From 2696e8c1499682f0b1f357d9035ed59f544892f8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Wed, 6 Aug 2025 18:49:52 -0700
Subject: [PATCH] [flang][cuda] Remove too restrictive assert for data transfer
 (#152398)

When the rhs is a an array element, the assert was triggered but this is
still a valid transfer. Remove the assert. The operation has a verifier
to check its validity.
---
 flang/lib/Lower/Bridge.cpp                   |  1 -
 flang/test/Lower/CUDA/cuda-data-transfer.cuf | 12 ++++++++++++
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 6b7efe6b57db..a5f0beff9ec2 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -4898,7 +4898,6 @@ private:
 
     // device = device
     if (lhsIsDevice && rhsIsDevice) {
-      assert(rhs.isVariable() && "CUDA Fortran assignment rhs is not legal");
       auto transferKindAttr = cuf::DataTransferKindAttr::get(
           builder.getContext(), cuf::DataTransferKind::DeviceDevice);
       cuf::DataTransferOp::create(builder, loc, rhsVal, lhsVal, shape,
diff --git a/flang/test/Lower/CUDA/cuda-data-transfer.cuf b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
index 3a4aff977b7a..9abf19b77869 100644
--- a/flang/test/Lower/CUDA/cuda-data-transfer.cuf
+++ b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
@@ -444,3 +444,15 @@ subroutine sub23(n)
 end subroutine
 
 ! CHECK-LABEL: func.func @_QPsub23
+
+subroutine sub24()
+  real, managed :: m
+  real, device :: d(4)
+  m = d(1)
+end
+
+! CHECK-LABEL: func.func @_QPsub24()
+! CHECK: %[[D:.*]]:2 = hlfir.declare %1(%2) {data_attr = #cuf.cuda<device>, uniq_name = "_QFsub24Ed"} : (!fir.ref<!fir.array<4xf32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<4xf32>>, !fir.ref<!fir.array<4xf32>>)
+! CHECK: %[[M:.*]]:2 = hlfir.declare %4 {data_attr = #cuf.cuda<managed>, uniq_name = "_QFsub24Em"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %[[D1:.*]] = hlfir.designate %[[D]]#0 (%c1{{.*}})  : (!fir.ref<!fir.array<4xf32>>, index) -> !fir.ref<f32>
+! CHECK: cuf.data_transfer %[[D1]] to %[[M]]#0 {transfer_kind = #cuf.cuda_transfer<device_device>} : !fir.ref<f32>, !fir.ref<f32>