/external/XNNPACK/src/qs8-gemm/gen/ |
D | 1x4c8-xw-minmax-ssse3.c | 83 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3() local 85 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3() 87 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__ssse3()
|
D | 1x4c8-minmax-ssse3-ld128.c | 85 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128() local 87 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128() 89 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld128()
|
D | 1x4c8-xw-minmax-sse2.c | 83 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() local 85 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() 87 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
|
D | 1x4c8-minmax-ssse3-ld64.c | 87 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64() local 89 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64() 91 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c8__ssse3_ld64()
|
D | 1x4c8-minmax-sse2-ld128.c | 85 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() local 87 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() 89 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
|
D | 1x4c8-minmax-sse2-ld64.c | 87 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() local 89 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() 91 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
|
D | 1x4c2-xw-minmax-sse2.c | 106 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2() local 108 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2() 110 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__sse2()
|
D | 1x4c2-xw-minmax-ssse3.c | 106 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3() local 108 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3() 110 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c2__ssse3()
|
D | 1x4c2-minmax-sse2-ld64.c | 113 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64() local 115 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64() 117 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld64()
|
D | 1x4c2-minmax-ssse3-ld64.c | 113 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld64() local 115 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld64() 117 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld64()
|
D | 1x4c2-minmax-ssse3-ld128.c | 113 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld128() local 115 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld128() 117 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c2__ssse3_ld128()
|
D | 1x4c2-minmax-sse2-ld128.c | 113 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128() local 115 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128() 117 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_1x4c2__sse2_ld128()
|
D | 2x4c8-minmax-ssse3-ld128.c | 106 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld128() local 109 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld128() 112 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld128()
|
D | 2x4c8-xw-minmax-sse2.c | 104 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2() local 107 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2() 110 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__sse2()
|
D | 2x4c8-minmax-ssse3-ld64.c | 108 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld64() local 111 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld64() 114 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_minmax_ukernel_2x4c8__ssse3_ld64()
|
D | 2x4c8-xw-minmax-ssse3.c | 104 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__ssse3() local 107 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__ssse3() 110 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__ssse3()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 1x4c8-minmax-ssse3-ld128.c | 98 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128() local 100 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128() 102 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld128()
|
D | 1x4c8-minmax-sse2-ld64.c | 100 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64() local 102 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64() 104 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld64()
|
D | 1x4c8-minmax-sse2-ld128.c | 98 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128() local 100 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128() 102 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c8__sse2_ld128()
|
D | 1x4c8-minmax-ssse3-ld64.c | 100 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld64() local 102 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld64() 104 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c8__ssse3_ld64()
|
D | 1x4c2-minmax-ssse3-ld128.c | 126 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld128() local 128 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld128() 130 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld128()
|
D | 1x4c2-minmax-sse2-ld128.c | 126 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128() local 128 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128() 130 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld128()
|
D | 1x4c2-minmax-ssse3-ld64.c | 126 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld64() local 128 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld64() 130 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c2__ssse3_ld64()
|
D | 1x4c2-minmax-sse2-ld64.c | 126 …const __m128i vabsacc0x0123 = _mm_sub_epi32(_mm_xor_si128(vacc0x0123, vnmask0x0123), vnmask0x0123); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64() local 128 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64() 130 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_1x4c2__sse2_ld64()
|
D | 2x4c8-minmax-ssse3-ld128.c | 121 const __m128i vabsacc0x0123 = _mm_abs_epi32(vacc0x0123); in xnn_qs8_igemm_minmax_ukernel_2x4c8__ssse3_ld128() local 124 const __m128i vabsacc0x1133 = _mm_shuffle_epi32(vabsacc0x0123, _MM_SHUFFLE(3, 3, 1, 1)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__ssse3_ld128() 127 const __m128i vabsprod0x02 = _mm_mul_epu32(vabsacc0x0123, vmultiplier); in xnn_qs8_igemm_minmax_ukernel_2x4c8__ssse3_ld128()
|