[flang-commits] [flang] [llvm] Fixing the location attribute added to mapInfoOp (PR #90764)

[Anchu Rajendran S via flang-commits]

https://github.com/anchuraj created https://github.com/llvm/llvm-project/pull/90764

Named location attribute added to `tgt_offload_entry` shall be used by runtime calls like `ompx_dump_mapping_tables` to print the information of variables that are mapped to the device. `ompx_dump_mapping_tables` was printing the wrong location information and this change fixes it.

A sample execution of example before the change:
```
omptarget device 0 info: OpenMP Host-Device pointer mappings after block at libomptarget:0:0:

omptarget device 0 info: Host Ptr           Target Ptr         Size (B) DynRefCount HoldRefCount Declaration

omptarget device 0 info: 0x0000000000206df0 0x00007f02cdc00000 20000000 1           0            <program-file-loc> at unknown:18:35
```

The change replaces unknown to the mapped symbol and location to the declaration location.

>From b15058ec91b0224c3b0fae79fb5b14a838e20ae4 Mon Sep 17 00:00:00 2001
From: Anchu Rajendran <asudhaku at amd.com>
Date: Wed, 1 May 2024 13:22:23 -0500
Subject: [PATCH] Fixing the location attribute added to mapInfoOp

Named location attribute added tgt_info shall be used by runtime calls
like ompx_dump_mapping_tables to print the information of variables that
are mapped to the device. Before this change, ompx_dump_mapping_tables
was printing the wrong location information.
---
 flang/lib/Lower/OpenMP/ClauseProcessor.cpp    |  4 +-
 flang/lib/Lower/OpenMP/OpenMP.cpp             |  6 ++-
 .../offloading/fortran/dump_map_tables.f90    | 38 +++++++++++++++++++
 3 files changed, 45 insertions(+), 3 deletions(-)
 create mode 100644 offload/test/offloading/fortran/dump_map_tables.f90

diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index 4c51b61f6bf029..1af954f3927b57 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -903,8 +903,10 @@ bool ClauseProcessor::processMap(
           // Explicit map captures are captured ByRef by default,
           // optimisation passes may alter this to ByCopy or other capture
           // types to optimise
+          auto context = converter.getFirOpBuilder().getContext();
+          auto location = mlir::NameLoc::get(mlir::StringAttr::get(context, asFortran.str()), symAddr.getLoc());
           mlir::Value mapOp = createMapInfoOp(
-              firOpBuilder, clauseLocation, symAddr, mlir::Value{},
+              firOpBuilder, location, symAddr, mlir::Value{},
               asFortran.str(), bounds, {},
               static_cast<
                   std::underlying_type_t<llvm::omp::OpenMPOffloadMappingFlags>>(
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index e932f7c284bca8..00d9eca19d3cc5 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -1645,9 +1645,11 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
           mapFlag |= llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_TO;
           mapFlag |= llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_FROM;
         }
-
+    ;
+        auto context = converter.getFirOpBuilder().getContext();
+        auto location = mlir::NameLoc::get(mlir::StringAttr::get(context, sym.name().ToString()), baseOp.getLoc());
         mlir::Value mapOp = createMapInfoOp(
-            firOpBuilder, baseOp.getLoc(), baseOp, mlir::Value{}, name.str(),
+            firOpBuilder, location, baseOp, mlir::Value{}, name.str(),
             bounds, {},
             static_cast<
                 std::underlying_type_t<llvm::omp::OpenMPOffloadMappingFlags>>(
diff --git a/offload/test/offloading/fortran/dump_map_tables.f90 b/offload/test/offloading/fortran/dump_map_tables.f90
new file mode 100644
index 00000000000000..93de49205b32ec
--- /dev/null
+++ b/offload/test/offloading/fortran/dump_map_tables.f90
@@ -0,0 +1,38 @@
+! Offloading test with a target region mapping a declare target
+! Fortran array writing some values to it and checking the host
+! correctly receives the updates made on the device.
+! REQUIRES: flang
+! UNSUPPORTED: nvptx64-nvidia-cuda-LTO
+! UNSUPPORTED: aarch64-unknown-linux-gnu
+! UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
+! UNSUPPORTED: x86_64-pc-linux-gnu
+! UNSUPPORTED: x86_64-pc-linux-gnu-LTO
+
+! RUN: %libomptarget-compile-fortran-run-and-check-generic
+
+program map_dump_example
+  INTERFACE
+    SUBROUTINE ompx_dump_mapping_tables() BIND(C)
+    END SUBROUTINE ompx_dump_mapping_tables
+  END INTERFACE
+
+  integer i,j,k,N
+  integer async_q(4)
+  real :: A(5000000),B(5000000),C(5000000)
+  N=5000000
+  do i=1, N
+    A(i)=0
+  enddo
+! clang-format off
+! CHECK: omptarget device 0 info: OpenMP Host-Device pointer mappings after block
+! CHECK-NEXT: omptarget device 0 info: Host Ptr Target Ptr Size (B) DynRefCount HoldRefCount Declaration
+! CHECK-NEXT: omptarget device 0 info: {{(0x[0-9a-f]{16})}} {{(0x[0-9a-f]{16})}}  20000000 1 0 {{.*}} at a(:n):21:11
+! clang-format on
+!$omp target enter data map(to:A(:N))
+  call ompx_dump_mapping_tables()
+!$omp target parallel do
+  do i=1, N
+    A(i)=B(i)*C(i)
+  enddo
+!$omp target exit data map(from:A)
+end program