Add test for 0 byte copy and copy one element. Fix 0 byte copy implementation

Author: Icohedron

llvm/lib/Target/DirectX/DXILCBufferAccess.cpp

@@ -199,8 +199,7 @@ static void replaceLoad(LoadInst *LI, CBufferResource &CBR,
 /// itself. Assumes the cbuffer global is an array, and the length of bytes to
 /// copy is divisible by array element allocation size.
 /// The memcpy source must also be a direct cbuffer global reference, not a GEP.
-static void replaceMemCpy(MemCpyInst *MCI, CBufferResource &CBR,
-                          SmallVectorImpl<WeakTrackingVH> &DeadInsts) {
+static void replaceMemCpy(MemCpyInst *MCI, CBufferResource &CBR) {
 
   ArrayType *ArrTy = dyn_cast<ArrayType>(CBR.getValueType());
   assert(ArrTy && "MemCpy lowering is only supported for array types");
@@ -219,7 +218,7 @@ static void replaceMemCpy(MemCpyInst *MCI, CBufferResource &CBR,
 
   // If length to copy is zero, no memcpy is needed
   if (ByteLength == 0) {
-    DeadInsts.push_back(MCI);
+    MCI->eraseFromParent();
     return;
   }
 
@@ -290,7 +289,7 @@ static void replaceAccessesWithHandle(CBufferResource &CBR) {
     // If we have a memcpy instruction, replace it with multiple accesses and
     // subsequent stores to the destination
     if (auto *MCI = dyn_cast<MemCpyInst>(Cur)) {
-      replaceMemCpy(MCI, CBR, DeadInsts);
+      replaceMemCpy(MCI, CBR);
       continue;
     }
 

llvm/test/CodeGen/DirectX/CBufferAccess/memcpy.ll

@@ -37,6 +37,18 @@ entry:
   %a7.copy = alloca [2 x i64], align 8
   %a8.copy = alloca [4 x i32], align 4
 
+  ; Try copying no elements
+; CHECK-NOT: memcpy
+  call void @llvm.memcpy.p0.p2.i32(ptr align 4 %a1.copy, ptr addrspace(2) align 4 @a1, i32 0, i1 false)
+
+  ; Try copying only the first element
+; CHECK:    [[CB:%.*]] = load target("dx.CBuffer", {{.*}})), ptr @CB.cb, align 4
+; CHECK:    [[LOAD:%.*]] = call { float, float, float, float } @llvm.dx.resource.load.cbufferrow.4.{{.*}}(target("dx.CBuffer", {{.*}})) [[CB]], i32 0)
+; CHECK:    [[X:%.*]] = extractvalue { float, float, float, float } [[LOAD]], 0
+; CHECK:    [[DEST:%.*]] = getelementptr inbounds i8, ptr [[A1_COPY:%.*]], i32 0
+; CHECK:    store float [[X]], ptr [[DEST]], align 4
+  call void @llvm.memcpy.p0.p2.i32(ptr align 4 %a1.copy, ptr addrspace(2) align 4 @a1, i32 4, i1 false)
+
 ; CHECK:    [[CB:%.*]] = load target("dx.CBuffer", {{.*}})), ptr @CB.cb, align 4
 ; CHECK:    [[LOAD:%.*]] = call { float, float, float, float } @llvm.dx.resource.load.cbufferrow.4.{{.*}}(target("dx.CBuffer", {{.*}})) [[CB]], i32 0)
 ; CHECK:    [[X:%.*]] = extractvalue { float, float, float, float } [[LOAD]], 0