Jolt Physics
A multi core friendly Game Physics Engine
Loading...
Searching...
No Matches
UVec4.inl
Go to the documentation of this file.
1// Jolt Physics Library (https://github.com/jrouwe/JoltPhysics)
2// SPDX-FileCopyrightText: 2021 Jorrit Rouwe
3// SPDX-License-Identifier: MIT
4
5JPH_NAMESPACE_BEGIN
6
7UVec4::UVec4(uint32 inX, uint32 inY, uint32 inZ, uint32 inW)
8{
9#if defined(JPH_USE_SSE)
10 mValue = _mm_set_epi32(int(inW), int(inZ), int(inY), int(inX));
11#elif defined(JPH_USE_NEON)
12 uint32x2_t xy = vcreate_u32(static_cast<uint64>(inX) | (static_cast<uint64>(inY) << 32));
13 uint32x2_t zw = vcreate_u32(static_cast<uint64>(inZ) | (static_cast<uint64>(inW) << 32));
14 mValue = vcombine_u32(xy, zw);
15#else
16 mU32[0] = inX;
17 mU32[1] = inY;
18 mU32[2] = inZ;
19 mU32[3] = inW;
20#endif
21}
22
23bool UVec4::operator == (UVec4Arg inV2) const
24{
25 return sEquals(*this, inV2).TestAllTrue();
26}
27
28template<uint32 SwizzleX, uint32 SwizzleY, uint32 SwizzleZ, uint32 SwizzleW>
29UVec4 UVec4::Swizzle() const
30{
31 static_assert(SwizzleX <= 3, "SwizzleX template parameter out of range");
32 static_assert(SwizzleY <= 3, "SwizzleY template parameter out of range");
33 static_assert(SwizzleZ <= 3, "SwizzleZ template parameter out of range");
34 static_assert(SwizzleW <= 3, "SwizzleW template parameter out of range");
35
36#if defined(JPH_USE_SSE)
37 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(SwizzleW, SwizzleZ, SwizzleY, SwizzleX));
38#elif defined(JPH_USE_NEON)
39 return JPH_NEON_SHUFFLE_U32x4(mValue, mValue, SwizzleX, SwizzleY, SwizzleZ, SwizzleW);
40#else
41 return UVec4(mU32[SwizzleX], mU32[SwizzleY], mU32[SwizzleZ], mU32[SwizzleW]);
42#endif
43}
44
45UVec4 UVec4::sZero()
46{
47#if defined(JPH_USE_SSE)
48 return _mm_setzero_si128();
49#elif defined(JPH_USE_NEON)
50 return vdupq_n_u32(0);
51#else
52 return UVec4(0, 0, 0, 0);
53#endif
54}
55
56UVec4 UVec4::sReplicate(uint32 inV)
57{
58#if defined(JPH_USE_SSE)
59 return _mm_set1_epi32(int(inV));
60#elif defined(JPH_USE_NEON)
61 return vdupq_n_u32(inV);
62#else
63 return UVec4(inV, inV, inV, inV);
64#endif
65}
66
67UVec4 UVec4::sLoadInt(const uint32 *inV)
68{
69#if defined(JPH_USE_SSE)
70 return _mm_castps_si128(_mm_load_ss(reinterpret_cast<const float*>(inV)));
71#elif defined(JPH_USE_NEON)
72 return vsetq_lane_u32(*inV, vdupq_n_u32(0), 0);
73#else
74 return UVec4(*inV, 0, 0, 0);
75#endif
76}
77
78UVec4 UVec4::sLoadInt4(const uint32 *inV)
79{
80#if defined(JPH_USE_SSE)
81 return _mm_loadu_si128(reinterpret_cast<const __m128i *>(inV));
82#elif defined(JPH_USE_NEON)
83 return vld1q_u32(inV);
84#else
85 return UVec4(inV[0], inV[1], inV[2], inV[3]);
86#endif
87}
88
89UVec4 UVec4::sLoadInt4Aligned(const uint32 *inV)
90{
91#if defined(JPH_USE_SSE)
92 return _mm_load_si128(reinterpret_cast<const __m128i *>(inV));
93#elif defined(JPH_USE_NEON)
94 return vld1q_u32(inV); // ARM doesn't make distinction between aligned or not
95#else
96 return UVec4(inV[0], inV[1], inV[2], inV[3]);
97#endif
98}
99
100template <const int Scale>
101UVec4 UVec4::sGatherInt4(const uint32 *inBase, UVec4Arg inOffsets)
102{
103#ifdef JPH_USE_AVX2
104 return _mm_i32gather_epi32(reinterpret_cast<const int *>(inBase), inOffsets.mValue, Scale);
105#else
106 const uint8 *base = reinterpret_cast<const uint8 *>(inBase);
107 uint32 x = *reinterpret_cast<const uint32 *>(base + inOffsets.GetX() * Scale);
108 uint32 y = *reinterpret_cast<const uint32 *>(base + inOffsets.GetY() * Scale);
109 uint32 z = *reinterpret_cast<const uint32 *>(base + inOffsets.GetZ() * Scale);
110 uint32 w = *reinterpret_cast<const uint32 *>(base + inOffsets.GetW() * Scale);
111 return UVec4(x, y, z, w);
112#endif
113}
114
115UVec4 UVec4::sMin(UVec4Arg inV1, UVec4Arg inV2)
116{
117#if defined(JPH_USE_SSE4_1)
118 return _mm_min_epu32(inV1.mValue, inV2.mValue);
119#elif defined(JPH_USE_NEON)
120 return vminq_u32(inV1.mValue, inV2.mValue);
121#else
122 UVec4 result;
123 for (int i = 0; i < 4; i++)
124 result.mU32[i] = min(inV1.mU32[i], inV2.mU32[i]);
125 return result;
126#endif
127}
128
129UVec4 UVec4::sMax(UVec4Arg inV1, UVec4Arg inV2)
130{
131#if defined(JPH_USE_SSE4_1)
132 return _mm_max_epu32(inV1.mValue, inV2.mValue);
133#elif defined(JPH_USE_NEON)
134 return vmaxq_u32(inV1.mValue, inV2.mValue);
135#else
136 UVec4 result;
137 for (int i = 0; i < 4; i++)
138 result.mU32[i] = max(inV1.mU32[i], inV2.mU32[i]);
139 return result;
140#endif
141}
142
143UVec4 UVec4::sEquals(UVec4Arg inV1, UVec4Arg inV2)
144{
145#if defined(JPH_USE_SSE)
146 return _mm_cmpeq_epi32(inV1.mValue, inV2.mValue);
147#elif defined(JPH_USE_NEON)
148 return vceqq_u32(inV1.mValue, inV2.mValue);
149#else
150 return UVec4(inV1.mU32[0] == inV2.mU32[0]? 0xffffffffu : 0,
151 inV1.mU32[1] == inV2.mU32[1]? 0xffffffffu : 0,
152 inV1.mU32[2] == inV2.mU32[2]? 0xffffffffu : 0,
153 inV1.mU32[3] == inV2.mU32[3]? 0xffffffffu : 0);
154#endif
155}
156
157UVec4 UVec4::sSelect(UVec4Arg inNotSet, UVec4Arg inSet, UVec4Arg inControl)
158{
159#if defined(JPH_USE_SSE4_1) && !defined(JPH_PLATFORM_WASM) // _mm_blendv_ps has problems on FireFox
160 return _mm_castps_si128(_mm_blendv_ps(_mm_castsi128_ps(inNotSet.mValue), _mm_castsi128_ps(inSet.mValue), _mm_castsi128_ps(inControl.mValue)));
161#elif defined(JPH_USE_SSE)
162 __m128 is_set = _mm_castsi128_ps(_mm_srai_epi32(inControl.mValue, 31));
163 return _mm_castps_si128(_mm_or_ps(_mm_and_ps(is_set, _mm_castsi128_ps(inSet.mValue)), _mm_andnot_ps(is_set, _mm_castsi128_ps(inNotSet.mValue))));
164#elif defined(JPH_USE_NEON)
165 return vbslq_u32(vreinterpretq_u32_s32(vshrq_n_s32(vreinterpretq_s32_u32(inControl.mValue), 31)), inSet.mValue, inNotSet.mValue);
166#else
167 UVec4 result;
168 for (int i = 0; i < 4; i++)
169 result.mU32[i] = (inControl.mU32[i] & 0x80000000u) ? inSet.mU32[i] : inNotSet.mU32[i];
170 return result;
171#endif
172}
173
174UVec4 UVec4::sOr(UVec4Arg inV1, UVec4Arg inV2)
175{
176#if defined(JPH_USE_SSE)
177 return _mm_or_si128(inV1.mValue, inV2.mValue);
178#elif defined(JPH_USE_NEON)
179 return vorrq_u32(inV1.mValue, inV2.mValue);
180#else
181 return UVec4(inV1.mU32[0] | inV2.mU32[0],
182 inV1.mU32[1] | inV2.mU32[1],
183 inV1.mU32[2] | inV2.mU32[2],
184 inV1.mU32[3] | inV2.mU32[3]);
185#endif
186}
187
188UVec4 UVec4::sXor(UVec4Arg inV1, UVec4Arg inV2)
189{
190#if defined(JPH_USE_SSE)
191 return _mm_xor_si128(inV1.mValue, inV2.mValue);
192#elif defined(JPH_USE_NEON)
193 return veorq_u32(inV1.mValue, inV2.mValue);
194#else
195 return UVec4(inV1.mU32[0] ^ inV2.mU32[0],
196 inV1.mU32[1] ^ inV2.mU32[1],
197 inV1.mU32[2] ^ inV2.mU32[2],
198 inV1.mU32[3] ^ inV2.mU32[3]);
199#endif
200}
201
202UVec4 UVec4::sAnd(UVec4Arg inV1, UVec4Arg inV2)
203{
204#if defined(JPH_USE_SSE)
205 return _mm_and_si128(inV1.mValue, inV2.mValue);
206#elif defined(JPH_USE_NEON)
207 return vandq_u32(inV1.mValue, inV2.mValue);
208#else
209 return UVec4(inV1.mU32[0] & inV2.mU32[0],
210 inV1.mU32[1] & inV2.mU32[1],
211 inV1.mU32[2] & inV2.mU32[2],
212 inV1.mU32[3] & inV2.mU32[3]);
213#endif
214}
215
216
217UVec4 UVec4::sNot(UVec4Arg inV1)
218{
219#if defined(JPH_USE_AVX512)
220 return _mm_ternarylogic_epi32(inV1.mValue, inV1.mValue, inV1.mValue, 0b01010101);
221#elif defined(JPH_USE_SSE)
222 return sXor(inV1, sReplicate(0xffffffff));
223#elif defined(JPH_USE_NEON)
224 return vmvnq_u32(inV1.mValue);
225#else
226 return UVec4(~inV1.mU32[0], ~inV1.mU32[1], ~inV1.mU32[2], ~inV1.mU32[3]);
227#endif
228}
229
230UVec4 UVec4::sSort4True(UVec4Arg inValue, UVec4Arg inIndex)
231{
232 // If inValue.z is false then shift W to Z
233 UVec4 v = UVec4::sSelect(inIndex.Swizzle<SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W, SWIZZLE_W>(), inIndex, inValue.SplatZ());
234
235 // If inValue.y is false then shift Z and further to Y and further
236 v = UVec4::sSelect(v.Swizzle<SWIZZLE_X, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W>(), v, inValue.SplatY());
237
238 // If inValue.x is false then shift X and further to Y and further
239 v = UVec4::sSelect(v.Swizzle<SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W>(), v, inValue.SplatX());
240
241 return v;
242}
243
244UVec4 UVec4::operator * (UVec4Arg inV2) const
245{
246#if defined(JPH_USE_SSE4_1)
247 return _mm_mullo_epi32(mValue, inV2.mValue);
248#elif defined(JPH_USE_NEON)
249 return vmulq_u32(mValue, inV2.mValue);
250#else
251 UVec4 result;
252 for (int i = 0; i < 4; i++)
253 result.mU32[i] = mU32[i] * inV2.mU32[i];
254 return result;
255#endif
256}
257
258UVec4 UVec4::operator + (UVec4Arg inV2)
259{
260#if defined(JPH_USE_SSE)
261 return _mm_add_epi32(mValue, inV2.mValue);
262#elif defined(JPH_USE_NEON)
263 return vaddq_u32(mValue, inV2.mValue);
264#else
265 return UVec4(mU32[0] + inV2.mU32[0],
266 mU32[1] + inV2.mU32[1],
267 mU32[2] + inV2.mU32[2],
268 mU32[3] + inV2.mU32[3]);
269#endif
270}
271
272UVec4 &UVec4::operator += (UVec4Arg inV2)
273{
274#if defined(JPH_USE_SSE)
275 mValue = _mm_add_epi32(mValue, inV2.mValue);
276#elif defined(JPH_USE_NEON)
277 mValue = vaddq_u32(mValue, inV2.mValue);
278#else
279 for (int i = 0; i < 4; ++i)
280 mU32[i] += inV2.mU32[i];
281#endif
282 return *this;
283}
284
285UVec4 UVec4::SplatX() const
286{
287#if defined(JPH_USE_SSE)
288 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(0, 0, 0, 0));
289#elif defined(JPH_USE_NEON)
290 return vdupq_laneq_u32(mValue, 0);
291#else
292 return UVec4(mU32[0], mU32[0], mU32[0], mU32[0]);
293#endif
294}
295
296UVec4 UVec4::SplatY() const
297{
298#if defined(JPH_USE_SSE)
299 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(1, 1, 1, 1));
300#elif defined(JPH_USE_NEON)
301 return vdupq_laneq_u32(mValue, 1);
302#else
303 return UVec4(mU32[1], mU32[1], mU32[1], mU32[1]);
304#endif
305}
306
307UVec4 UVec4::SplatZ() const
308{
309#if defined(JPH_USE_SSE)
310 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(2, 2, 2, 2));
311#elif defined(JPH_USE_NEON)
312 return vdupq_laneq_u32(mValue, 2);
313#else
314 return UVec4(mU32[2], mU32[2], mU32[2], mU32[2]);
315#endif
316}
317
318UVec4 UVec4::SplatW() const
319{
320#if defined(JPH_USE_SSE)
321 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(3, 3, 3, 3));
322#elif defined(JPH_USE_NEON)
323 return vdupq_laneq_u32(mValue, 3);
324#else
325 return UVec4(mU32[3], mU32[3], mU32[3], mU32[3]);
326#endif
327}
328
329Vec4 UVec4::ToFloat() const
330{
331#if defined(JPH_USE_SSE)
332 return _mm_cvtepi32_ps(mValue);
333#elif defined(JPH_USE_NEON)
334 return vcvtq_f32_u32(mValue);
335#else
336 return Vec4((float)mU32[0], (float)mU32[1], (float)mU32[2], (float)mU32[3]);
337#endif
338}
339
340Vec4 UVec4::ReinterpretAsFloat() const
341{
342#if defined(JPH_USE_SSE)
343 return Vec4(_mm_castsi128_ps(mValue));
344#elif defined(JPH_USE_NEON)
345 return vreinterpretq_f32_u32(mValue);
346#else
347 return *reinterpret_cast<const Vec4 *>(this);
348#endif
349}
350
351void UVec4::StoreInt4(uint32 *outV) const
352{
353#if defined(JPH_USE_SSE)
354 _mm_storeu_si128(reinterpret_cast<__m128i *>(outV), mValue);
355#elif defined(JPH_USE_NEON)
356 vst1q_u32(outV, mValue);
357#else
358 for (int i = 0; i < 4; ++i)
359 outV[i] = mU32[i];
360#endif
361}
362
363void UVec4::StoreInt4Aligned(uint32 *outV) const
364{
365#if defined(JPH_USE_SSE)
366 _mm_store_si128(reinterpret_cast<__m128i *>(outV), mValue);
367#elif defined(JPH_USE_NEON)
368 vst1q_u32(outV, mValue); // ARM doesn't make distinction between aligned or not
369#else
370 for (int i = 0; i < 4; ++i)
371 outV[i] = mU32[i];
372#endif
373}
374
375int UVec4::CountTrues() const
376{
377#if defined(JPH_USE_SSE)
378 return CountBits(_mm_movemask_ps(_mm_castsi128_ps(mValue)));
379#elif defined(JPH_USE_NEON)
380 return vaddvq_u32(vshrq_n_u32(mValue, 31));
381#else
382 return (mU32[0] >> 31) + (mU32[1] >> 31) + (mU32[2] >> 31) + (mU32[3] >> 31);
383#endif
384}
385
386int UVec4::GetTrues() const
387{
388#if defined(JPH_USE_SSE)
389 return _mm_movemask_ps(_mm_castsi128_ps(mValue));
390#elif defined(JPH_USE_NEON)
391 int32x4_t shift = JPH_NEON_INT32x4(0, 1, 2, 3);
392 return vaddvq_u32(vshlq_u32(vshrq_n_u32(mValue, 31), shift));
393#else
394 return (mU32[0] >> 31) | ((mU32[1] >> 31) << 1) | ((mU32[2] >> 31) << 2) | ((mU32[3] >> 31) << 3);
395#endif
396}
397
398bool UVec4::TestAnyTrue() const
399{
400 return GetTrues() != 0;
401}
402
403bool UVec4::TestAnyXYZTrue() const
404{
405 return (GetTrues() & 0b111) != 0;
406}
407
408bool UVec4::TestAllTrue() const
409{
410 return GetTrues() == 0b1111;
411}
412
413bool UVec4::TestAllXYZTrue() const
414{
415 return (GetTrues() & 0b111) == 0b111;
416}
417
418template <const uint Count>
419UVec4 UVec4::LogicalShiftLeft() const
420{
421 static_assert(Count <= 31, "Invalid shift");
422
423#if defined(JPH_USE_SSE)
424 return _mm_slli_epi32(mValue, Count);
425#elif defined(JPH_USE_NEON)
426 return vshlq_n_u32(mValue, Count);
427#else
428 return UVec4(mU32[0] << Count, mU32[1] << Count, mU32[2] << Count, mU32[3] << Count);
429#endif
430}
431
432template <const uint Count>
433UVec4 UVec4::LogicalShiftRight() const
434{
435 static_assert(Count <= 31, "Invalid shift");
436
437#if defined(JPH_USE_SSE)
438 return _mm_srli_epi32(mValue, Count);
439#elif defined(JPH_USE_NEON)
440 return vshrq_n_u32(mValue, Count);
441#else
442 return UVec4(mU32[0] >> Count, mU32[1] >> Count, mU32[2] >> Count, mU32[3] >> Count);
443#endif
444}
445
446template <const uint Count>
447UVec4 UVec4::ArithmeticShiftRight() const
448{
449 static_assert(Count <= 31, "Invalid shift");
450
451#if defined(JPH_USE_SSE)
452 return _mm_srai_epi32(mValue, Count);
453#elif defined(JPH_USE_NEON)
454 return vreinterpretq_u32_s32(vshrq_n_s32(vreinterpretq_s32_u32(mValue), Count));
455#else
456 return UVec4(uint32(int32_t(mU32[0]) >> Count),
457 uint32(int32_t(mU32[1]) >> Count),
458 uint32(int32_t(mU32[2]) >> Count),
459 uint32(int32_t(mU32[3]) >> Count));
460#endif
461}
462
463UVec4 UVec4::Expand4Uint16Lo() const
464{
465#if defined(JPH_USE_SSE)
466 return _mm_unpacklo_epi16(mValue, _mm_castps_si128(_mm_setzero_ps()));
467#elif defined(JPH_USE_NEON)
468 uint16x4_t value = vget_low_u16(vreinterpretq_u16_u32(mValue));
469 uint16x4_t zero = vdup_n_u16(0);
470 return vreinterpretq_u32_u16(vcombine_u16(vzip1_u16(value, zero), vzip2_u16(value, zero)));
471#else
472 return UVec4(mU32[0] & 0xffff,
473 (mU32[0] >> 16) & 0xffff,
474 mU32[1] & 0xffff,
475 (mU32[1] >> 16) & 0xffff);
476#endif
477}
478
479UVec4 UVec4::Expand4Uint16Hi() const
480{
481#if defined(JPH_USE_SSE)
482 return _mm_unpackhi_epi16(mValue, _mm_castps_si128(_mm_setzero_ps()));
483#elif defined(JPH_USE_NEON)
484 uint16x4_t value = vget_high_u16(vreinterpretq_u16_u32(mValue));
485 uint16x4_t zero = vdup_n_u16(0);
486 return vreinterpretq_u32_u16(vcombine_u16(vzip1_u16(value, zero), vzip2_u16(value, zero)));
487#else
488 return UVec4(mU32[2] & 0xffff,
489 (mU32[2] >> 16) & 0xffff,
490 mU32[3] & 0xffff,
491 (mU32[3] >> 16) & 0xffff);
492#endif
493}
494
495UVec4 UVec4::Expand4Byte0() const
496{
497#if defined(JPH_USE_SSE4_1)
498 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff03), int(0xffffff02), int(0xffffff01), int(0xffffff00)));
499#elif defined(JPH_USE_NEON)
500 uint8x16_t idx = JPH_NEON_UINT8x16(0x00, 0x7f, 0x7f, 0x7f, 0x01, 0x7f, 0x7f, 0x7f, 0x02, 0x7f, 0x7f, 0x7f, 0x03, 0x7f, 0x7f, 0x7f);
501 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
502#else
503 UVec4 result;
504 for (int i = 0; i < 4; i++)
505 result.mU32[i] = (mU32[0] >> (i * 8)) & 0xff;
506 return result;
507#endif
508}
509
510UVec4 UVec4::Expand4Byte4() const
511{
512#if defined(JPH_USE_SSE4_1)
513 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff07), int(0xffffff06), int(0xffffff05), int(0xffffff04)));
514#elif defined(JPH_USE_NEON)
515 uint8x16_t idx = JPH_NEON_UINT8x16(0x04, 0x7f, 0x7f, 0x7f, 0x05, 0x7f, 0x7f, 0x7f, 0x06, 0x7f, 0x7f, 0x7f, 0x07, 0x7f, 0x7f, 0x7f);
516 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
517#else
518 UVec4 result;
519 for (int i = 0; i < 4; i++)
520 result.mU32[i] = (mU32[1] >> (i * 8)) & 0xff;
521 return result;
522#endif
523}
524
525UVec4 UVec4::Expand4Byte8() const
526{
527#if defined(JPH_USE_SSE4_1)
528 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff0b), int(0xffffff0a), int(0xffffff09), int(0xffffff08)));
529#elif defined(JPH_USE_NEON)
530 uint8x16_t idx = JPH_NEON_UINT8x16(0x08, 0x7f, 0x7f, 0x7f, 0x09, 0x7f, 0x7f, 0x7f, 0x0a, 0x7f, 0x7f, 0x7f, 0x0b, 0x7f, 0x7f, 0x7f);
531 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
532#else
533 UVec4 result;
534 for (int i = 0; i < 4; i++)
535 result.mU32[i] = (mU32[2] >> (i * 8)) & 0xff;
536 return result;
537#endif
538}
539
540UVec4 UVec4::Expand4Byte12() const
541{
542#if defined(JPH_USE_SSE4_1)
543 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff0f), int(0xffffff0e), int(0xffffff0d), int(0xffffff0c)));
544#elif defined(JPH_USE_NEON)
545 uint8x16_t idx = JPH_NEON_UINT8x16(0x0c, 0x7f, 0x7f, 0x7f, 0x0d, 0x7f, 0x7f, 0x7f, 0x0e, 0x7f, 0x7f, 0x7f, 0x0f, 0x7f, 0x7f, 0x7f);
546 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
547#else
548 UVec4 result;
549 for (int i = 0; i < 4; i++)
550 result.mU32[i] = (mU32[3] >> (i * 8)) & 0xff;
551 return result;
552#endif
553}
554
555UVec4 UVec4::ShiftComponents4Minus(int inCount) const
556{
557#if defined(JPH_USE_SSE4_1) || defined(JPH_USE_NEON)
558 alignas(UVec4) static constexpr uint32 sFourMinusXShuffle[5][4] =
559 {
560 { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff },
561 { 0x0f0e0d0c, 0xffffffff, 0xffffffff, 0xffffffff },
562 { 0x0b0a0908, 0x0f0e0d0c, 0xffffffff, 0xffffffff },
563 { 0x07060504, 0x0b0a0908, 0x0f0e0d0c, 0xffffffff },
564 { 0x03020100, 0x07060504, 0x0b0a0908, 0x0f0e0d0c }
565 };
566#endif
567
568#if defined(JPH_USE_SSE4_1)
569 return _mm_shuffle_epi8(mValue, *reinterpret_cast<const UVec4::Type *>(sFourMinusXShuffle[inCount]));
570#elif defined(JPH_USE_NEON)
571 uint8x16_t idx = vreinterpretq_u8_u32(*reinterpret_cast<const UVec4::Type *>(sFourMinusXShuffle[inCount]));
572 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
573#else
574 UVec4 result = UVec4::sZero();
575 for (int i = 0; i < inCount; i++)
576 result.mU32[i] = mU32[i + 4 - inCount];
577 return result;
578#endif
579}
580
581JPH_NAMESPACE_END
uint8
std::uint8_t uint8
Definition Core.h:454
uint64
std::uint64_t uint64
Definition Core.h:457
JPH_NAMESPACE_END
#define JPH_NAMESPACE_END
Definition Core.h:379
uint32
std::uint32_t uint32
Definition Core.h:456
JPH_NAMESPACE_BEGIN
#define JPH_NAMESPACE_BEGIN
Definition Core.h:373
CountBits
uint CountBits(uint32 inValue)
Count the number of 1 bits in a value.
Definition Math.h:161
SWIZZLE_Z
@ SWIZZLE_Z
Use the Z component.
Definition Swizzle.h:14
SWIZZLE_W
@ SWIZZLE_W
Use the W component.
Definition Swizzle.h:15
SWIZZLE_X
@ SWIZZLE_X
Use the X component.
Definition Swizzle.h:12
SWIZZLE_Y
@ SWIZZLE_Y
Use the Y component.
Definition Swizzle.h:13
UVec4
Definition UVec4.h:12
UVec4::Swizzle
JPH_INLINE UVec4 Swizzle() const
Swizzle the elements in inV.
UVec4::sNot
static JPH_INLINE UVec4 sNot(UVec4Arg inV1)
Logical not (component wise)
Definition UVec4.inl:217
UVec4::GetZ
JPH_INLINE uint32 GetZ() const
Definition UVec4.h:104
UVec4::sMin
static JPH_INLINE UVec4 sMin(UVec4Arg inV1, UVec4Arg inV2)
Return the minimum value of each of the components.
Definition UVec4.inl:115
UVec4::LogicalShiftLeft
JPH_INLINE UVec4 LogicalShiftLeft() const
Shift all components by Count bits to the left (filling with zeros from the left)
UVec4::CountTrues
JPH_INLINE int CountTrues() const
Count the number of components that are true (true is when highest bit of component is set)
Definition UVec4.inl:375
UVec4::SplatY
JPH_INLINE UVec4 SplatY() const
Replicate the Y component to all components.
Definition UVec4.inl:296
UVec4::sSelect
static JPH_INLINE UVec4 sSelect(UVec4Arg inNotSet, UVec4Arg inSet, UVec4Arg inControl)
Component wise select, returns inNotSet when highest bit of inControl = 0 and inSet when highest bit ...
Definition UVec4.inl:157
UVec4::sLoadInt
static JPH_INLINE UVec4 sLoadInt(const uint32 *inV)
Load 1 int from memory and place it in the X component, zeros Y, Z and W.
Definition UVec4.inl:67
UVec4::Expand4Uint16Lo
JPH_INLINE UVec4 Expand4Uint16Lo() const
Takes the lower 4 16 bits and expands them to X, Y, Z and W.
Definition UVec4.inl:463
UVec4::sSort4True
static JPH_INLINE UVec4 sSort4True(UVec4Arg inValue, UVec4Arg inIndex)
Definition UVec4.inl:230
UVec4::GetY
JPH_INLINE uint32 GetY() const
Definition UVec4.h:103
UVec4::LogicalShiftRight
JPH_INLINE UVec4 LogicalShiftRight() const
Shift all components by Count bits to the right (filling with zeros from the right)
UVec4::sReplicate
static JPH_INLINE UVec4 sReplicate(uint32 inV)
Replicate int inV across all components.
Definition UVec4.inl:56
UVec4::SplatX
JPH_INLINE UVec4 SplatX() const
Replicate the X component to all components.
Definition UVec4.inl:285
UVec4::Expand4Byte4
JPH_INLINE UVec4 Expand4Byte4() const
Takes byte 4 .. 7 and expands them to X, Y, Z and W.
Definition UVec4.inl:510
UVec4::TestAllTrue
JPH_INLINE bool TestAllTrue() const
Test if all components are true (true is when highest bit of component is set)
Definition UVec4.inl:408
UVec4::Expand4Byte0
JPH_INLINE UVec4 Expand4Byte0() const
Takes byte 0 .. 3 and expands them to X, Y, Z and W.
Definition UVec4.inl:495
UVec4::GetTrues
JPH_INLINE int GetTrues() const
Store if X is true in bit 0, Y in bit 1, Z in bit 2 and W in bit 3 (true is when highest bit of compo...
Definition UVec4.inl:386
UVec4::TestAnyXYZTrue
JPH_INLINE bool TestAnyXYZTrue() const
Test if any of X, Y or Z components are true (true is when highest bit of component is set)
Definition UVec4.inl:403
UVec4::operator+=
JPH_INLINE UVec4 & operator+=(UVec4Arg inV2)
Add two integer vectors (component wise)
Definition UVec4.inl:272
UVec4::sGatherInt4
static JPH_INLINE UVec4 sGatherInt4(const uint32 *inBase, UVec4Arg inOffsets)
Gather 4 ints from memory at inBase + inOffsets[i] * Scale.
UVec4::sAnd
static JPH_INLINE UVec4 sAnd(UVec4Arg inV1, UVec4Arg inV2)
Logical and (component wise)
Definition UVec4.inl:202
UVec4::Type
{ uint32 mData[4] Type
Definition UVec4.h:22
UVec4::sEquals
static JPH_INLINE UVec4 sEquals(UVec4Arg inV1, UVec4Arg inV2)
Equals (component wise)
Definition UVec4.inl:143
UVec4::sOr
static JPH_INLINE UVec4 sOr(UVec4Arg inV1, UVec4Arg inV2)
Logical or (component wise)
Definition UVec4.inl:174
UVec4::GetW
JPH_INLINE uint32 GetW() const
Definition UVec4.h:105
UVec4::TestAllXYZTrue
JPH_INLINE bool TestAllXYZTrue() const
Test if X, Y and Z components are true (true is when highest bit of component is set)
Definition UVec4.inl:413
UVec4::ShiftComponents4Minus
JPH_INLINE UVec4 ShiftComponents4Minus(int inCount) const
Shift vector components by 4 - Count floats to the left, so if Count = 1 the resulting vector is (W,...
Definition UVec4.inl:555
UVec4::operator==
JPH_INLINE bool operator==(UVec4Arg inV2) const
Comparison.
Definition UVec4.inl:23
UVec4::sMax
static JPH_INLINE UVec4 sMax(UVec4Arg inV1, UVec4Arg inV2)
Return the maximum of each of the components.
Definition UVec4.inl:129
UVec4::SplatZ
JPH_INLINE UVec4 SplatZ() const
Replicate the Z component to all components.
Definition UVec4.inl:307
UVec4::mValue
Type mValue
Definition UVec4.h:211
UVec4::SplatW
JPH_INLINE UVec4 SplatW() const
Replicate the W component to all components.
Definition UVec4.inl:318
UVec4::StoreInt4
JPH_INLINE void StoreInt4(uint32 *outV) const
Store 4 ints to memory.
Definition UVec4.inl:351
UVec4::GetX
JPH_INLINE uint32 GetX() const
Get individual components.
Definition UVec4.h:102
UVec4::Expand4Byte8
JPH_INLINE UVec4 Expand4Byte8() const
Takes byte 8 .. 11 and expands them to X, Y, Z and W.
Definition UVec4.inl:525
UVec4::sLoadInt4Aligned
static JPH_INLINE UVec4 sLoadInt4Aligned(const uint32 *inV)
Load 4 ints from memory, aligned to 16 bytes.
Definition UVec4.inl:89
UVec4::sLoadInt4
static JPH_INLINE UVec4 sLoadInt4(const uint32 *inV)
Load 4 ints from memory.
Definition UVec4.inl:78
UVec4::Expand4Byte12
JPH_INLINE UVec4 Expand4Byte12() const
Takes byte 12 .. 15 and expands them to X, Y, Z and W.
Definition UVec4.inl:540
UVec4::sXor
static JPH_INLINE UVec4 sXor(UVec4Arg inV1, UVec4Arg inV2)
Logical xor (component wise)
Definition UVec4.inl:188
UVec4::Expand4Uint16Hi
JPH_INLINE UVec4 Expand4Uint16Hi() const
Takes the upper 4 16 bits and expands them to X, Y, Z and W.
Definition UVec4.inl:479
UVec4::sZero
static JPH_INLINE UVec4 sZero()
Vector with all zeros.
Definition UVec4.inl:45
UVec4::operator+
JPH_INLINE UVec4 operator+(UVec4Arg inV2)
Adds an integer value to all integer components (discards any overflow)
Definition UVec4.inl:258
UVec4::ArithmeticShiftRight
JPH_INLINE UVec4 ArithmeticShiftRight() const
Shift all components by Count bits to the right (shifting in the value of the highest bit)
UVec4::UVec4
UVec4()=default
Constructor.
UVec4::operator*
JPH_INLINE UVec4 operator*(UVec4Arg inV2) const
Multiplies each of the 4 integer components with an integer (discards any overflow)
Definition UVec4.inl:244
UVec4::ToFloat
JPH_INLINE Vec4 ToFloat() const
Convert each component from an int to a float.
Definition UVec4.inl:329
UVec4::ReinterpretAsFloat
JPH_INLINE Vec4 ReinterpretAsFloat() const
Reinterpret UVec4 as a Vec4 (doesn't change the bits)
Definition UVec4.inl:340
UVec4::StoreInt4Aligned
JPH_INLINE void StoreInt4Aligned(uint32 *outV) const
Store 4 ints to memory, aligned to 16 bytes.
Definition UVec4.inl:363
UVec4::TestAnyTrue
JPH_INLINE bool TestAnyTrue() const
Test if any of the components are true (true is when highest bit of component is set)
Definition UVec4.inl:398
UVec4::mU32
uint32 mU32[4]
Definition UVec4.h:212
Vec4
Definition Vec4.h:14