Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[libc] Partially implement 'rand' for the GPU #66167

Merged
merged 1 commit into from
Oct 19, 2023
Merged

[libc] Partially implement 'rand' for the GPU #66167

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions libc/config/gpu/entrypoints.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,8 @@ set(TARGET_LIBC_ENTRYPOINTS
libc.src.stdlib.lldiv
libc.src.stdlib.qsort
libc.src.stdlib.qsort_r
libc.src.stdlib.rand
libc.src.stdlib.srand
libc.src.stdlib.strtod
libc.src.stdlib.strtof
libc.src.stdlib.strtol
Expand Down
10 changes: 6 additions & 4 deletions libc/src/stdlib/rand.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,10 +15,12 @@ namespace LIBC_NAMESPACE {
// An implementation of the xorshift64star pseudo random number generator. This
// is a good general purpose generator for most non-cryptographics applications.
LLVM_LIBC_FUNCTION(int, rand, (void)) {
rand_next ^= rand_next >> 12;
rand_next ^= rand_next << 25;
rand_next ^= rand_next >> 27;
return static_cast<int>((rand_next * 0x2545F4914F6CDD1Dul) >> 32) & RAND_MAX;
unsigned long x = rand_next;
x ^= x >> 12;
x ^= x << 25;
x ^= x >> 27;
rand_next = x;
return static_cast<int>((x * 0x2545F4914F6CDD1Dul) >> 32) & RAND_MAX;
}

} // namespace LIBC_NAMESPACE
6 changes: 6 additions & 0 deletions libc/src/stdlib/rand_util.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,14 @@

namespace LIBC_NAMESPACE {

#ifdef LIBC_TARGET_ARCH_IS_GPU
// FIXME: Local GPU memory cannot be initialized so we cannot currently provide
// a standard compliant default value.
ThreadLocal<unsigned long> rand_next;
#else
// C standard 7.10p2: If 'rand' is called before 'srand' it is to proceed as if
// the 'srand' function was called with a value of '1'.
LIBC_THREAD_LOCAL unsigned long rand_next = 1;
#endif

} // namespace LIBC_NAMESPACE
22 changes: 22 additions & 0 deletions libc/src/stdlib/rand_util.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,11 +9,33 @@
#ifndef LLVM_LIBC_SRC_STDLIB_RAND_UTIL_H
#define LLVM_LIBC_SRC_STDLIB_RAND_UTIL_H

#include "src/__support/GPU/utils.h"
#include "src/__support/macros/attributes.h"

namespace LIBC_NAMESPACE {

#ifdef LIBC_TARGET_ARCH_IS_GPU
// Implement thread local storage on the GPU using local memory. Each thread
// gets its slot in the local memory array and is private to the group.
// TODO: We need to implement the 'thread_local' keyword on the GPU. This is an
// inefficient and incomplete stand-in until that is done.
template <typename T> class ThreadLocal {
private:
static constexpr long MAX_THREADS = 1024;
[[clang::loader_uninitialized]] static inline gpu::Local<T>
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What is gpu::Local? If you use AS(5), you just need one T, not 1024.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

gpu::local is an alias to opencl_local which is AS(5) in the backend. Global AS(5) memory doesn't work unless it's fully inlined into each calling kernel, which isn't something we can guarantee, especially because the seed is global state.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I want to allocate thread local variables on the kernel stack, using basically the same compile time allocation scheme as LDS. I'm getting push back on that because applications run out of stack space already and using more of it is considered hazardous.

AS(5) globals should be raising an error in the back end. They definitely aren't correctly lowered.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why should that cause pushback? We should be able to optimize any of this away in the general case. If we use this to handle thread_local semantics then it's a pay-for-what-you-use type situation I'd think.

storage[MAX_THREADS];

public:
LIBC_INLINE operator T() const { return storage[gpu::get_thread_id()]; }
LIBC_INLINE void operator=(const T &value) {
storage[gpu::get_thread_id()] = value;
}
};

extern ThreadLocal<unsigned long> rand_next;
#else
extern LIBC_THREAD_LOCAL unsigned long rand_next;
#endif

} // namespace LIBC_NAMESPACE

Expand Down
3 changes: 3 additions & 0 deletions libc/test/src/stdlib/rand_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,12 +23,15 @@ TEST(LlvmLibcRandTest, UnsetSeed) {
vals[i] = val;
}

// FIXME: The GPU implementation cannot initialize the seed correctly.
#ifndef LIBC_TARGET_ARCH_IS_GPU
// The C standard specifies that if 'srand' is never called it should behave
// as if 'srand' was called with a value of 1. If we seed the value with 1 we
// should get the same sequence as the unseeded version.
LIBC_NAMESPACE::srand(1);
for (size_t i = 0; i < 1000; ++i)
ASSERT_EQ(LIBC_NAMESPACE::rand(), vals[i]);
#endif
}

TEST(LlvmLibcRandTest, SetSeed) {
Expand Down