/external/XNNPACK/src/f32-vrnd/gen/ |
D | vrndd-wasmsimd-cvt-x4.c | 37 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local 39 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() 51 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local 53 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4()
|
D | vrndu-wasmsimd-cvt-x4.c | 37 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local 39 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() 52 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local 54 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4()
|
D | vrndd-sse2-x8.c | 60 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_f32_vrndd_ukernel__sse2_x8() local 61 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8() 71 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_f32_vrndd_ukernel__sse2_x8() local 72 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8()
|
D | vrndd-wasmsimd-cvt-x8.c | 65 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() local 67 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() 79 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() local 81 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8()
|
D | vrndu-wasmsimd-cvt-x8.c | 68 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local 70 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() 83 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local 85 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
|
D | vrndd-neon-x8.c | 65 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndd_ukernel__neon_x8() local 67 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x8() 77 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndd_ukernel__neon_x8() local 79 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x8()
|
D | vrndu-neon-x8.c | 68 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndu_ukernel__neon_x8() local 70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() 80 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndu_ukernel__neon_x8() local 83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8()
|
D | vrndu-sse2-x8.c | 66 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_f32_vrndu_ukernel__sse2_x8() local 67 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8() 79 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_f32_vrndu_ukernel__sse2_x8() local 80 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
|
D | vrndd-sse2-x4.c | 51 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_f32_vrndd_ukernel__sse2_x4() local 52 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x4()
|
D | vrndd-neon-x4.c | 55 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndd_ukernel__neon_x4() local 57 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x4()
|
D | vrndu-neon-x4.c | 56 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); in xnn_f32_vrndu_ukernel__neon_x4() local 59 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x4()
|
/external/XNNPACK/src/f32-vrnd/ |
D | vrndd-sse2.c.in | 42 const __m128 vprerndx${ABC[N:N+4]} = _mm_cvtepi32_ps(vintx${ABC[N:N+4]}); 45 …{ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]})); 62 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); variable 63 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); 73 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); variable 74 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
|
D | vrndd-wasmsimd-cvt.c.in | 42 const v128_t vprerndx${ABC[N:N+4]} = wasm_f32x4_convert_i32x4(vintx${ABC[N:N+4]}); 46 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vprerndx${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrnd… 65 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); variable 67 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); 79 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); variable 81 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask);
|
D | vrndu-wasmsimd-cvt.c.in | 42 const v128_t vprerndx${ABC[N:N+4]} = wasm_f32x4_convert_i32x4(vintx${ABC[N:N+4]}); 46 …const v128_t vrndx${ABC[N:N+4]} = wasm_v128_bitselect(vprerndx${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrnd… 68 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); variable 70 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); 83 const v128_t vprerndx = wasm_f32x4_convert_i32x4(vintx); variable 85 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask);
|
D | vrndd-neon.c.in | 40 const float32x4_t vprerndx${ABC[N:N+4]} = vcvtq_f32_s32(vintx${ABC[N:N+4]}); 46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$… 65 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); variable 67 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); 77 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); variable 79 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
|
D | vrndu-neon.c.in | 40 const float32x4_t vprerndx${ABC[N:N+4]} = vcvtq_f32_s32(vintx${ABC[N:N+4]}); 46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$… 68 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); variable 70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); 80 const float32x4_t vprerndx = vcvtq_f32_s32(vintx); variable 83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
|
D | vrndu-sse2.c.in | 42 const __m128 vprerndx${ABC[N:N+4]} = _mm_cvtepi32_ps(vintx${ABC[N:N+4]}); 45 …{ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]})); 68 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); variable 69 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); 81 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); variable 82 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
|
/external/XNNPACK/src/math/ |
D | roundu-scalar-cvt.c | 33 const float vprerndx = (float) (int32_t) vx; in xnn_math_f32_roundu__scalar_cvt() local 38 const float vrndx = XNN_UNPREDICTABLE(vabsx < vintegral_threshold) ? vprerndx : vx; in xnn_math_f32_roundu__scalar_cvt()
|
D | roundd-scalar-cvt.c | 33 const float vprerndx = (float) (int32_t) vx; in xnn_math_f32_roundd__scalar_cvt() local 38 const float vrndx = XNN_UNPREDICTABLE(vabsx < vintegral_threshold) ? vprerndx : vx; in xnn_math_f32_roundd__scalar_cvt()
|
D | roundd-wasmsimd-cvt.c | 36 const v128_t vprerndx = wasm_f32x4_convert_i32x4(wasm_i32x4_trunc_saturate_f32x4(vx)); in xnn_math_f32_roundd__wasmsimd_cvt() local 47 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_math_f32_roundd__wasmsimd_cvt()
|
D | roundd-neon-cvt.c | 35 const float32x4_t vprerndx = vcvtq_f32_s32(vcvtq_s32_f32(vx)); in xnn_math_f32_roundd__neon_cvt() local 46 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_math_f32_roundd__neon_cvt()
|
D | roundd-sse2-cvt.c | 43 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_math_f32_roundd__sse2_cvt() local 48 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_math_f32_roundd__sse2_cvt()
|
D | roundu-neon-cvt.c | 35 const float32x4_t vprerndx = vcvtq_f32_s32(vcvtq_s32_f32(vx)); in xnn_math_f32_roundu__neon_cvt() local 46 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_math_f32_roundu__neon_cvt()
|
D | roundu-wasmsimd-cvt.c | 36 const v128_t vprerndx = wasm_f32x4_convert_i32x4(wasm_i32x4_trunc_saturate_f32x4(vx)); in xnn_math_f32_roundu__wasmsimd_cvt() local 47 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_math_f32_roundu__wasmsimd_cvt()
|
D | roundu-sse2-cvt.c | 43 const __m128 vprerndx = _mm_cvtepi32_ps(vintx); in xnn_math_f32_roundu__sse2_cvt() local 48 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_math_f32_roundu__sse2_cvt()
|