At the same time, I multiply tow matrix using Eigen and standard C++. But the results are not the same, though very similar. I do not the reason. And I check all the details but no problem can be found.

This is due to floating-point rounding. With floating-point arithmetic, operators + and * are not associative anymore, so the result will depend on the operation order. Moreover, the result will also slightly differ whether you are using the FPU or SSE unit.

For example, matrix a and b are C++ style, and matrix c and d are eigen style. Now the sum of a*b and c*d is the same, but each coeff is not the same. For later derivation calculation, the different result will generate different results.

ggael wrote:This is due to floating-point rounding. With floating-point arithmetic, operators + and * are not associative anymore, so the result will depend on the operation order. Moreover, the result will also slightly differ whether you are using the FPU or SSE unit.

yes, that's normal.