Skip to content

Fix build error caused by TMA macro guard #3988

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

Merged
zcbenz merged 1 commit into from
Aug 3, 2026
Merged

Fix build error caused by TMA macro guard #3988

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
12 changes: 5 additions & 7 deletions mlx/backend/cuda/ptx.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,12 @@
#pragma once

#include <cuda.h>
#include <cuda_runtime.h>
#include <cute/arch/config.hpp>

namespace mlx::core {

namespace ptx {

#if (CUDART_VERSION >= 12080) && (__CUDA_ARCH__ >= 1000) && \
defined(__CUDA_ARCH_SPECIFIC__)
#if defined(CUTE_ARCH_TMA_SM90_ENABLED)

__device__ __forceinline__ void mbarrier_init(uint64_t* mbar, uint32_t count) {
uint32_t mbar_ptr = __cvta_generic_to_shared(mbar);
Expand Down Expand Up @@ -121,7 +119,7 @@ __device__ __forceinline__ void fence_proxy_async_shared_cta() {
asm volatile("fence.proxy.async.shared::cta;");
}

#endif // (CUDART_VERSION >= 12080) && (__CUDA_ARCH__ >= 1000) &&
// (__CUDA_ARCH_FAMILY_SPECIFIC__ >= 1000)
#endif // defined(CUTE_ARCH_TMA_SM90_ENABLED)

} // namespace ptx
} // namespace mlx::core
} // namespace mlx::core
8 changes: 4 additions & 4 deletions mlx/backend/cuda/quantized/fp_quantize.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,7 +55,7 @@ __device__ __forceinline__ void copy_2d_to_shared(
uint32_t num_bytes,
uint64_t* barrier,
const bool is_master_thread) {
#if (defined __CUDA_ARCH__) && (__CUDA_ARCH__ >= 1000)
#if defined(CUTE_ARCH_TMA_SM90_ENABLED)
if (is_master_thread) {
// Arrive and tell how many bytes are expected
ptx::mbarrier_arrive_expect_tx(barrier, num_bytes);
Expand All @@ -66,7 +66,7 @@ __device__ __forceinline__ void copy_2d_to_shared(
// Other threads just arrive
ptx::mbarrier_arrive(barrier);
}
#endif // #if (defined __CUDA_ARCH__) && (__CUDA_ARCH__ >= 1000)
#endif // defined(CUTE_ARCH_TMA_SM90_ENABLED)
}

namespace cg = cooperative_groups;
Expand Down Expand Up @@ -228,7 +228,7 @@ __global__ void __launch_bounds__(THREADS_PER_BLOCK)
uint8_t* __restrict__ scales,
const size_t rows,
const size_t cols) {
#if (defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 1000)
#if defined(CUTE_ARCH_TMA_SM90_ENABLED)
using Tx2 = Vector2_t<T>;
using Tx4 = Vector4_t<T>;

Expand Down Expand Up @@ -408,7 +408,7 @@ __global__ void __launch_bounds__(THREADS_PER_BLOCK)
ptx::mbarrier_invalidate(&mbar[iter]);
}
}
#endif // __CUDA_ARCH__ >= 1000
#endif // defined(CUTE_ARCH_TMA_SM90_ENABLED)
}

// TODO: add kernel with tma instructions
Expand Down
Loading