Add support for building CTS natively on Windows#2722
Conversation
| ### Building CTS on Windows | ||
|
|
||
| For Windows environments, it is strongly recommended to build CTS using [MSYS2](https://www.msys2.org/), | ||
| the MinGW-w64 (GCC) toolchain, and Ninja. | ||
| All commands in the following sections should be run from an MSYS2 MinGW64 shell. |
There was a problem hiding this comment.
Can you give use more information why that is?
There was a problem hiding this comment.
For example, using the MSVC compiler on Windows, in which long double is treated as identical to double, so it eliminates extended‑precision support and can introduce precision loss in tests that rely on higher‑precision floating‑point behavior, such as brute_math FP64 for asinh, atanh, , expm1, tgamma , etc.
There was a problem hiding this comment.
I'm having trouble thinking that this justifies the OpenCL group recommending MSY2 over anything else.
But I'm not a Windows user, so I will let that to people with more expertise.
There was a problem hiding this comment.
If we want to advertise support for MYS2, regardless of whether we recommend it over MSVC on Windows, I think we should introduce CI testing for that configuration. MSVC builds on windows are already tested in CI, but without CI testing we could easily break a MYS2 build again after this PR is merged.
rjodinchr
left a comment
rjodinchr
left a comment
There was a problem hiding this comment.
Fine as far as I can see. We need some people involved with windows to review
3 participants
Adding instructions and support to build CTS on Windows using MSYS2 gcc.