GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-05-02 00:34:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merged eigen/eigen into default

Browse Source
This commit is contained in:
Benoit Steiner 2016-12-01 07:13:17 -08:00
commit 7ff26ddcbb
4 changed files with 49 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Eigen/src/Core/products/GeneralBlockPanelKernel.h
View File

@ -972,7 +972,7 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
EIGEN_ASM_COMMENT("begin step of gebp micro kernel 3pX4"); \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 3pX4"); \
EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \
internal::prefetch(blA+(3*K+16)*LhsProgress); \ internal::prefetch(blA+(3*K+16)*LhsProgress); \
if (EIGEN_ARCH_ARM) internal::prefetch(blB+(4*K+16)*RhsProgress); /* Bug 953 */ \ if (EIGEN_ARCH_ARM) { internal::prefetch(blB+(4*K+16)*RhsProgress); } /* Bug 953 */ \
traits.loadLhs(&blA[(0+3*K)*LhsProgress], A0); \ traits.loadLhs(&blA[(0+3*K)*LhsProgress], A0); \
traits.loadLhs(&blA[(1+3*K)*LhsProgress], A1); \ traits.loadLhs(&blA[(1+3*K)*LhsProgress], A1); \
traits.loadLhs(&blA[(2+3*K)*LhsProgress], A2); \ traits.loadLhs(&blA[(2+3*K)*LhsProgress], A2); \

16
Eigen/src/SparseCore/SparseMatrix.h
View File

@ -786,30 +786,38 @@ class SparseMatrix
EIGEN_DBG_SPARSE( EIGEN_DBG_SPARSE(
s << "Nonzero entries:\n"; s << "Nonzero entries:\n";
if(m.isCompressed()) if(m.isCompressed())
{
for (Index i=0; i<m.nonZeros(); ++i) for (Index i=0; i<m.nonZeros(); ++i)
s << "(" << m.m_data.value(i) << "," << m.m_data.index(i) << ") "; s << "(" << m.m_data.value(i) << "," << m.m_data.index(i) << ") ";
}
else else
{
for (Index i=0; i<m.outerSize(); ++i) for (Index i=0; i<m.outerSize(); ++i)
{ {
Index p = m.m_outerIndex[i]; Index p = m.m_outerIndex[i];
Index pe = m.m_outerIndex[i]+m.m_innerNonZeros[i]; Index pe = m.m_outerIndex[i]+m.m_innerNonZeros[i];
Index k=p; Index k=p;
for (; k<pe; ++k) for (; k<pe; ++k) {
s << "(" << m.m_data.value(k) << "," << m.m_data.index(k) << ") "; s << "(" << m.m_data.value(k) << "," << m.m_data.index(k) << ") ";
for (; k<m.m_outerIndex[i+1]; ++k) }
for (; k<m.m_outerIndex[i+1]; ++k) {
s << "(_,_) "; s << "(_,_) ";
} }
}
}
s << std::endl; s << std::endl;
s << std::endl; s << std::endl;
s << "Outer pointers:\n"; s << "Outer pointers:\n";
for (Index i=0; i<m.outerSize(); ++i) for (Index i=0; i<m.outerSize(); ++i) {
s << m.m_outerIndex[i] << " "; s << m.m_outerIndex[i] << " ";
}
s << " $" << std::endl; s << " $" << std::endl;
if(!m.isCompressed()) if(!m.isCompressed())
{ {
s << "Inner non zeros:\n"; s << "Inner non zeros:\n";
for (Index i=0; i<m.outerSize(); ++i) for (Index i=0; i<m.outerSize(); ++i) {
s << m.m_innerNonZeros[i] << " "; s << m.m_innerNonZeros[i] << " ";
}
s << " $" << std::endl; s << " $" << std::endl;
} }
s << std::endl; s << std::endl;

21
Eigen/src/SparseLU/SparseLU_gemm_kernel.h
View File

@ -114,12 +114,12 @@ void sparselu_gemm(Index m, Index n, Index d, const Scalar* A, Index lda, const
KMADD(c0, a1, b10, t0) \ KMADD(c0, a1, b10, t0) \
KMADD(c1, a1, b11, t1) \ KMADD(c1, a1, b11, t1) \
a1 = pload<Packet>(A1+i+(I+1)*PacketSize); \ a1 = pload<Packet>(A1+i+(I+1)*PacketSize); \
if(RK==4) KMADD(c0, a2, b20, t0) \ if(RK==4){ KMADD(c0, a2, b20, t0) }\
if(RK==4) KMADD(c1, a2, b21, t1) \ if(RK==4){ KMADD(c1, a2, b21, t1) }\
if(RK==4) a2 = pload<Packet>(A2+i+(I+1)*PacketSize); \ if(RK==4){ a2 = pload<Packet>(A2+i+(I+1)*PacketSize); }\
if(RK==4) KMADD(c0, a3, b30, t0) \ if(RK==4){ KMADD(c0, a3, b30, t0) }\
if(RK==4) KMADD(c1, a3, b31, t1) \ if(RK==4){ KMADD(c1, a3, b31, t1) }\
if(RK==4) a3 = pload<Packet>(A3+i+(I+1)*PacketSize); \ if(RK==4){ a3 = pload<Packet>(A3+i+(I+1)*PacketSize); }\
pstore(C0+i+(I)*PacketSize, c0); \ pstore(C0+i+(I)*PacketSize, c0); \
pstore(C1+i+(I)*PacketSize, c1) pstore(C1+i+(I)*PacketSize, c1)
@ -131,6 +131,7 @@ void sparselu_gemm(Index m, Index n, Index d, const Scalar* A, Index lda, const
prefetch((A1+i+(5)*PacketSize)); prefetch((A1+i+(5)*PacketSize));
if(RK==4) prefetch((A2+i+(5)*PacketSize)); if(RK==4) prefetch((A2+i+(5)*PacketSize));
if(RK==4) prefetch((A3+i+(5)*PacketSize)); if(RK==4) prefetch((A3+i+(5)*PacketSize));
WORK(0); WORK(0);
WORK(1); WORK(1);
WORK(2); WORK(2);
@ -208,10 +209,10 @@ void sparselu_gemm(Index m, Index n, Index d, const Scalar* A, Index lda, const
a0 = pload<Packet>(A0+i+(I+1)*PacketSize); \ a0 = pload<Packet>(A0+i+(I+1)*PacketSize); \
KMADD(c0, a1, b10, t0) \ KMADD(c0, a1, b10, t0) \
a1 = pload<Packet>(A1+i+(I+1)*PacketSize); \ a1 = pload<Packet>(A1+i+(I+1)*PacketSize); \
if(RK==4) KMADD(c0, a2, b20, t0) \ if(RK==4){ KMADD(c0, a2, b20, t0) }\
if(RK==4) a2 = pload<Packet>(A2+i+(I+1)*PacketSize); \ if(RK==4){ a2 = pload<Packet>(A2+i+(I+1)*PacketSize); }\
if(RK==4) KMADD(c0, a3, b30, t0) \ if(RK==4){ KMADD(c0, a3, b30, t0) }\
if(RK==4) a3 = pload<Packet>(A3+i+(I+1)*PacketSize); \ if(RK==4){ a3 = pload<Packet>(A3+i+(I+1)*PacketSize); }\
pstore(C0+i+(I)*PacketSize, c0); pstore(C0+i+(I)*PacketSize, c0);
// agressive vectorization and peeling // agressive vectorization and peeling

2
test/main.h
View File

@ -90,7 +90,7 @@ inline void on_temporary_creation(long int size) {
#define VERIFY_EVALUATION_COUNT(XPR,N) {\ #define VERIFY_EVALUATION_COUNT(XPR,N) {\
nb_temporaries = 0; \ nb_temporaries = 0; \
XPR; \ XPR; \
if(nb_temporaries!=N) std::cerr << "nb_temporaries == " << nb_temporaries << "\n"; \ if(nb_temporaries!=N) { std::cerr << "nb_temporaries == " << nb_temporaries << "\n"; }\
VERIFY( (#XPR) && nb_temporaries==N ); \ VERIFY( (#XPR) && nb_temporaries==N ); \
} }