Visible to Intel only — GUID: GUID-3D339375-3B88-4B38-8780-393F585FE13C
Visible to Intel only — GUID: GUID-3D339375-3B88-4B38-8780-393F585FE13C
DPCT1125
Message
The type “<type name>” defined in function “<function name>” is used as the parameter type in all functions in the call path from the corresponding sycl::handler::parallel_for() to the current function. You may need to adjust the definition location of the type.
Detailed Help
SYCL* 2020 does not support declaring a local memory variable in place. A local accessor needs to be declared in the command group scope and passed to all functions using the local memory. So, the definition location of the local memory variable type needs to be changed. Note: The DPC++ compiler extension sycl_ext_oneapi_local_memory (https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/supported/sycl_ext_oneapi_local_memory.asciidoc) supports declaring a local memory variable in place, but only in the kernel function scope.
Suggestions to Fix
For example, this original CUDA* code:
template <typename T> struct kernel_type_t {
using Type = T;
};
template <typename T> __global__ void device(int a, int b) {
using Tk = typename kernel_type_t<T>::Type;
__shared__ Tk mem[256];
...
}
template <typename T> __global__ void kernel(int a, int b) { device<T>(a, b); }
template <typename T> void foo() {
int i;
kernel<T><<<1, 1>>>(i, i);
}
results in the following migrated SYCL code:
template <typename T> struct kernel_type_t {
using Type = T;
};
/*
DPCT1125:1: The type "Tk" defined in function "device" is used as the parameter
type in all functions in the call path from the corresponding
sycl::handler::parallel_for() to the current function. You may need to adjust
the definition location of the type.
*/
template <typename T> void device(int a, int b, Tk *mem) {
using Tk = typename kernel_type_t<T>::Type;
...
}
/*
DPCT1125:2: The type "Tk" defined in function "device" is used as the parameter
type in all functions in the call path from the corresponding
sycl::handler::parallel_for() to the current function. You may need to adjust
the definition location of the type.
*/
template <typename T>
void kernel(int a, int b, Tk *mem) {
device<T>(a, b, mem);
}
template <typename T> void foo() {
int i;
/*
DPCT1126:0: The type "Tk" defined in function "device" is used as the
parameter type in all functions in the call path from the
sycl::handler::parallel_for() to the function "device". You may need to adjust
the definition location of the type.
*/
dpct::get_in_order_queue().submit([&](sycl::handler &cgh) {
sycl::local_accessor<Tk, 1> mem_acc_ct1(sycl::range<1>(256), cgh);
cgh.parallel_for(
sycl::nd_range<3>(sycl::range<3>(1, 1, 1), sycl::range<3>(1, 1, 1)),
[=](sycl::nd_item<3> item_ct1) {
kernel<T>(
i, i,
mem_acc_ct1.template get_multi_ptr<sycl::access::decorated::no>()
.get());
});
});
}
which needs to be rewritten to:
template <typename T> struct kernel_type_t {
using Type = T;
};
template <typename T> void device(int a, int b, typename kernel_type_t<T>::Type *mem) {
...
}
template <typename T>
void kernel(int a, int b, typename kernel_type_t<T>::Type *mem) {
device<T>(a, b, mem);
}
template <typename T> void foo() {
int i;
using Tk = typename kernel_type_t<T>::Type;
dpct::get_in_order_queue().submit([&](sycl::handler &cgh) {
sycl::local_accessor<Tk, 1> mem_acc_ct1(sycl::range<1>(256), cgh);
cgh.parallel_for(
sycl::nd_range<3>(sycl::range<3>(1, 1, 1), sycl::range<3>(1, 1, 1)),
[=](sycl::nd_item<3> item_ct1) {
kernel<T>(
i, i,
mem_acc_ct1.template get_multi_ptr<sycl::access::decorated::no>().get());
});
});
}