我不确定我理解这里发生的事情的细微差别,并希望得到解释。
我从模板化的包装器lapack_gesvd_nothrow
调用了几个重载函数。从那里,我打电话给这样的个人fxns:
inline void lapack_gesvd(char *jobu, char *jobvt,
int *m, int *n,
float *a, int *lda,
float *s,
float *u, int *ldu,
float *vt, int *ldvt,
float *work, int *lwork,
int *info) {
sgesvd_(jobu, jobvt, m, n,
a, lda, s, u, ldu,
vt, ldvt, work, lwork,
info);
}
inline void lapack_gesvd(char *jobu, char *jobvt,
int *m, int *n,
nm::Complex64 *a, int *lda,
nm::Complex64 *s,
nm::Complex64 *u, int *ldu,
nm::Complex64 *vt, int *ldvt,
nm::Complex64 *work, int *lwork, float *rwork,
int *info) {
cgesvd_(jobu, jobvt, m, n,
a, lda, s, u, ldu,
vt, ldvt, work, lwork,
rwork, info);
}
直到我宣布第二种类型的重载(我已经重载浮动和双重,没有问题),但是现在它抛出了一些错误,似乎不能很好地计算我的论点。
我从一个接受参数的函数中调用它:
template <typename DType, typename CType>
static int lapack_gesvd_nothrow(char *jobu, char *jobvt,
int m, int n,
void *a, int lda,
void *s,
void *u, int ldu,
void *vt, int ldvt,
void *work, int lwork,
int info, void *rwork) {
....
DType* UPCASE = reinterpret_cast<DType*>(lowercase);
....
if (typeid(DType) == typeid(CType)) {
lapack_gesvd(jobu, jobvt, &m, &n, A, &lda, S, U, &ldu, VT, &ldvt, WORK, &lwork, &info);
} else {
CType* RWORK = reinterpret_cast<CType*>(rwork);
lapack_gesvd(jobu, jobvt, &m, &n, A, &lda, S, U, &ldu, VT, &ldvt, WORK, &lwork, RWORK, &info);
}
我实际上只在两者之间做reinterpret_casts
。
这是它显然正在寻找的fxn:
error: no matching function for call to ‘lapack_gesvd(char*&, char*&, int*, int*, float*&, int*, float*&, float*&, int*, float*&, int*, float*&, int*, float*&, int*)
以下是候选人匹配:
candidates are:
note: void nm::math::lapack_gesvd(char*, char*, int*, int*, float*, int*, float*, float*, int*, float*, int*, float*, int*, int*)
note: candidate expects 14 arguments, 15 provided
void nm::math::lapack_gesvd(char*, char*, int*, int*, double*, int*, double*, double*, int*, double*, int*, double*, int*, int*)
note: candidate expects 14 arguments, 15 provided
void nm::math::lapack_gesvd(char*, char*, int*, int*, nm::Complex64*, int*, nm::Complex64*, nm::Complex64*, int*, nm::Complex64*, int*, nm::Complex64*, int*, float*, int*)
note: no known conversion for argument 5 from ‘float*’ to ‘nm::Complex64* {aka nm::Complex<float>*}’
我很困惑为什么解除引用现在出现,当它显示需要error: invalid conversion from 'int' to 'int*'
没有它,并且在最近的重载之前需要。
您的解释和解决方案将非常感谢!谢谢!
修改
可能归结为:
如果我拨打lapack_gesvd_nothrow<float, float>(...)
然后执行typeid比较if (typeid(DType) == typeid(CType))
...我会得到预期的答案吗?现在,它似乎不是这样。如何正确检查模板类型以进行比较?
答案 0 :(得分:1)
<强> TL; DR 强>
你这样做:
if (false) {
// some language rule violation here
} else {
// correct code here
}
if-else的两边都需要编译。
<强>解决方案强>
您可以部分专门化模板。
template <class U, class V>
void foo(...){
//assume U and V are different
}
template <class U>
void foo<U, U>(...){
//assume both types are the same
}
为什么会这样?
编译器替换编译时指定的类型。这种方式如果你初始化,它将改变if with float
lapack_gesvd_nothrow<float, float>(...)
if (typeid(float) == typeid(float))
这样,代码最终会像这样:
if (typeid(float) == typeid(float)){
lapack_gesvd(jobu, jobvt, &m, &n, A, &lda, S, U, &ldu, VT, &ldvt, WORK, &lwork, &info);
} else {
CType* RWORK = reinterpret_cast<CType*>(rwork);
lapack_gesvd(jobu, jobvt, &m, &n, A, &lda, S, U, &ldu, VT, &ldvt, WORK, &lwork, RWORK, &info);
}
编译器会看到A的类型是float。 if-else语句的第一部分是正确的。虽然,在第二部分不是。以下是编译器将如何看到它的示例:
void foo(float, float){};
void foo(int, int, int){};
template <class U>(){
...
U a, b, c;
if (...)
foo(a, b);
else
foo(a, b, c);
...
}
//Will be changed to
float a, b, c;
if (...)
foo(a, b);
else
foo(a, b, c);
导致else部分出现编译错误,因为没有foo(float,float,float)。