x4c2-minmax-sse2-ld64.c | 124 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64() local 126 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64() 132 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64() 140 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64()
|
D | 2x4c8-minmax-sse2-ld128.c | 118 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld128() local 121 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld128() 130 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld128() 142 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld128()
|
D | 2x4c8-minmax-sse2-ld64.c | 120 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld64() local 123 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld64() 132 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld64() 144 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__sse2_ld64()
|
D | 1x4c8-minmax-ssse3-ld128.c | 96 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128() local 104 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128() 112 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128()
|
D | 3x4c8-minmax-sse2-ld128.c | 140 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_3x4c8__sse2_ld128() local 144 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_3x4c8__sse2_ld128() 156 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__sse2_ld128() 172 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__sse2_ld128()
|
/external/XNNPACK/src/qu8-gemm/ |
D | 2x4c8-minmax-sse2.c | 114 const __m128i vnmask0x0123 = _mm_cmpgt_epi32(_mm_setzero_si128(), vacc0x0123); in xnn_qu8_gemm_minmax_ukernel_2x4c8__sse2() local 117 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qu8_gemm_minmax_ukernel_2x4c8__sse2() 126 const __m128i vnmask0x02 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(2, 2, 0, 0)); in xnn_qu8_gemm_minmax_ukernel_2x4c8__sse2() 138 const __m128i vnmask0x13 = _mm_shuffle_epi32(vnmask0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qu8_gemm_minmax_ukernel_2x4c8__sse2()
|