[SYCL] Change uint to uint32_t to avoid build error on Windows

[v-klochkov]

Signed-off-by: Vyacheslav N Klochkov vyacheslav.n.klochkov@intel.com

[vladimirlaz]

sycl/include/CL/sycl/types.hpp also contains uint:
sycl/include/CL/sycl/types.hpp:1881:DECLARE_SYCL_VEC(uint)

and sycl/test/basic_tests/types.cpp also have:
sycl/test/basic_tests/types.cpp:48: CHECK_TYPE(uint);
and sycl/test/sub_group/sg.cl:
sycl/test/sub_group/sg.cl:10: uint local_id;
sycl/test/sub_group/sg.cl:11: uint local_range;
sycl/test/sub_group/sg.cl:12: uint max_local_range;
sycl/test/sub_group/sg.cl:13: uint group_id;
sycl/test/sub_group/sg.cl:14: uint group_range;
sycl/test/sub_group/sg.cl:15: uint uniform_group_range;
sycl/test/sub_group/sg.cl:18: uint id = get_global_id(0);

and sycl/test/struct_param/struct_kernel_param.cpp:
sycl/test/struct_param/struct_kernel_param.cpp:87: uint ice2 = 888;
sycl/test/struct_param/struct_kernel_param.cpp:88: uint result[4] = {0};

I guess we will have failures for compilation on windows as well

[v-klochkov]

Vladimir, those other places where uint is used do not cause any errors on Windows.
For commands.h/cpp I had choice: to use cl::sycl::uint or use uint32_t. I chose the 2nd.

[vladimirlaz]

Vladimir, those other places where uint is used do not cause any errors on Windows.
For commands.h/cpp I had choice: to use cl::sycl::uint or use uint32_t. I chose the 2nd.

Could you explain why we have warning in one place and do not have in the other for the similar code?

[v-klochkov]

Errors, not warnings.
Initially I thought it was something wrong with namespaces, but after an additional quick investigation I realized that the alternative fix to my patch is to simply add add '#include <types.hpp>' to commands.h.

It is rather the matter of taste here (cl::sycl::uint vs uint32_t).
uint32_t is used much more often in SYCL sources/headers than 'uint'. So, I would stick to the current 'uint32_t' fix.

[vladimirlaz]

Errors, not warnings.
Initially I thought it was something wrong with namespaces, but after an additional quick investigation I realized that the alternative fix to my patch is to simply add add '#include <types.hpp>' to commands.h.

It is rather the matter of taste here (cl::sycl::uint vs uint32_t).
uint32_t is used much more often in SYCL sources/headers than 'uint'. So, I would stick to the current 'uint32_t' fix.

Thank you for clarification. LGTM.

[bader]

You fixed only forward declarations, but not definitions.
Are they use something different from uint?

[bader]

This change seems to be unrelated to the fix.
Suggest reverting.

[v-klochkov]

This change IS related to the fix.
The fix changed the 1st argument of passGlobalAccessorAsArg() and passLocalAccessorAsArg from uint to uint32_t and ArgumentID is passed to there.

BTW, your comment helped me to find couple places where I did not move the patch properly from WIndows (where it was complete) to Linux (where I also tested the patch and from where uploaded fix for review). Those 2 places are fixed now (changed uint to uint32_t there).

[bader]

Are you going to replace uint I?

[v-klochkov]

Yes, thank you. Sorry, I did 'git commit' without doing 'git add', that is why uint -> uint32_t for this parameter 'I' was not updated here 20 minutes ago. It is fixed now.