|
✍ Additional Ostinato build parameters.
It could be interesting to investigate the assembly listings produced by building with XD_XMTX_CONCAT_VEC and XD_USE_LA flags XD_XMTX_CONCAT_VEC allows to choose an xt_xmtx (3x4 transformation matrix) concatenation implementation: By default it is XD_XMTX_CONCAT_VEC=0 - a 'direct' implementaion without loops. CXX=clang++ ./build.sh ... -DXD_XMTX_CONCAT_VEC=0|1|2
Crosscore contains a set of templates for linear algebra operations.
By default 4x4 matrix operations, such as multiplication, vector/point trasnformation and matrix inversion, make use of them.
XD_USE_LA allows to switch this off.
CXX=clang++ ./build.sh ... -DXD_USE_LA=0|1
Note:
Clang produces an erronous code when the sources are compiled for webassembly with simd instructions enabled and XD_USE_LA=0. Can be some Clang/LLVM problem.
./build.sh wasm -msimd128 -DXD_USE_LA=0
|