/external/XNNPACK/src/f32-vrnd/gen/ |
D | vrndu-sse2-x8.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndu_ukernel__sse2_x8() local 38 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndu_ukernel__sse2_x8() 39 …nst __m128 vrndmask4567 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx4567, vmagic)… in xnn_f32_vrndu_ukernel__sse2_x8() 47 … const __m128 vadjmask0123 = _mm_or_ps(_mm_cmpge_ps(vrndx0123, vx0123), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() 48 … const __m128 vadjmask4567 = _mm_or_ps(_mm_cmpge_ps(vrndx4567, vx4567), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() 65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndu_ukernel__sse2_x8() 68 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() 78 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndu_ukernel__sse2_x8() 81 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8()
|
D | vrndne-sse2-x8.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndne_ukernel__sse2_x8() local 37 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndne_ukernel__sse2_x8() 38 …nst __m128 vrndmask4567 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx4567, vmagic)… in xnn_f32_vrndne_ukernel__sse2_x8() 55 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndne_ukernel__sse2_x8() 65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndne_ukernel__sse2_x8()
|
D | vrndz-sse2-x8.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndz_ukernel__sse2_x8() local 37 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndz_ukernel__sse2_x8() 38 …nst __m128 vrndmask4567 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx4567, vmagic)… in xnn_f32_vrndz_ukernel__sse2_x8() 55 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndz_ukernel__sse2_x8() 65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndz_ukernel__sse2_x8()
|
D | vrndd-sse2-x8.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndd_ukernel__sse2_x8() local 38 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndd_ukernel__sse2_x8() 39 …nst __m128 vrndmask4567 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx4567, vmagic)… in xnn_f32_vrndd_ukernel__sse2_x8() 59 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndd_ukernel__sse2_x8() 70 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndd_ukernel__sse2_x8()
|
D | vrndu-sse2-x4.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndu_ukernel__sse2_x4() local 36 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndu_ukernel__sse2_x4() 42 … const __m128 vadjmask0123 = _mm_or_ps(_mm_cmpge_ps(vrndx0123, vx0123), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x4() 54 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndu_ukernel__sse2_x4() 57 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x4()
|
D | vrndne-sse2-x4.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndne_ukernel__sse2_x4() local 35 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndne_ukernel__sse2_x4() 47 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndne_ukernel__sse2_x4()
|
D | vrndz-sse2-x4.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndz_ukernel__sse2_x4() local 35 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndz_ukernel__sse2_x4() 47 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndz_ukernel__sse2_x4()
|
D | vrndd-sse2-x4.c | 28 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); in xnn_f32_vrndd_ukernel__sse2_x4() local 36 …nst __m128 vrndmask0123 = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx0123, vmagic)… in xnn_f32_vrndd_ukernel__sse2_x4() 50 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndd_ukernel__sse2_x4()
|
/external/XNNPACK/src/f32-vrnd/ |
D | vrndu-sse2.c.in | 27 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); variable 39 …dmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${ABC[N:N+4]}, vma… 48 …C[N:N+4]} = _mm_or_ps(_mm_cmpge_ps(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}), _mm_castsi128_ps(vmagic)); 67 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); 70 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); 80 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); 83 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic));
|
D | vrndne-sse2.c.in | 27 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); variable 38 …dmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${ABC[N:N+4]}, vma… 57 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); 67 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic)));
|
D | vrndz-sse2.c.in | 27 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); variable 38 …dmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${ABC[N:N+4]}, vma… 57 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); 67 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic)));
|
D | vrndd-sse2.c.in | 27 const __m128i vmagic = _mm_load_si128((const __m128i*) params->sse2.sign_mask); variable 39 …dmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${ABC[N:N+4]}, vma… 61 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); 72 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic)));
|
/external/XNNPACK/src/math/ |
D | roundu-sse2-cvt.c | 24 const __m128i vmagic = _mm_set1_epi32(0x80000000); in xnn_math_f32_roundu__sse2_cvt() local 39 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_math_f32_roundu__sse2_cvt() 54 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_math_f32_roundu__sse2_cvt()
|
D | roundne-sse2-cvt.c | 24 const __m128i vmagic = _mm_set1_epi32(0x80000000); in xnn_math_f32_roundne__sse2_cvt() local 37 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_math_f32_roundne__sse2_cvt()
|
D | roundz-sse2-cvt.c | 24 const __m128i vmagic = _mm_set1_epi32(0x80000000); in xnn_math_f32_roundz__sse2_cvt() local 37 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_math_f32_roundz__sse2_cvt()
|
D | roundd-sse2-cvt.c | 24 const __m128i vmagic = _mm_set1_epi32(0x80000000); in xnn_math_f32_roundd__sse2_cvt() local 39 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_math_f32_roundd__sse2_cvt()
|