JCS-Prod/ProtoTank
1
0
Fork 0
Code Issues Pull Requests Projects Releases Activity

Join perspective divide with projection transform

Browse Source
This commit is contained in:
JackCarterSmith 2024-10-03 18:11:25 +02:00
parent 9ff5d34908
commit 294f806900
Signed by: JackCarterSmith
GPG Key ID: 832E52F4E23F8F24
4 changed files with 360 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
Engine/Graphics/3DRenderer.cpp
View File

@ -39,10 +39,11 @@ void Graphic3DRenderer::Draw(sf::RenderTexture& context) {
M3D_MATRIX MVPMat = modelMat * viewProjMat;
M3D_MATRIX viewportMat = M3D_TransformMatrixViewport(1280.0f, 324.f, 0.0f, 0.0f);
sf::Vertex v_tri[4];
auto cubeMesh = testObj.GetObjectMesh();
M3D_F4 projVertices[cubeMesh.vertices.size()];
M3D_V3Transform(projVertices, sizeof(M3D_F4), (M3D_F3*)cubeMesh.vertices.data(), sizeof(Vertex), cubeMesh.vertices.size(), MVPMat);
// Do the vertices projection and perspective divide
M3D_F3 projVertices[cubeMesh.vertices.size()];
M3D_V3TransformPersDiv(projVertices, sizeof(M3D_F3), (M3D_F3*)cubeMesh.vertices.data(), sizeof(Vertex), cubeMesh.vertices.size(), MVPMat);
auto indicePtr = (uint32_t*)cubeMesh.parts[0].indices.data();
for (uint32_t i = 0; i < cubeMesh.parts[0].indices.size(); i += 3) {
@ -50,17 +51,15 @@ void Graphic3DRenderer::Draw(sf::RenderTexture& context) {
if (i+2 > cubeMesh.parts[0].indices.size())
break;
// Simple clipping
//TODO: implement complete Cohen-Sutherland algo or similar
if ((projVertices[indicePtr[i]]).z > 0 &&
(projVertices[indicePtr[i+1]]).z > 0 &&
(projVertices[indicePtr[i+2]]).z > 0) {
M3D_VECTOR V1 = M3D_V4LoadF4(&projVertices[indicePtr[i]]);
M3D_VECTOR V2 = M3D_V4LoadF4(&projVertices[indicePtr[i+1]]);
M3D_VECTOR V3 = M3D_V4LoadF4(&projVertices[indicePtr[i+2]]);
V1 = M3D_V4Divide(V1, M3D_V4SplatW(V1));
V2 = M3D_V4Divide(V2, M3D_V4SplatW(V2));
V3 = M3D_V4Divide(V3, M3D_V4SplatW(V3));
M3D_VECTOR V1 = M3D_V4LoadF3(&projVertices[indicePtr[i]]);
M3D_VECTOR V2 = M3D_V4LoadF3(&projVertices[indicePtr[i+1]]);
M3D_VECTOR V3 = M3D_V4LoadF3(&projVertices[indicePtr[i+2]]);
V1 = M3D_V3Transform(V1, viewportMat);
V2 = M3D_V3Transform(V2, viewportMat);

19
Engine/Utils/3DMaths.hpp
View File

@ -42,10 +42,17 @@
#endif
#define M3D_UNPACK3INTO4(l1, l2, l3) \
M3D_VECTOR V3 = _mm_shuffle_ps(l2, l3, _MM_SHUFFLE(0, 0, 3, 2));\
M3D_VECTOR V2 = _mm_shuffle_ps(l2, l1, _MM_SHUFFLE(3, 3, 1, 0));\
V2 = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 0, 2));\
M3D_VECTOR V4 = _mm_castsi128_ps(_mm_srli_si128(_mm_castps_si128(L3), 32 / 8))
M3D_VECTOR V3 = _mm_shuffle_ps(l2, l3, _MM_SHUFFLE(0, 0, 3, 2));\
M3D_VECTOR V2 = _mm_shuffle_ps(l2, l1, _MM_SHUFFLE(3, 3, 1, 0));\
V2 = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 0, 2));\
M3D_VECTOR V4 = _mm_castsi128_ps(_mm_srli_si128(_mm_castps_si128(L3), 32 / 8))
#define M3D_PACK4INTO3(v2x) \
v2x = _mm_shuffle_ps(V2, V3, _MM_SHUFFLE(1, 0, 2, 1));\
V2 = _mm_shuffle_ps(V2, V1, _MM_SHUFFLE(2, 2, 0, 0));\
V1 = _mm_shuffle_ps(V1, V2, _MM_SHUFFLE(0, 2, 1, 0));\
V3 = _mm_shuffle_ps(V3, V4, _MM_SHUFFLE(0, 0, 2, 2));\
V3 = _mm_shuffle_ps(V3, V4, _MM_SHUFFLE(2, 1, 2, 0))
#endif
//
@ -339,7 +346,9 @@ M3D_MATRIX M3D_MTranspose(M3D_MATRIX M) noexcept;
// Vector/Matrix operation
//
M3D_VECTOR M3D_V3Transform(M3D_VECTOR V, M3D_MATRIX M) noexcept;
M3D_F4* M3D_V3Transform(M3D_F4* pOutputStream, size_t OutputStride, const M3D_F3* pInputStream, size_t InputStride, size_t VectorCount, M3D_MATRIX M) noexcept;
void M3D_V3Transform(M3D_F4* pOutputStream, size_t OutputStride, const M3D_F3* pInputStream, size_t InputStride, size_t VectorCount, M3D_MATRIX M) noexcept;
M3D_VECTOR M3D_V3TransformPersDiv(M3D_VECTOR V, M3D_MATRIX M) noexcept;
void M3D_V3TransformPersDiv(M3D_F3* pOutputStream, size_t OutputStride, const M3D_F3* pInputStream, size_t InputStride, size_t VectorCount, M3D_MATRIX M) noexcept;
//

331
Engine/Utils/3DMaths.inl
View File

@ -1130,7 +1130,7 @@ inline M3D_VECTOR M3D_V3Transform(M3D_VECTOR V, M3D_MATRIX M) noexcept {
#endif
}
inline M3D_F4* M3D_V3Transform(
inline void M3D_V3Transform(
M3D_F4* pOutputStream,
size_t OutputStride,
const M3D_F3* pInputStream,
@ -1162,8 +1162,6 @@ inline M3D_F4* M3D_V3Transform(
pInputVector += InputStride;
pOutputVector += OutputStride;
}
return pOutputStream;
#else
size_t i = 0;
size_t four = VectorCount >> 2;
@ -1346,8 +1344,333 @@ inline M3D_F4* M3D_V3Transform(
}
M3D_SFENCE();
#endif
}
return pOutputStream;
inline M3D_VECTOR M3D_V3TransformPersDiv(M3D_VECTOR V, M3D_MATRIX M) noexcept {
M3D_VECTOR Z = M3D_V4SplatZ(V);
M3D_VECTOR Y = M3D_V4SplatY(V);
M3D_VECTOR X = M3D_V4SplatX(V);
M3D_VECTOR Result = M3D_V4MultiplyAdd(Z, M.rows[2], M.rows[3]);
Result = M3D_V4MultiplyAdd(Y, M.rows[1], Result);
Result = M3D_V4MultiplyAdd(X, M.rows[0], Result);
M3D_VECTOR W = M3D_V4SplatW(Result);
return M3D_V4Divide(Result, W);
}
inline void M3D_V3TransformPersDiv(
M3D_F3* pOutputStream,
size_t OutputStride,
const M3D_F3* pInputStream,
size_t InputStride,
size_t VectorCount,
M3D_MATRIX M
) noexcept {
auto pInputVector = reinterpret_cast<const uint8_t*>(pInputStream);
auto pOutputVector = reinterpret_cast<uint8_t*>(pOutputStream);
const M3D_VECTOR row0 = M.rows[0];
const M3D_VECTOR row1 = M.rows[1];
const M3D_VECTOR row2 = M.rows[2];
const M3D_VECTOR row3 = M.rows[3];
#ifdef DISABLE_INTRINSICS
for (size_t i = 0; i < VectorCount; i++)
{
M3D_VECTOR V = M3D_V4LoadF3(reinterpret_cast<const M3D_F3*>(pInputVector));
M3D_VECTOR Z = M3D_V4SplatZ(V);
M3D_VECTOR Y = M3D_V4SplatY(V);
M3D_VECTOR X = M3D_V4SplatX(V);
M3D_VECTOR Result = M3D_V4MultiplyAdd(Z, row2, row3);
Result = M3D_V4MultiplyAdd(Y, row1, Result);
Result = M3D_V4MultiplyAdd(X, row0, Result);
M3D_VECTOR W = M3D_V4SplatW(Result);
Result = M3D_V4Divide(Result, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), Result);
pInputVector += InputStride;
pOutputVector += OutputStride;
}
#else
size_t i = 0;
size_t four = VectorCount >> 2;
if (four > 0) {
if (InputStride == sizeof(M3D_F3)) {
if (OutputStride == sizeof(M3D_F3)) {
if (!(reinterpret_cast<uintptr_t>(pOutputStream) & 0xF)) {
// Packed input, aligned & packed output
for (size_t j = 0; j < four; ++j) {
__m128 V1 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector));
__m128 L2 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 16));
__m128 L3 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 32));
pInputVector += sizeof(M3D_F3) * 4;
// Unpack the 4 vectors (.w components are junk)
M3D_UNPACK3INTO4(V1, L2, L3);
// Result 1
M3D_VECTOR Z = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2));
M3D_VECTOR Y = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1));
M3D_VECTOR X = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0));
M3D_VECTOR vTemp = M3D_FMADD_PS(Z, row2, row3);
M3D_VECTOR vTemp2 = _mm_mul_ps(Y, row1);
M3D_VECTOR vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
M3D_VECTOR W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V1 = _mm_div_ps(vTemp, W);
// Result 2
Z = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V2 = _mm_div_ps(vTemp, W);
// Result 3
Z = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V3 = _mm_div_ps(vTemp, W);
// Result 4
Z = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V4 = _mm_div_ps(vTemp, W);
// Pack and store the vectors
M3D_PACK4INTO3(vTemp);
M3D_STREAM_PS(reinterpret_cast<float*>(pOutputVector), V1);
M3D_STREAM_PS(reinterpret_cast<float*>(pOutputVector + 16), vTemp);
M3D_STREAM_PS(reinterpret_cast<float*>(pOutputVector + 32), V3);
pOutputVector += sizeof(M3D_F3) * 4;
i += 4;
}
} else {
// Packed input, unaligned & packed output
for (size_t j = 0; j < four; ++j) {
__m128 V1 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector));
__m128 L2 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 16));
__m128 L3 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 32));
pInputVector += sizeof(M3D_F3) * 4;
// Unpack the 4 vectors (.w components are junk)
M3D_UNPACK3INTO4(V1, L2, L3);
// Result 1
M3D_VECTOR Z = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2));
M3D_VECTOR Y = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1));
M3D_VECTOR X = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0));
M3D_VECTOR vTemp = M3D_FMADD_PS(Z, row2, row3);
M3D_VECTOR vTemp2 = _mm_mul_ps(Y, row1);
M3D_VECTOR vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
M3D_VECTOR W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V1 = _mm_div_ps(vTemp, W);
// Result 2
Z = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V2 = _mm_div_ps(vTemp, W);
// Result 3
Z = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V3 = _mm_div_ps(vTemp, W);
// Result 4
Z = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
V4 = _mm_div_ps(vTemp, W);
// Pack and store the vectors
M3D_PACK4INTO3(vTemp);
_mm_storeu_ps(reinterpret_cast<float*>(pOutputVector), V1);
_mm_storeu_ps(reinterpret_cast<float*>(pOutputVector + 16), vTemp);
_mm_storeu_ps(reinterpret_cast<float*>(pOutputVector + 32), V3);
pOutputVector += sizeof(M3D_F3) * 4;
i += 4;
}
}
} else {
// Packed input, unpacked output
for (size_t j = 0; j < four; ++j)
{
__m128 V1 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector));
__m128 L2 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 16));
__m128 L3 = _mm_loadu_ps(reinterpret_cast<const float*>(pInputVector + 32));
pInputVector += sizeof(M3D_F3) * 4;
// Unpack the 4 vectors (.w components are junk)
M3D_UNPACK3INTO4(V1, L2, L3);
// Result 1
M3D_VECTOR Z = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2));
M3D_VECTOR Y = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1));
M3D_VECTOR X = M3D_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0));
M3D_VECTOR vTemp = M3D_FMADD_PS(Z, row2, row3);
M3D_VECTOR vTemp2 = _mm_mul_ps(Y, row1);
M3D_VECTOR vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
M3D_VECTOR W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
vTemp = _mm_div_ps(vTemp, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), vTemp);
pOutputVector += OutputStride;
// Result 2
Z = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
vTemp = _mm_div_ps(vTemp, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), vTemp);
pOutputVector += OutputStride;
// Result 3
Z = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
vTemp = _mm_div_ps(vTemp, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), vTemp);
pOutputVector += OutputStride;
// Result 4
Z = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2));
Y = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1));
X = M3D_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0));
vTemp = M3D_FMADD_PS(Z, row2, row3);
vTemp2 = _mm_mul_ps(Y, row1);
vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
vTemp = _mm_div_ps(vTemp, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), vTemp);
pOutputVector += OutputStride;
i += 4;
}
}
}
}
for (; i < VectorCount; i++) {
M3D_VECTOR V = M3D_V4LoadF3(reinterpret_cast<const M3D_F3*>(pInputVector));
pInputVector += InputStride;
M3D_VECTOR Z = M3D_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2));
M3D_VECTOR Y = M3D_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1));
M3D_VECTOR X = M3D_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0));
M3D_VECTOR vTemp = M3D_FMADD_PS(Z, row2, row3);
M3D_VECTOR vTemp2 = _mm_mul_ps(Y, row1);
M3D_VECTOR vTemp3 = _mm_mul_ps(X, row0);
vTemp = _mm_add_ps(vTemp, vTemp2);
vTemp = _mm_add_ps(vTemp, vTemp3);
M3D_VECTOR W = M3D_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3));
vTemp = _mm_div_ps(vTemp, W);
M3D_V4StoreF3(reinterpret_cast<M3D_F3*>(pOutputVector), vTemp);
pOutputVector += OutputStride;
}
M3D_SFENCE();
#endif
}

13
Engine/World/WorldObject.hpp
View File

@ -5,14 +5,21 @@
#include "../Utils/MeshHelper.hpp"
class WorldObject {};
class WorldObject {
public:
virtual const Mesh& GetObjectMesh() const = 0;
protected:
WorldObject() = default;
};
template<class D>
class WorldObjectAbstract : public WorldObject {
class WorldObjectAbstract : virtual public WorldObject {
public:
virtual ~WorldObjectAbstract() = 0;
const Mesh& GetObjectMesh() const noexcept { return mMesh; }
const Mesh& GetObjectMesh() const noexcept override { return mMesh; }
protected:
inline static Mesh mMesh;