1 .. _libc_gpu_motivation:
3 ==========================
4 Motivation and Limitations
5 ==========================
7 .. contents:: Table of Contents
14 This project aims to provide a large subset of the C standard library to users
15 of GPU accelerators. We deliberately choose to only implement a subset of the C
16 library as some features are not expressly useful or easily implemented on the
17 GPU. This will be discussed further in `Limitations <libc_gpu_limitations>`_.
18 The main motivation behind this project is to provide the well understood C
19 library as a firm base for GPU development.
21 The main idea behind this project is that programming GPUs can be as
22 straightforward as programming on CPUs. This project aims to validate the GPU as
23 a more general-purpose target. The implementations here will also enable more
24 complex implementations of other libraries on the GPU, such as ``libc++``.
26 Host services and C library features are currently provided sparsely by the
27 different GPU vendors. We wish to provide these functions more completely and
28 make their implementations available in a common format. This is useful for
29 targets like OpenMP offloading or SYCL which wish to unify the offloading
30 toolchain. We also aim to provide these functions in a format compatible with
31 offloading in ``Clang`` so that we can treat the C library for the GPU as a
32 standard static library.
34 A valuable use for providing C library features on the GPU is for testing. For
35 this reason we build `tests on the GPU <libc_gpu_testing>`_ that can run a unit
36 test as if it were being run on the CPU. This also helps users port applications
37 that traditionally were run on the CPU. With this support, we can expand test
38 coverage for the GPU backend to the existing LLVM C library tests.
40 .. _libc_gpu_limitations:
45 We only implement a subset of the standard C library. The GPU does not
46 currently support thread local variables in all cases, so variables like
47 ``errno`` are atomic and global. Furthermore, the GPU under the OpenCL execution
48 model cannot safely provide a mutex interface. This means that features like
49 file buffering are not implemented on the GPU. We can also not easily provide
50 threading features on the GPU due to the execution model so these will be
51 ignored, as will features like ``locale`` or ``time``.