{
        // S4=A12-S2, P3=S4*B22, P1=A11*B11
        auto f = [a12, X, eSub, aHeight, numberThread, core, this](int tId) {
            auto xAddr = mStack[X.stackIndex] + X.offsetBytes;
            MNNMATRIX_SUB_MULTITHREAD(xAddr, mStack[a12.stackIndex] + a12.offsetBytes, xAddr, eSub, X.lineStrideBytes, a12.lineStrideBytes, X.lineStrideBytes, aHeight, core);
        };
        mFunctions.emplace_back(std::make_pair(f, numberThread));
        auto code = _generateMatMul(eSub, lSub, hSub, X, b22, c11, Empty, currentDepth, {});
        if (code != NO_ERROR) {
            return code;
        }
        code = _generateMatMul(eSub, lSub, hSub, a11, b11, CX, Empty, currentDepth, {});
        if (code != NO_ERROR) {
            return code;
        }
    }