/external/XNNPACK/src/f32-vrnd/gen/ |
D | vrndu-wasmsimd-addsub-x4.c | 38 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() local 39 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() 40 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() 41 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() 52 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() local 53 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() 54 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() 55 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4()
|
D | vrndu-wasmsimd-cvt-x4.c | 39 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local 40 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 41 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 42 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 54 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local 55 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 56 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 57 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4()
|
D | vrndu-wasmsimd-cvt-x8.c | 70 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local 71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 72 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 85 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local 86 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 87 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 88 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
|
D | vrndu-wasmsimd-addsub-x8.c | 68 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() local 69 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() 70 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() 71 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() 82 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() local 83 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() 84 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() 85 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
|
D | vrndd-wasmsimd-addsub-x4.c | 38 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4() local 39 const v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4() 50 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4() local 51 v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4()
|
D | vrndd-wasmsimd-cvt-x4.c | 39 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local 40 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() 41 const v128_t vy = wasm_f32x4_sub(vrndx, vadj); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() 53 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local 54 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() 55 v128_t vy = wasm_f32x4_sub(vrndx, vadj); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4()
|
D | vrndu-neon-x8.c | 70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() local 71 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() 72 const float32x4_t vadjrndx = vaddq_f32(vrndx, vone); in xnn_f32_vrndu_ukernel__neon_x8() 74 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_f32_vrndu_ukernel__neon_x8() 83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() local 84 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() 85 const float32x4_t vadjrndx = vaddq_f32(vrndx, vone); in xnn_f32_vrndu_ukernel__neon_x8() 87 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_f32_vrndu_ukernel__neon_x8()
|
D | vrndu-sse2-x8.c | 67 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8() local 68 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() 69 const __m128 vadjrndx = _mm_add_ps(vrndx, vone); in xnn_f32_vrndu_ukernel__sse2_x8() 70 const __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_f32_vrndu_ukernel__sse2_x8() 80 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8() local 81 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() 82 const __m128 vadjrndx = _mm_add_ps(vrndx, vone); in xnn_f32_vrndu_ukernel__sse2_x8() 83 __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
|
D | vrndd-sse2-x8.c | 61 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8() local 62 const __m128 vy = _mm_sub_ps(vrndx, _mm_and_ps(_mm_cmpgt_ps(vrndx, vx), vone)); in xnn_f32_vrndd_ukernel__sse2_x8() 72 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8() local 73 __m128 vy = _mm_sub_ps(vrndx, _mm_and_ps(_mm_cmpgt_ps(vrndx, vx), vone)); in xnn_f32_vrndd_ukernel__sse2_x8()
|
D | vrndd-wasmsimd-addsub-x8.c | 62 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8() local 63 const v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8() 74 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8() local 75 v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8()
|
D | vrndd-wasmsimd-cvt-x8.c | 67 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() local 68 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() 69 const v128_t vy = wasm_f32x4_sub(vrndx, vadj); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() 81 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() local 82 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() 83 v128_t vy = wasm_f32x4_sub(vrndx, vadj); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8()
|
/external/XNNPACK/src/f32-vrnd/ |
D | vrndu-wasmsimd-addsub.c.in | 47 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vx${ABC[N:N+4]}, vrndabsx${ABC[N:N+4]}, vrnd… 50 …djmask${ABC[N:N+4]} = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx${ABC[N:N+4]}, vrndx${ABC[N:N+4]})); 53 const v128_t vadjrndx${ABC[N:N+4]} = wasm_f32x4_add(vrndx${ABC[N:N+4]}, vone); 56 …const v128_t vy${ABC[N:N+4]} = wasm_v128_bitselect(vrndx${ABC[N:N+4]}, vadjrndx${ABC[N:N+4]}, vadj… 70 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); variable 71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); 72 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); 73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); 84 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); variable 85 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); [all …]
|
D | vrndu-wasmsimd-cvt.c.in | 46 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vprerndx${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrnd… 49 …djmask${ABC[N:N+4]} = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx${ABC[N:N+4]}, vrndx${ABC[N:N+4]})); 52 const v128_t vadjrndx${ABC[N:N+4]} = wasm_f32x4_add(vrndx${ABC[N:N+4]}, vone); 55 …const v128_t vy${ABC[N:N+4]} = wasm_v128_bitselect(vrndx${ABC[N:N+4]}, vadjrndx${ABC[N:N+4]}, vadj… 70 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); variable 71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); 72 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); 73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); 85 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); variable 86 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); [all …]
|
D | vrndu-neon.c.in | 46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$… 49 uint32x4_t vadjmask${ABC[N:N+4]} = vcgeq_f32(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}); 52 const float32x4_t vadjrndx${ABC[N:N+4]} = vaddq_f32(vrndx${ABC[N:N+4]}, vone); 58 …const float32x4_t vy${ABC[N:N+4]} = vbslq_f32(vadjmask${ABC[N:N+4]}, vrndx${ABC[N:N+4]}, vadjrndx$… 70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); variable 71 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); 72 const float32x4_t vadjrndx = vaddq_f32(vrndx, vone); 74 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); 83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); variable 84 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); [all …]
|
D | vrndu-sse2.c.in | 45 …const __m128 vrndx${ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _m… 48 …const __m128 vadjmask${ABC[N:N+4]} = _mm_or_ps(_mm_cmpge_ps(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}), … 51 const __m128 vadjrndx${ABC[N:N+4]} = _mm_add_ps(vrndx${ABC[N:N+4]}, vone); 54 …const __m128 vy${ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vrndx${ABC[N:N+4]}, vadjmask${ABC[N:N+4]}), _m… 69 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); variable 70 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); 71 const __m128 vadjrndx = _mm_add_ps(vrndx, vone); 72 const __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); 82 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); variable 83 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); [all …]
|
D | vrndd-wasmsimd-addsub.c.in | 47 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vx${ABC[N:N+4]}, vrndabsx${ABC[N:N+4]}, vrnd… 50 … v128_t vy${ABC[N:N+4]} = wasm_f32x4_sub(vrndx${ABC[N:N+4]}, wasm_v128_and(wasm_f32x4_lt(vx${ABC[N… 64 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); variable 65 const v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone)); 76 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); variable 77 v128_t vy = wasm_f32x4_sub(vrndx, wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone));
|
D | vrndd-sse2.c.in | 45 …const __m128 vrndx${ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _m… 48 …const __m128 vy${ABC[N:N+4]} = _mm_sub_ps(vrndx${ABC[N:N+4]}, _mm_and_ps(_mm_cmpgt_ps(vrndx${ABC[N… 63 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); variable 64 const __m128 vy = _mm_sub_ps(vrndx, _mm_and_ps(_mm_cmpgt_ps(vrndx, vx), vone)); 74 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); variable 75 __m128 vy = _mm_sub_ps(vrndx, _mm_and_ps(_mm_cmpgt_ps(vrndx, vx), vone));
|
D | vrndd-wasmsimd-cvt.c.in | 46 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vprerndx${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrnd… 49 …const v128_t vadj${ABC[N:N+4]} = wasm_v128_and(wasm_f32x4_lt(vx${ABC[N:N+4]}, vrndx${ABC[N:N+4]}),… 52 const v128_t vy${ABC[N:N+4]} = wasm_f32x4_sub(vrndx${ABC[N:N+4]}, vadj${ABC[N:N+4]}); 67 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); variable 68 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); 69 const v128_t vy = wasm_f32x4_sub(vrndx, vadj); 81 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); variable 82 const v128_t vadj = wasm_v128_and(wasm_f32x4_lt(vx, vrndx), vone); 83 v128_t vy = wasm_f32x4_sub(vrndx, vadj);
|
D | vrndd-neon.c.in | 46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$… 49 const uint32x4_t vadjmask${ABC[N:N+4]} = vcgtq_f32(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}); 55 const float32x4_t vy${ABC[N:N+4]} = vsubq_f32(vrndx${ABC[N:N+4]}, vadjrndx${ABC[N:N+4]}); 67 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); variable 68 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); 70 const float32x4_t vy = vsubq_f32(vrndx, vadjrndx); 79 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); variable 80 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); 82 const float32x4_t vy = vsubq_f32(vrndx, vadjrndx);
|
/external/XNNPACK/src/math/ |
D | roundu-scalar-addsub.c | 46 const float vrndx = copysignf(vrndabsx, vx); in xnn_math_f32_roundu__scalar_addsub() local 49 const float vprey = XNN_UNPREDICTABLE(vrndx < vx) ? vrndx + vone : vrndx; in xnn_math_f32_roundu__scalar_addsub() 53 const float vy = copysignf(vprey, vrndx); in xnn_math_f32_roundu__scalar_addsub()
|
D | roundu-scalar-cvt.c | 38 const float vrndx = XNN_UNPREDICTABLE(vabsx < vintegral_threshold) ? vprerndx : vx; in xnn_math_f32_roundu__scalar_cvt() local 42 const float vprey = XNN_UNPREDICTABLE(vrndx >= vx) ? vrndx : vrndx + vone; in xnn_math_f32_roundu__scalar_cvt()
|
D | roundd-scalar-cvt.c | 38 const float vrndx = XNN_UNPREDICTABLE(vabsx < vintegral_threshold) ? vprerndx : vx; in xnn_math_f32_roundd__scalar_cvt() local 41 const float vadjrndx = copysignf(vrndx, vx); in xnn_math_f32_roundd__scalar_cvt() 45 const float vy = XNN_UNPREDICTABLE(vrndx <= vx) ? vadjrndx : vrndx - vone; in xnn_math_f32_roundd__scalar_cvt()
|
D | roundd-scalar-addsub.c | 46 const float vrndx = copysignf(vrndabsx, vx); in xnn_math_f32_roundd__scalar_addsub() local 49 const float vy = XNN_UNPREDICTABLE(vrndx > vx) ? vrndx - vone : vrndx; in xnn_math_f32_roundd__scalar_addsub()
|
D | roundu-neon-cvt.c | 46 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_math_f32_roundu__neon_cvt() local 52 const uint32x4_t vadjmask = vorrq_u32(vcgeq_f32(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__neon_cvt() 57 const float32x4_t vadjrndx = vaddq_f32(vrndx, vone); in xnn_math_f32_roundu__neon_cvt() 65 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_math_f32_roundu__neon_cvt()
|
D | roundu-wasmsimd-cvt.c | 47 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_math_f32_roundu__wasmsimd_cvt() local 53 const v128_t vadjmask = wasm_v128_or(wasm_f32x4_ge(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__wasmsimd_cvt() 58 const v128_t vadjrndx = wasm_f32x4_add(vrndx, vone); in xnn_math_f32_roundu__wasmsimd_cvt() 66 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_math_f32_roundu__wasmsimd_cvt()
|