/external/XNNPACK/src/qs8-requantization/ |
D | fp32-sse2.c | 33 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qs8_requantize_fp32__sse2() local 76 const __m128i xy_clamped = _mm_max_epi16(_mm_min_epi16(xy_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__sse2() 77 const __m128i zw_clamped = _mm_max_epi16(_mm_min_epi16(zw_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__sse2()
|
D | fp32-sse4.c | 33 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qs8_requantize_fp32__sse4() local 77 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__sse4()
|
D | precise-ssse3.c | 42 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qs8_requantize_precise__ssse3() local 103 const __m128i xy_clamped = _mm_max_epi16(_mm_min_epi16(xy_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__ssse3() 104 const __m128i zw_clamped = _mm_max_epi16(_mm_min_epi16(zw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__ssse3()
|
D | precise-sse2.c | 42 const __m128i vqmin = _mm_set1_epi8((short) qmin); in xnn_qs8_requantize_precise__sse2() local 108 const __m128i xy_clamped = _mm_max_epi16(_mm_min_epi16(xy_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__sse2() 109 const __m128i zw_clamped = _mm_max_epi16(_mm_min_epi16(zw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__sse2()
|
D | q31-neon.c | 50 const int8x16_t vqmin = vdupq_n_s8(qmin); in xnn_qs8_requantize_q31__neon() local 92 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__neon()
|
D | fp32-neon.c | 34 const int8x16_t vqmin = vdupq_n_s8(qmin); in xnn_qs8_requantize_fp32__neon() local 77 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_fp32__neon()
|
D | q31-ssse3.c | 48 const __m128i vqmin = _mm_set1_epi16((short) qmin); in xnn_qs8_requantize_q31__ssse3() local 160 const __m128i xy_clamped = _mm_max_epi16(_mm_min_epi16(xy_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__ssse3() 161 const __m128i zw_clamped = _mm_max_epi16(_mm_min_epi16(zw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__ssse3()
|
D | precise-sse4.c | 42 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qs8_requantize_precise__sse4() local 96 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__sse4()
|
D | q31-sse2.c | 48 const __m128i vqmin = _mm_set1_epi16((short) qmin); in xnn_qs8_requantize_q31__sse2() local 160 const __m128i xy_clamped = _mm_max_epi16(_mm_min_epi16(xy_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__sse2() 161 const __m128i zw_clamped = _mm_max_epi16(_mm_min_epi16(zw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__sse2()
|
D | q31-sse4.c | 48 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qs8_requantize_q31__sse4() local 112 const __m128i xyzw_clamped = _mm_max_epi8(_mm_min_epi8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_q31__sse4()
|
D | q31-wasmsimd.c | 51 const v128_t vqmin = wasm_i8x16_splat(qmin); in xnn_qs8_requantize_q31__wasmsimd() local 115 const v128_t xyzw_clamped = wasm_i8x16_min(wasm_i8x16_max(xyzw_packed, vqmin), vqmax); in xnn_qs8_requantize_q31__wasmsimd()
|
D | precise-neon.c | 46 const int8x16_t vqmin = vdupq_n_s8(qmin); in xnn_qs8_requantize_precise__neon() local 129 const int8x16_t xyzw_clamped = vmaxq_s8(vminq_s8(xyzw_packed, vqmax), vqmin); in xnn_qs8_requantize_precise__neon()
|
/external/XNNPACK/src/qu8-requantization/ |
D | fp32-sse2.c | 33 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_fp32__sse2() local 77 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_fp32__sse2()
|
D | q31-neon.c | 50 const uint8x16_t vqmin = vdupq_n_u8(qmin); in xnn_qu8_requantize_q31__neon() local 92 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__neon()
|
D | fp32-neon.c | 34 const uint8x16_t vqmin = vdupq_n_u8(qmin); in xnn_qu8_requantize_fp32__neon() local 77 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_fp32__neon()
|
D | precise-sse4.c | 42 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_precise__sse4() local 96 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__sse4()
|
D | precise-ssse3.c | 42 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_precise__ssse3() local 104 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__ssse3()
|
D | q31-sse4.c | 48 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_q31__sse4() local 112 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__sse4()
|
D | precise-sse2.c | 42 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_precise__sse2() local 109 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__sse2()
|
D | q31-wasmsimd.c | 51 const v128_t vqmin = wasm_i8x16_splat((int8_t) qmin); in xnn_qu8_requantize_q31__wasmsimd() local 115 const v128_t xyzw_clamped = wasm_u8x16_min(wasm_u8x16_max(xyzw_packed, vqmin), vqmax); in xnn_qu8_requantize_q31__wasmsimd()
|
D | precise-neon.c | 46 const uint8x16_t vqmin = vdupq_n_u8(qmin); in xnn_qu8_requantize_precise__neon() local 129 const uint8x16_t xyzw_clamped = vmaxq_u8(vminq_u8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_precise__neon()
|
D | q31-ssse3.c | 48 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_q31__ssse3() local 161 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__ssse3()
|
D | q31-sse2.c | 48 const __m128i vqmin = _mm_set1_epi8((char) qmin); in xnn_qu8_requantize_q31__sse2() local 161 const __m128i xyzw_clamped = _mm_max_epu8(_mm_min_epu8(xyzw_packed, vqmax), vqmin); in xnn_qu8_requantize_q31__sse2()
|