[DAGCombiner] Preserve debug location of original load in fold (conv (load x))

This patch fixes a debug information loss issue during the combine of a conversion (e.g., bitcast)
with a load into a new load: `fold (conv (load x)) -> (load (conv*)x)`.

The newly created load node was incorrectly using the debug location (`SDLoc`) of the conversion
operation (the `conv` node, `N`) instead of the location of the original load operation (the `load`
node, `LN0`). The location of the conversion operation often points to compiler-internal
instructions and provides little value for source-level debugging. In contrast, the original load's
location accurately represents the source of the data access in the user's code.

This change ensures the new load inherits the debug location from `LN0` by using `SDLoc(LN0)`,
which improves debugging experience and fixes a test case failure observed in the Triton compiler.

Author: jwu10003

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -16703,7 +16703,7 @@ SDValue DAGCombiner::visitBITCAST(SDNode *N) {
         }
       }
       SDValue Load =
-          DAG.getLoad(VT, SDLoc(N), LN0->getChain(), LN0->getBasePtr(),
+          DAG.getLoad(VT, SDLoc(LN0), LN0->getChain(), LN0->getBasePtr(),
                       LN0->getMemOperand());
       DAG.ReplaceAllUsesOfValueWith(N0.getValue(1), Load.getValue(1));
       return Load;

llvm/test/CodeGen/AMDGPU/combine-conv-load.ll

@@ -0,0 +1,41 @@
+; RUN: llc -mtriple=amdgcn -mcpu=gfx942  < %s | FileCheck %s
+
+; CHECK-LABEL:  test:
+; CHECK:        .loc    1 8 16                          ; test.py:8:16
+; CHECK-NEXT:   s_load_dword
+
+; Function Attrs: alwaysinline mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite)
+define amdgpu_kernel void @test(ptr addrspace(1) inreg readonly captures(none) %0, ptr addrspace(1) inreg writeonly captures(none) %1, ptr addrspace(1) inreg readnone captures(none) %2, ptr addrspace(1) inreg readnone captures(none) %3) local_unnamed_addr #0 !dbg !4 {
+  %5 = tail call i32 @llvm.amdgcn.workitem.id.x(), !dbg !7
+  %6 = and i32 %5, 255, !dbg !7
+  %7 = icmp eq i32 %6, 0, !dbg !7
+  br i1 %7, label %8, label %10, !dbg !7
+
+8:                                                ; preds = %4
+  %9 = load <1 x float>, ptr addrspace(1) %0, align 4, !dbg !8, !amdgpu.noclobber !6
+  store <1 x float> %9, ptr addrspace(1) %1, align 4, !dbg !7
+  br label %10, !dbg !7
+
+10:                                               ; preds = %8, %4
+  ret void, !dbg !9
+}
+
+; Function Attrs: alwaysinline nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare noundef range(i32 0, 1024) i32 @llvm.amdgcn.workitem.id.x() #1
+
+attributes #0 = { alwaysinline mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite) "amdgpu-agpr-alloc"="0" "amdgpu-flat-work-group-size"="1,256" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="1,1" "denormal-fp-math-f32"="ieee" "uniform-work-group-size"="false" }
+attributes #1 = { alwaysinline nocallback nofree nosync nounwind speculatable willreturn memory(none) }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!2, !3}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C, file: !1, producer: "triton", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly)
+!1 = !DIFile(filename: "test.py", directory: "/path")
+!2 = !{i32 2, !"Debug Info Version", i32 3}
+!3 = !{i32 1, !"amdhsa_code_object_version", i32 500}
+!4 = distinct !DISubprogram(name: "test", linkageName: "test", scope: !1, file: !1, line: 7, type: !5, scopeLine: 7, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0)
+!5 = !DISubroutineType(cc: DW_CC_normal, types: !6)
+!6 = !{}
+!7 = !DILocation(line: 9, column: 20, scope: !4)
+!8 = !DILocation(line: 8, column: 16, scope: !4)
+!9 = !DILocation(line: 9, column: 4, scope: !4)