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 inV1, UVec4Arg inV2, UVec4Arg inControl)
158{
159#if defined(JPH_USE_SSE4_1)
160 return _mm_castps_si128(_mm_blendv_ps(_mm_castsi128_ps(inV1.mValue), _mm_castsi128_ps(inV2.mValue), _mm_castsi128_ps(inControl.mValue)));
161#elif defined(JPH_USE_NEON)
162 return vbslq_u32(vreinterpretq_u32_s32(vshrq_n_s32(vreinterpretq_s32_u32(inControl.mValue), 31)), inV2.mValue, inV1.mValue);
163#else
164 UVec4 result;
165 for (int i = 0; i < 4; i++)
166 result.mU32[i] = inControl.mU32[i] ? inV2.mU32[i] : inV1.mU32[i];
167 return result;
168#endif
169}
170
171UVec4 UVec4::sOr(UVec4Arg inV1, UVec4Arg inV2)
172{
173#if defined(JPH_USE_SSE)
174 return _mm_or_si128(inV1.mValue, inV2.mValue);
175#elif defined(JPH_USE_NEON)
176 return vorrq_u32(inV1.mValue, inV2.mValue);
177#else
178 return UVec4(inV1.mU32[0] | inV2.mU32[0],
179 inV1.mU32[1] | inV2.mU32[1],
180 inV1.mU32[2] | inV2.mU32[2],
181 inV1.mU32[3] | inV2.mU32[3]);
182#endif
183}
184
185UVec4 UVec4::sXor(UVec4Arg inV1, UVec4Arg inV2)
186{
187#if defined(JPH_USE_SSE)
188 return _mm_xor_si128(inV1.mValue, inV2.mValue);
189#elif defined(JPH_USE_NEON)
190 return veorq_u32(inV1.mValue, inV2.mValue);
191#else
192 return UVec4(inV1.mU32[0] ^ inV2.mU32[0],
193 inV1.mU32[1] ^ inV2.mU32[1],
194 inV1.mU32[2] ^ inV2.mU32[2],
195 inV1.mU32[3] ^ inV2.mU32[3]);
196#endif
197}
198
199UVec4 UVec4::sAnd(UVec4Arg inV1, UVec4Arg inV2)
200{
201#if defined(JPH_USE_SSE)
202 return _mm_and_si128(inV1.mValue, inV2.mValue);
203#elif defined(JPH_USE_NEON)
204 return vandq_u32(inV1.mValue, inV2.mValue);
205#else
206 return UVec4(inV1.mU32[0] & inV2.mU32[0],
207 inV1.mU32[1] & inV2.mU32[1],
208 inV1.mU32[2] & inV2.mU32[2],
209 inV1.mU32[3] & inV2.mU32[3]);
210#endif
211}
212
213
214UVec4 UVec4::sNot(UVec4Arg inV1)
215{
216#if defined(JPH_USE_AVX512)
217 return _mm_ternarylogic_epi32(inV1.mValue, inV1.mValue, inV1.mValue, 0b01010101);
218#elif defined(JPH_USE_SSE)
219 return sXor(inV1, sReplicate(0xffffffff));
220#elif defined(JPH_USE_NEON)
221 return vmvnq_u32(inV1.mValue);
222#else
223 return UVec4(~inV1.mU32[0], ~inV1.mU32[1], ~inV1.mU32[2], ~inV1.mU32[3]);
224#endif
225}
226
227UVec4 UVec4::sSort4True(UVec4Arg inValue, UVec4Arg inIndex)
228{
229 // If inValue.z is false then shift W to Z
230 UVec4 v = UVec4::sSelect(inIndex.Swizzle<SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W, SWIZZLE_W>(), inIndex, inValue.SplatZ());
231
232 // If inValue.y is false then shift Z and further to Y and further
233 v = UVec4::sSelect(v.Swizzle<SWIZZLE_X, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W>(), v, inValue.SplatY());
234
235 // If inValue.x is false then shift X and further to Y and further
236 v = UVec4::sSelect(v.Swizzle<SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W>(), v, inValue.SplatX());
237
238 return v;
239}
240
241UVec4 UVec4::operator * (UVec4Arg inV2) const
242{
243#if defined(JPH_USE_SSE4_1)
244 return _mm_mullo_epi32(mValue, inV2.mValue);
245#elif defined(JPH_USE_NEON)
246 return vmulq_u32(mValue, inV2.mValue);
247#else
248 UVec4 result;
249 for (int i = 0; i < 4; i++)
250 result.mU32[i] = mU32[i] * inV2.mU32[i];
251 return result;
252#endif
253}
254
255UVec4 UVec4::operator + (UVec4Arg inV2)
256{
257#if defined(JPH_USE_SSE)
258 return _mm_add_epi32(mValue, inV2.mValue);
259#elif defined(JPH_USE_NEON)
260 return vaddq_u32(mValue, inV2.mValue);
261#else
262 return UVec4(mU32[0] + inV2.mU32[0],
263 mU32[1] + inV2.mU32[1],
264 mU32[2] + inV2.mU32[2],
265 mU32[3] + inV2.mU32[3]);
266#endif
267}
268
269UVec4 &UVec4::operator += (UVec4Arg inV2)
270{
271#if defined(JPH_USE_SSE)
272 mValue = _mm_add_epi32(mValue, inV2.mValue);
273#elif defined(JPH_USE_NEON)
274 mValue = vaddq_u32(mValue, inV2.mValue);
275#else
276 for (int i = 0; i < 4; ++i)
277 mU32[i] += inV2.mU32[i];
278#endif
279 return *this;
280}
281
282UVec4 UVec4::SplatX() const
283{
284#if defined(JPH_USE_SSE)
285 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(0, 0, 0, 0));
286#elif defined(JPH_USE_NEON)
287 return vdupq_laneq_u32(mValue, 0);
288#else
289 return UVec4(mU32[0], mU32[0], mU32[0], mU32[0]);
290#endif
291}
292
293UVec4 UVec4::SplatY() const
294{
295#if defined(JPH_USE_SSE)
296 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(1, 1, 1, 1));
297#elif defined(JPH_USE_NEON)
298 return vdupq_laneq_u32(mValue, 1);
299#else
300 return UVec4(mU32[1], mU32[1], mU32[1], mU32[1]);
301#endif
302}
303
304UVec4 UVec4::SplatZ() const
305{
306#if defined(JPH_USE_SSE)
307 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(2, 2, 2, 2));
308#elif defined(JPH_USE_NEON)
309 return vdupq_laneq_u32(mValue, 2);
310#else
311 return UVec4(mU32[2], mU32[2], mU32[2], mU32[2]);
312#endif
313}
314
315UVec4 UVec4::SplatW() const
316{
317#if defined(JPH_USE_SSE)
318 return _mm_shuffle_epi32(mValue, _MM_SHUFFLE(3, 3, 3, 3));
319#elif defined(JPH_USE_NEON)
320 return vdupq_laneq_u32(mValue, 3);
321#else
322 return UVec4(mU32[3], mU32[3], mU32[3], mU32[3]);
323#endif
324}
325
326Vec4 UVec4::ToFloat() const
327{
328#if defined(JPH_USE_SSE)
329 return _mm_cvtepi32_ps(mValue);
330#elif defined(JPH_USE_NEON)
331 return vcvtq_f32_u32(mValue);
332#else
333 return Vec4((float)mU32[0], (float)mU32[1], (float)mU32[2], (float)mU32[3]);
334#endif
335}
336
337Vec4 UVec4::ReinterpretAsFloat() const
338{
339#if defined(JPH_USE_SSE)
340 return Vec4(_mm_castsi128_ps(mValue));
341#elif defined(JPH_USE_NEON)
342 return vreinterpretq_f32_u32(mValue);
343#else
344 return *reinterpret_cast<const Vec4 *>(this);
345#endif
346}
347
348void UVec4::StoreInt4(uint32 *outV) const
349{
350#if defined(JPH_USE_SSE)
351 _mm_storeu_si128(reinterpret_cast<__m128i *>(outV), mValue);
352#elif defined(JPH_USE_NEON)
353 vst1q_u32(outV, mValue);
354#else
355 for (int i = 0; i < 4; ++i)
356 outV[i] = mU32[i];
357#endif
358}
359
360void UVec4::StoreInt4Aligned(uint32 *outV) const
361{
362#if defined(JPH_USE_SSE)
363 _mm_store_si128(reinterpret_cast<__m128i *>(outV), mValue);
364#elif defined(JPH_USE_NEON)
365 vst1q_u32(outV, mValue); // ARM doesn't make distinction between aligned or not
366#else
367 for (int i = 0; i < 4; ++i)
368 outV[i] = mU32[i];
369#endif
370}
371
372int UVec4::CountTrues() const
373{
374#if defined(JPH_USE_SSE)
375 return CountBits(_mm_movemask_ps(_mm_castsi128_ps(mValue)));
376#elif defined(JPH_USE_NEON)
377 return vaddvq_u32(vshrq_n_u32(mValue, 31));
378#else
379 return (mU32[0] >> 31) + (mU32[1] >> 31) + (mU32[2] >> 31) + (mU32[3] >> 31);
380#endif
381}
382
383int UVec4::GetTrues() const
384{
385#if defined(JPH_USE_SSE)
386 return _mm_movemask_ps(_mm_castsi128_ps(mValue));
387#elif defined(JPH_USE_NEON)
388 int32x4_t shift = JPH_NEON_INT32x4(0, 1, 2, 3);
389 return vaddvq_u32(vshlq_u32(vshrq_n_u32(mValue, 31), shift));
390#else
391 return (mU32[0] >> 31) | ((mU32[1] >> 31) << 1) | ((mU32[2] >> 31) << 2) | ((mU32[3] >> 31) << 3);
392#endif
393}
394
395bool UVec4::TestAnyTrue() const
396{
397 return GetTrues() != 0;
398}
399
400bool UVec4::TestAnyXYZTrue() const
401{
402 return (GetTrues() & 0b111) != 0;
403}
404
405bool UVec4::TestAllTrue() const
406{
407 return GetTrues() == 0b1111;
408}
409
410bool UVec4::TestAllXYZTrue() const
411{
412 return (GetTrues() & 0b111) == 0b111;
413}
414
415template <const uint Count>
416UVec4 UVec4::LogicalShiftLeft() const
417{
418 static_assert(Count <= 31, "Invalid shift");
419
420#if defined(JPH_USE_SSE)
421 return _mm_slli_epi32(mValue, Count);
422#elif defined(JPH_USE_NEON)
423 return vshlq_n_u32(mValue, Count);
424#else
425 return UVec4(mU32[0] << Count, mU32[1] << Count, mU32[2] << Count, mU32[3] << Count);
426#endif
427}
428
429template <const uint Count>
430UVec4 UVec4::LogicalShiftRight() const
431{
432 static_assert(Count <= 31, "Invalid shift");
433
434#if defined(JPH_USE_SSE)
435 return _mm_srli_epi32(mValue, Count);
436#elif defined(JPH_USE_NEON)
437 return vshrq_n_u32(mValue, Count);
438#else
439 return UVec4(mU32[0] >> Count, mU32[1] >> Count, mU32[2] >> Count, mU32[3] >> Count);
440#endif
441}
442
443template <const uint Count>
444UVec4 UVec4::ArithmeticShiftRight() const
445{
446 static_assert(Count <= 31, "Invalid shift");
447
448#if defined(JPH_USE_SSE)
449 return _mm_srai_epi32(mValue, Count);
450#elif defined(JPH_USE_NEON)
451 return vreinterpretq_u32_s32(vshrq_n_s32(vreinterpretq_s32_u32(mValue), Count));
452#else
453 return UVec4(uint32(int32_t(mU32[0]) >> Count),
454 uint32(int32_t(mU32[1]) >> Count),
455 uint32(int32_t(mU32[2]) >> Count),
456 uint32(int32_t(mU32[3]) >> Count));
457#endif
458}
459
460UVec4 UVec4::Expand4Uint16Lo() const
461{
462#if defined(JPH_USE_SSE)
463 return _mm_unpacklo_epi16(mValue, _mm_castps_si128(_mm_setzero_ps()));
464#elif defined(JPH_USE_NEON)
465 uint16x4_t value = vget_low_u16(vreinterpretq_u16_u32(mValue));
466 uint16x4_t zero = vdup_n_u16(0);
467 return vreinterpretq_u32_u16(vcombine_u16(vzip1_u16(value, zero), vzip2_u16(value, zero)));
468#else
469 return UVec4(mU32[0] & 0xffff,
470 (mU32[0] >> 16) & 0xffff,
471 mU32[1] & 0xffff,
472 (mU32[1] >> 16) & 0xffff);
473#endif
474}
475
476UVec4 UVec4::Expand4Uint16Hi() const
477{
478#if defined(JPH_USE_SSE)
479 return _mm_unpackhi_epi16(mValue, _mm_castps_si128(_mm_setzero_ps()));
480#elif defined(JPH_USE_NEON)
481 uint16x4_t value = vget_high_u16(vreinterpretq_u16_u32(mValue));
482 uint16x4_t zero = vdup_n_u16(0);
483 return vreinterpretq_u32_u16(vcombine_u16(vzip1_u16(value, zero), vzip2_u16(value, zero)));
484#else
485 return UVec4(mU32[2] & 0xffff,
486 (mU32[2] >> 16) & 0xffff,
487 mU32[3] & 0xffff,
488 (mU32[3] >> 16) & 0xffff);
489#endif
490}
491
492UVec4 UVec4::Expand4Byte0() const
493{
494#if defined(JPH_USE_SSE4_1)
495 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff03), int(0xffffff02), int(0xffffff01), int(0xffffff00)));
496#elif defined(JPH_USE_NEON)
497 uint8x16_t idx = JPH_NEON_UINT8x16(0x00, 0x7f, 0x7f, 0x7f, 0x01, 0x7f, 0x7f, 0x7f, 0x02, 0x7f, 0x7f, 0x7f, 0x03, 0x7f, 0x7f, 0x7f);
498 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
499#else
500 UVec4 result;
501 for (int i = 0; i < 4; i++)
502 result.mU32[i] = (mU32[0] >> (i * 8)) & 0xff;
503 return result;
504#endif
505}
506
507UVec4 UVec4::Expand4Byte4() const
508{
509#if defined(JPH_USE_SSE4_1)
510 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff07), int(0xffffff06), int(0xffffff05), int(0xffffff04)));
511#elif defined(JPH_USE_NEON)
512 uint8x16_t idx = JPH_NEON_UINT8x16(0x04, 0x7f, 0x7f, 0x7f, 0x05, 0x7f, 0x7f, 0x7f, 0x06, 0x7f, 0x7f, 0x7f, 0x07, 0x7f, 0x7f, 0x7f);
513 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
514#else
515 UVec4 result;
516 for (int i = 0; i < 4; i++)
517 result.mU32[i] = (mU32[1] >> (i * 8)) & 0xff;
518 return result;
519#endif
520}
521
522UVec4 UVec4::Expand4Byte8() const
523{
524#if defined(JPH_USE_SSE4_1)
525 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff0b), int(0xffffff0a), int(0xffffff09), int(0xffffff08)));
526#elif defined(JPH_USE_NEON)
527 uint8x16_t idx = JPH_NEON_UINT8x16(0x08, 0x7f, 0x7f, 0x7f, 0x09, 0x7f, 0x7f, 0x7f, 0x0a, 0x7f, 0x7f, 0x7f, 0x0b, 0x7f, 0x7f, 0x7f);
528 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
529#else
530 UVec4 result;
531 for (int i = 0; i < 4; i++)
532 result.mU32[i] = (mU32[2] >> (i * 8)) & 0xff;
533 return result;
534#endif
535}
536
537UVec4 UVec4::Expand4Byte12() const
538{
539#if defined(JPH_USE_SSE4_1)
540 return _mm_shuffle_epi8(mValue, _mm_set_epi32(int(0xffffff0f), int(0xffffff0e), int(0xffffff0d), int(0xffffff0c)));
541#elif defined(JPH_USE_NEON)
542 uint8x16_t idx = JPH_NEON_UINT8x16(0x0c, 0x7f, 0x7f, 0x7f, 0x0d, 0x7f, 0x7f, 0x7f, 0x0e, 0x7f, 0x7f, 0x7f, 0x0f, 0x7f, 0x7f, 0x7f);
543 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
544#else
545 UVec4 result;
546 for (int i = 0; i < 4; i++)
547 result.mU32[i] = (mU32[3] >> (i * 8)) & 0xff;
548 return result;
549#endif
550}
551
552UVec4 UVec4::ShiftComponents4Minus(int inCount) const
553{
554#if defined(JPH_USE_SSE4_1) || defined(JPH_USE_NEON)
555 alignas(UVec4) static constexpr uint32 sFourMinusXShuffle[5][4] =
556 {
557 { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff },
558 { 0x0f0e0d0c, 0xffffffff, 0xffffffff, 0xffffffff },
559 { 0x0b0a0908, 0x0f0e0d0c, 0xffffffff, 0xffffffff },
560 { 0x07060504, 0x0b0a0908, 0x0f0e0d0c, 0xffffffff },
561 { 0x03020100, 0x07060504, 0x0b0a0908, 0x0f0e0d0c }
562 };
563#endif
564
565#if defined(JPH_USE_SSE4_1)
566 return _mm_shuffle_epi8(mValue, *reinterpret_cast<const UVec4::Type *>(sFourMinusXShuffle[inCount]));
567#elif defined(JPH_USE_NEON)
568 uint8x16_t idx = vreinterpretq_u8_u32(*reinterpret_cast<const UVec4::Type *>(sFourMinusXShuffle[inCount]));
569 return vreinterpretq_u32_s8(vqtbl1q_s8(vreinterpretq_s8_u32(mValue), idx));
570#else
571 UVec4 result = UVec4::sZero();
572 for (int i = 0; i < inCount; i++)
573 result.mU32[i] = mU32[i + 4 - inCount];
574 return result;
575#endif
576}
577
578JPH_NAMESPACE_END
uint8
std::uint8_t uint8
Definition Core.h:453
uint64
std::uint64_t uint64
Definition Core.h:456
JPH_NAMESPACE_END
#define JPH_NAMESPACE_END
Definition Core.h:378
uint32
std::uint32_t uint32
Definition Core.h:455
JPH_NAMESPACE_BEGIN
#define JPH_NAMESPACE_BEGIN
Definition Core.h:372
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:214
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:372
UVec4::SplatY
JPH_INLINE UVec4 SplatY() const
Replicate the Y component to all components.
Definition UVec4.inl:293
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:460
UVec4::sSort4True
static JPH_INLINE UVec4 sSort4True(UVec4Arg inValue, UVec4Arg inIndex)
Definition UVec4.inl:227
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:282
UVec4::Expand4Byte4
JPH_INLINE UVec4 Expand4Byte4() const
Takes byte 4 .. 7 and expands them to X, Y, Z and W.
Definition UVec4.inl:507
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:405
UVec4::Expand4Byte0
JPH_INLINE UVec4 Expand4Byte0() const
Takes byte 0 .. 3 and expands them to X, Y, Z and W.
Definition UVec4.inl:492
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:383
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:400
UVec4::operator+=
JPH_INLINE UVec4 & operator+=(UVec4Arg inV2)
Add two integer vectors (component wise)
Definition UVec4.inl:269
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:199
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:171
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:410
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:552
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:304
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:315
UVec4::StoreInt4
JPH_INLINE void StoreInt4(uint32 *outV) const
Store 4 ints to memory.
Definition UVec4.inl:348
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:522
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:537
UVec4::sXor
static JPH_INLINE UVec4 sXor(UVec4Arg inV1, UVec4Arg inV2)
Logical xor (component wise)
Definition UVec4.inl:185
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:476
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:255
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:241
UVec4::sSelect
static JPH_INLINE UVec4 sSelect(UVec4Arg inV1, UVec4Arg inV2, UVec4Arg inControl)
Component wise select, returns inV1 when highest bit of inControl = 0 and inV2 when highest bit of in...
Definition UVec4.inl:157
UVec4::ToFloat
JPH_INLINE Vec4 ToFloat() const
Convert each component from an int to a float.
Definition UVec4.inl:326
UVec4::ReinterpretAsFloat
JPH_INLINE Vec4 ReinterpretAsFloat() const
Reinterpret UVec4 as a Vec4 (doesn't change the bits)
Definition UVec4.inl:337
UVec4::StoreInt4Aligned
JPH_INLINE void StoreInt4Aligned(uint32 *outV) const
Store 4 ints to memory, aligned to 16 bytes.
Definition UVec4.inl:360
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:395
UVec4::mU32
uint32 mU32[4]
Definition UVec4.h:212
Vec4
Definition Vec4.h:14