AxioDL/metaforce
2
0
Fork 0
mirror of https://github.com/AxioDL/metaforce.git synced 2025-12-08 17:44:56 +00:00
Code Issues Projects Releases Wiki Activity

SIMD refactor

Browse Source
This commit is contained in:
Jack Andersen
2018-12-07 15:31:11 -10:00
parent 1b073abb76
commit 54c466276b
13 changed files with 107 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
hecl/lib/Blender/Connection.cpp
View File

@@ -830,6 +830,8 @@ void PyOutStream::linkBackground(const char* target, const char* sceneName)
void PyOutStream::AABBToBMesh(const atVec3f& min, const atVec3f& max)
{
athena::simd_floats minf(min.simd);
athena::simd_floats maxf(max.simd);
format("bm = bmesh.new()\n"
"bm.verts.new((%f,%f,%f))\n"
"bm.verts.new((%f,%f,%f))\n"
@@ -852,14 +854,14 @@ void PyOutStream::AABBToBMesh(const atVec3f& min, const atVec3f& max)
"bm.edges.new((bm.verts[6], bm.verts[2]))\n"
"bm.edges.new((bm.verts[6], bm.verts[4]))\n"
"bm.edges.new((bm.verts[6], bm.verts[7]))\n",
min.vec[0], min.vec[1], min.vec[2],
max.vec[0], min.vec[1], min.vec[2],
min.vec[0], max.vec[1], min.vec[2],
max.vec[0], max.vec[1], min.vec[2],
min.vec[0], min.vec[1], max.vec[2],
max.vec[0], min.vec[1], max.vec[2],
min.vec[0], max.vec[1], max.vec[2],
max.vec[0], max.vec[1], max.vec[2]);
minf[0], minf[1], minf[2],
maxf[0], minf[1], minf[2],
minf[0], maxf[1], minf[2],
maxf[0], maxf[1], minf[2],
minf[0], minf[1], maxf[2],
maxf[0], minf[1], maxf[2],
minf[0], maxf[1], maxf[2],
maxf[0], maxf[1], maxf[2]);
}
void PyOutStream::centerView()
@@ -2333,9 +2335,9 @@ DataStream::getBoneMatrices(std::string_view name)
{
float val;
m_parent->_readBuf(&val, 4);
matOut[i].vec[j] = val;
matOut[i].simd[j] = val;
}
reinterpret_cast<atVec4f&>(matOut[i]).vec[3] = 0.f;
reinterpret_cast<atVec4f&>(matOut[i]).simd[3] = 0.f;
}
ret.emplace(std::make_pair(std::move(name), std::move(matOut)));
@@ -2355,8 +2357,8 @@ bool DataStream::renderPvs(std::string_view path, const atVec3f& location)
m_parent->getBlendPath().getAbsolutePath().data());
char req[256];
snprintf(req, 256, "RENDERPVS %s %f %f %f", path.data(),
location.vec[0], location.vec[1], location.vec[2]);
athena::simd_floats f(location.simd);
snprintf(req, 256, "RENDERPVS %s %f %f %f", path.data(), f[0], f[1], f[2]);
m_parent->_writeStr(req);
char readBuf[256];

43
hecl/lib/Blender/HMDL.cpp
View File

@@ -11,18 +11,30 @@ namespace hecl::blender
atVec3f MtxVecMul4RM(const Matrix4f& mtx, const Vector3f& vec)
{
atVec3f res;
res.vec[0] = mtx[0].vec[0] * vec.val.vec[0] + mtx[0].vec[1] * vec.val.vec[1] + mtx[0].vec[2] * vec.val.vec[2] + mtx[0].vec[3];
res.vec[1] = mtx[1].vec[0] * vec.val.vec[0] + mtx[1].vec[1] * vec.val.vec[1] + mtx[1].vec[2] * vec.val.vec[2] + mtx[1].vec[3];
res.vec[2] = mtx[2].vec[0] * vec.val.vec[0] + mtx[2].vec[1] * vec.val.vec[1] + mtx[2].vec[2] * vec.val.vec[2] + mtx[2].vec[3];
athena::simd_floats resf;
athena::simd_floats mtxf[3];
for (int i = 0; i < 3; ++i)
mtx[i].simd.copy_to(mtxf[i]);
athena::simd_floats vecf(vec.val.simd);
resf[0] = mtxf[0][0] * vecf[0] + mtxf[0][1] * vecf[1] + mtxf[0][2] * vecf[2] + mtxf[0][3];
resf[1] = mtxf[1][0] * vecf[0] + mtxf[1][1] * vecf[1] + mtxf[1][2] * vecf[2] + mtxf[1][3];
resf[2] = mtxf[2][0] * vecf[0] + mtxf[2][1] * vecf[1] + mtxf[2][2] * vecf[2] + mtxf[2][3];
res.simd.copy_from(resf);
return res;
}
atVec3f MtxVecMul3RM(const Matrix4f& mtx, const Vector3f& vec)
{
atVec3f res;
res.vec[0] = mtx[0].vec[0] * vec.val.vec[0] + mtx[0].vec[1] * vec.val.vec[1] + mtx[0].vec[2] * vec.val.vec[2];
res.vec[1] = mtx[1].vec[0] * vec.val.vec[0] + mtx[1].vec[1] * vec.val.vec[1] + mtx[1].vec[2] * vec.val.vec[2];
res.vec[2] = mtx[2].vec[0] * vec.val.vec[0] + mtx[2].vec[1] * vec.val.vec[1] + mtx[2].vec[2] * vec.val.vec[2];
athena::simd_floats resf;
athena::simd_floats mtxf[3];
for (int i = 0; i < 3; ++i)
mtx[i].simd.copy_to(mtxf[i]);
athena::simd_floats vecf(vec.val.simd);
resf[0] = mtxf[0][0] * vecf[0] + mtxf[0][1] * vecf[1] + mtxf[0][2] * vecf[2];
resf[1] = mtxf[1][0] * vecf[0] + mtxf[1][1] * vecf[1] + mtxf[1][2] * vecf[2];
resf[2] = mtxf[2][0] * vecf[0] + mtxf[2][1] * vecf[1] + mtxf[2][2] * vecf[2];
res.simd.copy_from(resf);
return res;
}
@@ -113,15 +125,11 @@ HMDLBuffers Mesh::getHMDLBuffers(bool absoluteCoords, PoolSkinIndex& poolSkinInd
vboW.writeVec3fLittle(preXfPos);
atVec3f preXfNorm = MtxVecMul3RM(sceneXf, norm[v.iNorm]);
float mag =
preXfNorm.vec[0] * preXfNorm.vec[0] +
preXfNorm.vec[1] * preXfNorm.vec[1] +
preXfNorm.vec[2] * preXfNorm.vec[2];
athena::simd_floats f(preXfNorm.simd * preXfNorm.simd);
float mag = f[0] + f[1] + f[2];
if (mag > FLT_EPSILON)
mag = 1.f / std::sqrt(mag);
preXfNorm.vec[0] *= mag;
preXfNorm.vec[1] *= mag;
preXfNorm.vec[2] *= mag;
preXfNorm.simd *= mag;
vboW.writeVec3fLittle(preXfNorm);
}
else
@@ -133,9 +141,10 @@ HMDLBuffers Mesh::getHMDLBuffers(bool absoluteCoords, PoolSkinIndex& poolSkinInd
for (size_t i=0 ; i<colorLayerCount ; ++i)
{
const Vector3f& c = color[v.iColor[i]];
vboW.writeUByte(std::max(0, std::min(255, int(c.val.vec[0] * 255))));
vboW.writeUByte(std::max(0, std::min(255, int(c.val.vec[1] * 255))));
vboW.writeUByte(std::max(0, std::min(255, int(c.val.vec[2] * 255))));
athena::simd_floats f(c.val.simd);
vboW.writeUByte(std::max(0, std::min(255, int(f[0] * 255))));
vboW.writeUByte(std::max(0, std::min(255, int(f[1] * 255))));
vboW.writeUByte(std::max(0, std::min(255, int(f[2] * 255))));
vboW.writeUByte(255);
}
@@ -158,7 +167,7 @@ HMDLBuffers Mesh::getHMDLBuffers(bool absoluteCoords, PoolSkinIndex& poolSkinInd
for (const SkinBind& bind : binds)
if (bind.boneIdx == *it)
{
vec.vec[j] = bind.weight;
vec.simd[j] = bind.weight;
break;
}
++it;