Fix offset bug in matrix multiply cascade #5302

Job	Run time
ubuntu-22.04 llvm assert=ON rtti=ON	9m 30s
ubuntu-22.04 llvm assert=ON rtti=OFF	9m 33s
ubuntu-22.04 llvm assert=OFF rtti=ON	9m 32s
ubuntu-22.04 llvm assert=OFF rtti=OFF	9m 18s
ubuntu-22.04 gcc assert=ON rtti=ON	11m 29s
ubuntu-22.04 gcc assert=ON rtti=OFF	11m 1s
ubuntu-22.04 gcc assert=OFF rtti=ON	11m 5s
ubuntu-22.04 gcc assert=OFF rtti=OFF	10m 54s
ubuntu-24.04 llvm assert=ON rtti=ON	11m 32s
ubuntu-24.04 llvm assert=ON rtti=OFF	9m 24s
ubuntu-24.04 llvm assert=OFF rtti=ON	9m 37s
ubuntu-24.04 llvm assert=OFF rtti=OFF	9m 27s
ubuntu-24.04 gcc assert=ON rtti=ON	12m 9s
ubuntu-24.04 gcc assert=ON rtti=OFF	11m 42s
ubuntu-24.04 gcc assert=OFF rtti=ON	12m 13s
ubuntu-24.04 gcc assert=OFF rtti=OFF	11m 51s
ubuntu-20.04 gcc-11 assert=ON rtti=OFF	10m 39s
ubuntu-20.04 gcc-11 assert=OFF rtti=OFF	10m 30s
ubuntu-20.04 gcc-11 assert=ON rtti=ON	10m 7s
ubuntu-20.04 gcc-11 assert=OFF rtti=ON	11m 11s
windows-2019 msvc assert=ON rtti=ON	11m 4s
windows-2019 msvc assert=ON rtti=OFF	10m 46s
windows-2019 msvc assert=OFF rtti=ON	11m 51s
windows-2019 msvc assert=OFF rtti=OFF	10m 59s
	4h 17m 24s

Provide feedback