[flang-commits] [flang] Revert "[flang][cuda] Add entry points for cudastreamsynchronize (#181932)" (PR #182657)

[via flang-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-flang-fir-hlfir

Author: Valentin Clement (バレンタイン クレメン) (clementval)

<details>
<summary>Changes</summary>



This is causing some testing issue. Reverting for now.

---
Full diff: https://github.com/llvm/llvm-project/pull/182657.diff


2 Files Affected:

- (modified) flang/lib/Optimizer/Builder/CUDAIntrinsicCall.cpp (-10) 
- (modified) flang/test/Lower/CUDA/cuda-default-stream.cuf (-11) 


``````````diff
diff --git a/flang/lib/Optimizer/Builder/CUDAIntrinsicCall.cpp b/flang/lib/Optimizer/Builder/CUDAIntrinsicCall.cpp
index e5b07aa3861a2..0403a43a845cd 100644
--- a/flang/lib/Optimizer/Builder/CUDAIntrinsicCall.cpp
+++ b/flang/lib/Optimizer/Builder/CUDAIntrinsicCall.cpp
@@ -403,16 +403,6 @@ static constexpr IntrinsicHandler cudaHandlers[]{
          &CI::genCUDASetDefaultStream),
      {{{"stream", asValue}}},
      /*isElemental=*/false},
-    {"cudastreamsynchronize",
-     static_cast<CUDAIntrinsicLibrary::ExtendedGenerator>(
-         &CI::genCUDAStreamSynchronize),
-     {{{"stream", asValue}}},
-     /*isElemental=*/false},
-    {"cudastreamsynchronizenull",
-     static_cast<CUDAIntrinsicLibrary::ElementalGenerator>(
-         &CI::genCUDAStreamSynchronizeNull),
-     {},
-     /*isElemental=*/false},
     {"fence_proxy_async",
      static_cast<CUDAIntrinsicLibrary::SubroutineGenerator>(
          &CI::genFenceProxyAsync),
diff --git a/flang/test/Lower/CUDA/cuda-default-stream.cuf b/flang/test/Lower/CUDA/cuda-default-stream.cuf
index 5bc78af22e84b..b9fe1f7949595 100644
--- a/flang/test/Lower/CUDA/cuda-default-stream.cuf
+++ b/flang/test/Lower/CUDA/cuda-default-stream.cuf
@@ -39,14 +39,3 @@ end subroutine
 ! CHECK: %{{.*}} = fir.call @_FortranACUFGetDefaultStream() fastmath<contract> : () -> i64
 ! CHECK: %{{.*}} = fir.call @_FortranACUFGetDefaultStream() fastmath<contract> : () -> i64
 
-subroutine stream_synchronize
-  use cuda_runtime_api
-  integer(kind=cuda_stream_kind) :: strm
-  integer :: istat
-  istat = cudastreamsynchronize(strm)
-  istat = cudastreamsynchronize()
-end subroutine
-
-! CHECK-LABEL: func.func @_QPstream_synchronize()
-! CHECK: %{{.*}} = fir.call @_FortranACUFStreamSynchronize(%{{.*}}) fastmath<contract> : (i64) -> i32
-! CHECK: %{{.*}} = fir.call @_FortranACUFStreamSynchronizeNull() fastmath<contract> : () -> i32

``````````

</details>


https://github.com/llvm/llvm-project/pull/182657