| 1 | #include "../simd/integer.h" |
| 2 | |
| 3 | #if GLM_ARCH & GLM_ARCH_SSE2_BIT |
| 4 | |
| 5 | namespace glm{ |
| 6 | namespace detail |
| 7 | { |
| 8 | template<qualifier Q> |
| 9 | struct compute_bitfieldReverseStep<4, uint, Q, true, true> |
| 10 | { |
| 11 | GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v, uint Mask, uint Shift) |
| 12 | { |
| 13 | __m128i const set0 = v.data; |
| 14 | |
| 15 | __m128i const set1 = _mm_set1_epi32(static_cast<int>(Mask)); |
| 16 | __m128i const and1 = _mm_and_si128(set0, set1); |
| 17 | __m128i const sft1 = _mm_slli_epi32(and1, Shift); |
| 18 | |
| 19 | __m128i const set2 = _mm_andnot_si128(set0, _mm_set1_epi32(-1)); |
| 20 | __m128i const and2 = _mm_and_si128(set0, set2); |
| 21 | __m128i const sft2 = _mm_srai_epi32(and2, Shift); |
| 22 | |
| 23 | __m128i const or0 = _mm_or_si128(sft1, sft2); |
| 24 | |
| 25 | return or0; |
| 26 | } |
| 27 | }; |
| 28 | |
| 29 | template<qualifier Q> |
| 30 | struct compute_bitfieldBitCountStep<4, uint, Q, true, true> |
| 31 | { |
| 32 | GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v, uint Mask, uint Shift) |
| 33 | { |
| 34 | __m128i const set0 = v.data; |
| 35 | |
| 36 | __m128i const set1 = _mm_set1_epi32(static_cast<int>(Mask)); |
| 37 | __m128i const and0 = _mm_and_si128(set0, set1); |
| 38 | __m128i const sft0 = _mm_slli_epi32(set0, Shift); |
| 39 | __m128i const and1 = _mm_and_si128(sft0, set1); |
| 40 | __m128i const add0 = _mm_add_epi32(and0, and1); |
| 41 | |
| 42 | return add0; |
| 43 | } |
| 44 | }; |
| 45 | }//namespace detail |
| 46 | |
| 47 | # if GLM_ARCH & GLM_ARCH_AVX_BIT |
| 48 | template<> |
| 49 | GLM_FUNC_QUALIFIER int bitCount(uint x) |
| 50 | { |
| 51 | return _mm_popcnt_u32(x); |
| 52 | } |
| 53 | |
| 54 | # if(GLM_MODEL == GLM_MODEL_64) |
| 55 | template<> |
| 56 | GLM_FUNC_QUALIFIER int bitCount(detail::uint64 x) |
| 57 | { |
| 58 | return static_cast<int>(_mm_popcnt_u64(x)); |
| 59 | } |
| 60 | # endif//GLM_MODEL |
| 61 | # endif//GLM_ARCH |
| 62 | |
| 63 | }//namespace glm |
| 64 | |
| 65 | #endif//GLM_ARCH & GLM_ARCH_SSE2_BIT |