/external/XNNPACK/src/qu8-requantization/ |
D | fp32-sse2.c | 77 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_fp32__sse2() local 90 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_fp32__sse2()
|
D | q31-neon.c | 92 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__neon() local 122 vst1q_u8(output, xyzw_clamped); in xnn_qu8_requantize_q31__neon()
|
D | fp32-neon.c | 77 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_fp32__neon() local 79 vst1q_u8(output, xyzw_clamped); in xnn_qu8_requantize_fp32__neon()
|
D | precise-sse4.c | 96 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__sse4() local 114 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_precise__sse4()
|
D | precise-ssse3.c | 104 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__ssse3() local 121 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_precise__ssse3()
|
D | q31-sse4.c | 112 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__sse4() local 133 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_q31__sse4()
|
D | precise-sse2.c | 109 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__sse2() local 127 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_precise__sse2()
|
D | q31-wasmsimd.c | 115 const v128_t xyzw_clamped = wasm_u8x16_min(wasm_u8x16_max(xyzw_packed, vqmin), vqmax); in xnn_qu8_requantize_q31__wasmsimd() local 134 wasm_v128_store(output, xyzw_clamped); in xnn_qu8_requantize_q31__wasmsimd()
|
D | precise-neon.c | 129 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__neon() local 163 vst1q_u8(output, xyzw_clamped); in xnn_qu8_requantize_precise__neon()
|
D | q31-ssse3.c | 161 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__ssse3() local 185 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_q31__ssse3()
|
D | q31-sse2.c | 161 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__sse2() local 184 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qu8_requantize_q31__sse2()
|
/external/XNNPACK/src/qs8-requantization/ |
D | fp32-sse4.c | 77 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__sse4() local 90 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_fp32__sse4()
|
D | fp32-sse2.c | 78 const __m128i xyzw_clamped = _mm_packs_epi16(xy_clamped, zw_clamped); in xnn_qs8_requantize_fp32__sse2() local 91 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_fp32__sse2()
|
D | q31-neon.c | 92 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__neon() local 122 vst1q_s8(output, xyzw_clamped); in xnn_qs8_requantize_q31__neon()
|
D | fp32-neon.c | 77 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__neon() local 90 vst1q_s8(output, xyzw_clamped); output += 16; in xnn_qs8_requantize_fp32__neon()
|
D | precise-sse4.c | 96 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__sse4() local 114 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_precise__sse4()
|
D | q31-sse4.c | 112 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__sse4() local 133 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_q31__sse4()
|
D | precise-ssse3.c | 105 const __m128i xyzw_clamped = _mm_packs_epi16(xy_clamped, zw_clamped); in xnn_qs8_requantize_precise__ssse3() local 122 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_precise__ssse3()
|
D | q31-wasmsimd.c | 115 const v128_t xyzw_clamped = wasm_i8x16_min(wasm_i8x16_max(xyzw_packed, vqmin), vqmax); in xnn_qs8_requantize_q31__wasmsimd() local 134 wasm_v128_store(output, xyzw_clamped); in xnn_qs8_requantize_q31__wasmsimd()
|
D | precise-sse2.c | 110 const __m128i xyzw_clamped = _mm_packs_epi16(xy_clamped, zw_clamped); in xnn_qs8_requantize_precise__sse2() local 128 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_precise__sse2()
|
D | precise-neon.c | 129 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__neon() local 163 vst1q_s8(output, xyzw_clamped); in xnn_qs8_requantize_precise__neon()
|
D | q31-ssse3.c | 162 const __m128i xyzw_clamped = _mm_packs_epi16(xy_clamped, zw_clamped); in xnn_qs8_requantize_q31__ssse3() local 186 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_q31__ssse3()
|
D | q31-sse2.c | 162 const __m128i xyzw_clamped = _mm_packs_epi16(xy_clamped, zw_clamped); in xnn_qs8_requantize_q31__sse2() local 185 _mm_storeu_si128((__m128i*) output, xyzw_clamped); in xnn_qs8_requantize_q31__sse2()
|