Previously we used a single port to implement the RPC. This was
sufficient for single threaded tests but can potentially cause deadlocks
when using multiple threads. The reason for this is that GPUs make no
forward progress guarantees. Therefore one group of threads waiting on
another group of threads can spin forever because there is no guarantee
that the other threads will continue executing. The typical workaround
for this is to allocate enough memory that a sufficiently large number
of work groups can make progress. As long as this number is somewhat
close to the amount of total concurrency we can obtain reliable
execution around a shared resource.
This patch enables using multiple ports by widening the arrays to a
predetermined size and indexes into them. Empty ports are currently
obtained via a trivial linker scan. This should be imporoved in the
future for performance reasons. Portions of D148191 were applied to
achieve parallel support.
Depends on D149581
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D149598
The GPU has a different execution model to standard `_start`
implementations. On the GPU, all threads are active at the start of a
kernel. In order to correctly intitialize and call the constructors we
want single threaded semantics. Previously, this was done using a
makeshift global barrier with atomics. However, it should be easier to
simply put the portions of the code that must be single threaded in
separate kernels and then call those with only one thread. Generally,
mixing global state between kernel launches makes optimizations more
difficult, similarly to calling a function outside of the TU, but for
testing it is better to be correct.
Depends on D149527 D148943
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D149581
The execution model of the GPU expects that groups of threads will
execute in lock-step in SIMD fashion. It's both important for
performance and correctness that we treat this as the smallest possible
granularity for an RPC operation. Thus, we map multiple threads to a
single larger buffer and ship that across the wire.
This patch makes the necessary changes to support executing the RPC on
the GPU with multiple threads. This requires some workarounds to mimic
the model when handling the protocol from the CPU. I'm not completely
happy with some of the workarounds required, but I think it should work.
Uses some of the implementation details from D148191.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D148943
This patch adds the necessary hacks to support global constructors and
destructors. This is an incredibly hacky process caused by the primary
fact that Nvidia does not provide any binary tools and very little
linker support. We first had to emit references to these functions and
their priority in D149451. Then we dig them out of the module once it's
loaded to manually create the list that the linker should have made for
us. This patch also contains a few Nvidia specific hacks, but it passes
the test, albeit with a stack size warning from `ptxas` for the
callback. But this should be fine given the resource usage of a common
test.
This also adds a dependency on LLVM to the NVPTX loader, which hopefully doesn't
cause problems with our CUDA buildbot.
Depends on D149451
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D149527
The implementation of the test printing currently expects a null
terminated C-string. However, the `write_to_stderr` interface uses a
string view, which doesn't need to be null terminated. This patch
changes the printing interface to directly use `fwrite` instead rather
than relying on a null terminator.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D149493
Currently, the RPC interface with the loader is only tested if the other
tests fail. This test adds a direct test that runs a simple integer
increment over the RPC handshake 10000 times.
Depends on https://reviews.llvm.org/D148288
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D148342
This patch reworks the RPC interface to allow more generic memory
operations using the shared better. This patch decomposes the entire RPC
interface into opening a port and calling `send` or `recv` on it.
The `send` function sends a single packet of the length of the buffer.
The `recv` function is paired with the `send` call to then use the data.
So, any aribtrary combination of sending packets is possible. The only
restriction is that the client initiates the exchange with a `send`
while the server consumes it with a `recv`.
The operation of this is driven by two independent state machines that
tracks the buffer ownership during loads / stores. We keep track of two
so that we can transition between a send state and a recv state without
an extra wait. State transitions are observed via bit toggling, e.g.
This interface supports an efficient `send -> ack -> send -> ack -> send`
interface and allows for the last send to be ignored without checking
the ack.
A following patch will add some more comprehensive testing to this interface. I
I informally made an RPC call that simply incremented an integer and it took
roughly 10 microsends to complete an RPC call.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D148288
We will want to test the GPU `libc` with multiple threads in the future.
This patch adds the `--threads` and `--blocks` option to set the `x`
dimension of the kernel. Using CUDA terminology instead of OpenCL for
familiarity.
Depends on D148288 D148342
Reviewed By: jdoerfert, sivachandra, tra
Differential Revision: https://reviews.llvm.org/D148485
It is currently used only by math tests so moving it to test/src/math
keeps it closer to where it is used.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D148497
Along the way, the utility Timer has been moved to the math differential
test directory.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D148450
Summary:
This part was ignored and we just hoped that shutting down the runtime
freed these correctly. But it's best to be specific and free the memory
we've allocated.
Summary:
This implementation was buggy and inefficient. Fine-grained memory can
only be allocated on a page-level granularity. Which means that each
allocated string used about 4096 bytes. This is wasteful in general, and
also allowed for buggy behaviour. The previous copying of the
environment vector only worked because the large buffer size meant that
we would typically have a null byte after the allocated memory. However
this would break if the vector was larger than a page. This patch
allocates everything into a single buffer. It makes it easier to free,
use, and it more correct.
This patch adds the necessary build infrastructure to build and run the
integration tests on NVIDIA GPUs. The NVIDIA `nvlink` linker utility is
what is ultimately used to combine these files into a single executable
image. Unfortunately, their tool does not support static libraries. So
we need to link with every object directly instead. This could be solved
by impelementing a "wrapper" utility around `nvlink` like we used to use
for OpenMP. But for now this should be sufficient.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D146861
This patch adds the necessary code to impelement the existing RPC client
/ server interface when targeting NVPTX GPUs. This follows closely to
the implementation in the AMDGPU version. This does not yet enable unit
testing as the `nvlink` linker does not support static libraries. So
that will need to be worked around.
I am ignoring the RPC duplication between the AMDGPU and NVPTX loaders. This
will be changed completely later so there's no point unifying the code at this
stage. The implementation was tested manually with the following file and
compilation flags.
```
namespace __llvm_libc {
void write_to_stderr(const char *msg);
void quick_exit(int);
} // namespace __llvm_libc
using namespace __llvm_libc;
int main(int argc, char **argv, char **envp) {
for (int i = 0; i < argc; ++i) {
write_to_stderr(argv[i]);
write_to_stderr("\n");
}
quick_exit(255);
}
```
```
$ clang++ crt1.o rpc_client.o quick_exit.o io.o main.cpp --target=nvptx64-nvidia-cuda -march=sm_70 -o image
$ ./nvptx_loader image 1 2 3
image
1
2
3
$ echo $?
255
```
Depends on D146681
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D146846
This patch adds a loader utility targeting the CUDA driver API to launch
NVPTX images called `nvptx_loader`. This takes a GPU image on the
command line and launches the `_start` kernel with the appropriate
arguments. The `_start` kernel is provided by the already implemented
`nvptx/start.cpp`. So, an application with a `main` function can be
compiled and run as follows.
```
clang++ --target=nvptx64-nvidia-cuda main.cpp crt1.o -march=sm_70 -o image
./nvptx_loader image args to kernel
```
This implementation is not tested and does not yet support RPC. This
requires further development to work around NVIDIA specific limitations
in atomics and linking.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D146681
This patch performs the same operation to copy over the `argv` array to
the `envp` array. This allows the GPU tests to use environment
variables.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D146322
This patch enables integration tests running on the GPU. This uses the
RPC interface implemented in D145913 to compile the necessary
dependencies for the integration test object. We can then use this to
compile the objects for the GPU directly and execute them using the AMD
HSA loader combined with its RPC server. For example, the compiler is
performing the following actions to execute the integration tests.
```
$ clang++ --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nostdlib -flto -ffreestanding \
crt1.o io.o quick_exit.o test.o rpc_client.o args_test.o -o image
$ ./amdhsa_loader image 1 2 5
args_test.cpp:24: Expected 'my_streq(argv[3], "3")' to be true, but is false
```
This currently only works with a single threaded client implementation
running on AMDGPU. Further work will implement multiple clients for AMD
and the ability to run on NVPTX as well.
Depends on D145913
Reviewed By: sivachandra, JonChesterfield
Differential Revision: https://reviews.llvm.org/D146256
This patch adds initial support for an RPC client / server architecture.
The GPU is unable to perform several system utilities on its own, so in
order to implement features like printing or memory allocation we need
to be able to communicate with the executing process. This is done via a
buffer of "sharable" memory. That is, a buffer with a unified pointer
that both the client and server can use to communicate.
The implementation here is based off of Jon Chesterfields minimal RPC
example in his work. We use an `inbox` and `outbox` to communicate
between if there is an RPC request and to signify when work is done.
We use a fixed-size buffer for the communication channel. This is fixed
size so that we can ensure that there is enough space for all
compute-units on the GPU to issue work to any of the ports. Right now
the implementation is single threaded so there is only a single buffer
that is not shared.
This implementation still has several features missing to be complete.
Such as multi-threaded support and asynchrnonous calls.
Depends on D145912
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D145913
Clean up some warnings from running libc-lint for these folders.
Reviewed By: michaelrj, sivachandra
Differential Revision: https://reviews.llvm.org/D146048
This library is linked into libc-hdrgen, which is compiled with
DISABLE_LLVM_LINK_LLVM_DYLIB. This option avoids linking with
libLLVM.so when LLVM_LINK_LLVM_DYLIB is specified. Unfortunately,
this can cause wierd linkage errors where symbols are defined
multiple times, with one definition coming from static linkage
and another definition coming from libLLVM.so. This is usually
apparent as a link error with options defined multiple times.
This patch adds DISABLE_LLVM_LINK_LLVM_DYLIB to this library, to
get consistent linkage of libc-hdrgen.
Differential Revision: https://reviews.llvm.org/D145839
The entrypoint has been added to the various entrypoint lists. The libc
code style doc has been updated with information on how errno should be
set from the libc runtime code.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D145179
This is the first attempt to get some testing support for GPUs in LLVM's
libc. We want to be able to compile for and call generic code while on
the device. This is difficult as most GPU applications also require the
support of large runtimes that may contain their own bugs (e.g. CUDA /
HIP / OpenMP / OpenCL / SYCL). The proposed solution is to provide a
"loader" utility that allows us to execute a "main" function on the GPU.
This patch implements a simple loader utility targeting the AMDHSA
runtime called `amdhsa_loader` that takes a GPU program as its first
argument. It will then attempt to load a predetermined `_start` kernel
inside that image and launch execution. The `_start` symbol is provided
by a `start` utility function that will be linked alongside the
application. Thus, this should allow us to run arbitrary code on the
user's GPU with the following steps for testing.
```
clang++ Start.cpp --target=amdgcn-amd-amdhsa -mcpu=<arch> -ffreestanding -nogpulib -nostdinc -nostdlib -c
clang++ Main.cpp --target=amdgcn-amd-amdhsa -mcpu=<arch> -nogpulib -nostdinc -nostdlib -c
clang++ Start.o Main.o --target=amdgcn-amd-amdhsa -o image
amdhsa_loader image <args, ...>
```
We determine the `-mcpu` value using the `amdgpu-arch` utility provided
either by `clang` or `rocm`. If `amdgpu-arch` isn't found or returns an
error we shouldn't run the tests as the machine does not have a valid
HSA compatible GPU. Alternatively we could make this utility in-source
to avoid the external dependency.
This patch provides a single test for this untility that simply checks
to see if we can compile an application containing a simple `main`
function and execute it.
The proposed solution in the future is to create an alternate
implementation of the LibcTest.cpp source that can be compiled and
launched using this utility. This approach should allow us to use the
same test sources as the other applications.
This is primarily a prototype, suggestions for how to better integrate
this with the existing LibC infastructure would be greatly appreciated.
The loader code should also be cleaned up somewhat. An implementation
for NVPTX will need to be written as well.
Reviewed By: sivachandra, JonChesterfield
Differential Revision: https://reviews.llvm.org/D139839
This part of the effort to make all test related pieces into the `test`
directory. This helps is excluding test related pieces in a straight
forward manner if LLVM_INCLUDE_TESTS is OFF. Future patches will also move
the MPFR wrapper and testutils into the 'test' directory.
WIFEXITED and friends expect an `int *` but these methods
were marked `const` so they instead got a `const int *`.
This macros aren't actually modifying their argument, but
we were never using these functions on an immutable
`ProcessStatus` type anyway.
To improve code generation for C++ code that directly includes our
headers, the external function definitions will now be marked noexcept.
This may not be necessary for the internal definitions since we build
with the -fno-exceptions flag.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D141095
The config currently includes ctype, math, stdlib, inttypes and string
functions.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D140378
Implement a high-precision floating point class using UInt<> as its
mantissa. This will be used in accurate pass for double precision math
functions.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D136799
Implement full multiplication `UInt<A> * UInt<B> -> UInt<A + B>` and
`quick_mul_hi` that returns the higher half of the product `UInt<A> * UInt<A>`.
These 2 functions will be used for dyadic floating point class.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D138541
This patch contains the initial support for building LLVM's libc as a
target for the GPU. Currently this only supports a handful of very basic
functions that can be implemented without an operating system. The GPU
code is build using the existing OpenMP toolchain. This allows us to
minimally change the existing codebase and get a functioning static
library. This patch allows users to create a static library called
`libcgpu.a` that contains fat binaries containing device IR.
Current limitations are the lack of test support and the fact that only
one target OS can be built at a time. That is, the user cannot get a
`libc` for Linux and one for the GPU simultaneously.
This introduces two new CMake variables to control the behavior
`LLVM_LIBC_TARET_OS` is exported so the user can now specify it to equal
`"gpu"`. `LLVM_LIBC_GPU_ARCHITECTURES` is also used to configure how
many targets to build for at once.
Depends on D138607
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D138608
When using `LLVM_ENABLE_RUNTIMES=libc` we need to perform a few extra
steps to include LLVM utilities similar to if we were performing a
standalone build. Libc depends on the tablegen utilities and the LLVM
libraries when performing a full build. When using an
`LLVM_ENABLE_PROJECTS=libc` build these are included as a part of the
greater LLVM build, but here we need to perform it maunally. This patch
should allow using `LLVM_LIBC_FULL_BUILD=ON` when building with
runtimes.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D138040
This is the first piece of scanf. It's very similar in design to printf,
and so much of the code is copied from that. There were potential issues
with conflicting macros so I've also renamed the "ASSERT_FORMAT_EQ"
macro for printf to "ASSERT_PFORMAT_EQ".
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D136288
