diff --git a/llvm/lib/Transforms/Utils/InlineFunction.cpp b/llvm/lib/Transforms/Utils/InlineFunction.cpp
--- a/llvm/lib/Transforms/Utils/InlineFunction.cpp
+++ b/llvm/lib/Transforms/Utils/InlineFunction.cpp
@@ -2194,9 +2194,11 @@
         CI->setTailCallKind(ChildTCK);
         InlinedMustTailCalls |= CI->isMustTailCall();
 
-        // Calls inlined through a 'nounwind' call site should be marked
-        // 'nounwind'.
-        if (MarkNoUnwind)
+        // Call sites inlined through a 'nounwind' call site should be
+        // 'nounwind' as well. However, avoid marking call sites explicitly
+        // where possible. This helps expose more opportunities for CSE after
+        // inlining, commonly when the callee is an intrinsic.
+        if (MarkNoUnwind && !CI->doesNotThrow())
           CI->setDoesNotThrow();
       }
     }
diff --git a/llvm/test/Transforms/Coroutines/coro-retcon-once-value.ll b/llvm/test/Transforms/Coroutines/coro-retcon-once-value.ll
--- a/llvm/test/Transforms/Coroutines/coro-retcon-once-value.ll
+++ b/llvm/test/Transforms/Coroutines/coro-retcon-once-value.ll
@@ -53,7 +53,7 @@
 ; CHECK-NEXT:    store i32* [[ARRAY:%.*]], i32** [[TMP0]], align 8
 ; CHECK-NEXT:    [[LOAD_I:%.*]] = load i32, i32* [[ARRAY]], align 4
 ; CHECK-NEXT:    [[LOAD_POS_I:%.*]] = icmp sgt i32 [[LOAD_I]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[LOAD_I]], i32 0) #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[LOAD_I]], i32 0)
 ; CHECK-NEXT:    call void @print(i32 [[TMP1]])
 ; CHECK-NEXT:    [[CONT_CAST:%.*]] = select i1 [[LOAD_POS_I]], void (i8*, i1)* @f.resume.0, void (i8*, i1)* @f.resume.1
 ; CHECK-NEXT:    call void [[CONT_CAST]](i8* nonnull [[DOTSUB]], i1 zeroext false)
diff --git a/llvm/test/Transforms/Inline/noalias-calls-always.ll b/llvm/test/Transforms/Inline/noalias-calls-always.ll
--- a/llvm/test/Transforms/Inline/noalias-calls-always.ll
+++ b/llvm/test/Transforms/Inline/noalias-calls-always.ll
@@ -34,11 +34,11 @@
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META0:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0i8(i64 512, i8* [[L_I]])
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A:%.*]], i8* align 16 [[B:%.*]], i64 16, i1 false) #[[ATTR6:[0-9]+]], !noalias !3
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C:%.*]], i64 16, i1 false) #[[ATTR6]], !noalias !0
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR6]], !alias.scope !5
-; CHECK-NEXT:    call void @hey() #[[ATTR6]], !noalias !5
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR6]], !noalias !0
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A:%.*]], i8* align 16 [[B:%.*]], i64 16, i1 false), !noalias !3
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C:%.*]], i64 16, i1 false), !noalias !0
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false), !alias.scope !5
+; CHECK-NEXT:    call void @hey(), !noalias !5
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false), !noalias !0
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0i8(i64 512, i8* [[L_I]])
 ; CHECK-NEXT:    ret void
 ;
@@ -75,11 +75,11 @@
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0i8(i64 512, i8* [[L_I]])
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A:%.*]], i8* align 16 [[B:%.*]], i64 16, i1 false) #[[ATTR6]], !noalias !9
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C:%.*]], i64 16, i1 false) #[[ATTR6]], !noalias !6
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR6]], !alias.scope !11
-; CHECK-NEXT:    call void @hey() #[[ATTR6]], !noalias !11
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR6]], !noalias !6
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A:%.*]], i8* align 16 [[B:%.*]], i64 16, i1 false), !noalias !9
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C:%.*]], i64 16, i1 false), !noalias !6
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false), !alias.scope !11
+; CHECK-NEXT:    call void @hey(), !noalias !11
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false), !noalias !6
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0i8(i64 512, i8* [[L_I]])
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/Inline/noalias-calls.ll b/llvm/test/Transforms/Inline/noalias-calls.ll
--- a/llvm/test/Transforms/Inline/noalias-calls.ll
+++ b/llvm/test/Transforms/Inline/noalias-calls.ll
@@ -37,11 +37,11 @@
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META0:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0i8(i64 512, i8* [[L_I]])
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[B]], i64 16, i1 false) #[[ATTR3]], !noalias !3
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !noalias !0
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !alias.scope !5
-; CHECK-NEXT:    call void @hey() #[[ATTR3]], !noalias !5
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !noalias !0
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[B]], i64 16, i1 false), !noalias !3
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C]], i64 16, i1 false), !noalias !0
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false), !alias.scope !5
+; CHECK-NEXT:    call void @hey(), !noalias !5
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false), !noalias !0
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0i8(i64 512, i8* [[L_I]])
 ; CHECK-NEXT:    ret void
 ;
@@ -80,11 +80,11 @@
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0i8(i64 512, i8* [[L_I]])
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[B]], i64 16, i1 false) #[[ATTR3]], !noalias !9
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !noalias !6
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !alias.scope !11
-; CHECK-NEXT:    call void @hey() #[[ATTR3]], !noalias !11
-; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false) #[[ATTR3]], !noalias !6
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[B]], i64 16, i1 false), !noalias !9
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[B]], i8* align 16 [[C]], i64 16, i1 false), !noalias !6
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[A]], i8* align 16 [[C]], i64 16, i1 false), !alias.scope !11
+; CHECK-NEXT:    call void @hey(), !noalias !11
+; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[L_I]], i8* align 16 [[C]], i64 16, i1 false), !noalias !6
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0i8(i64 512, i8* [[L_I]])
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/Inline/noalias-cs.ll b/llvm/test/Transforms/Inline/noalias-cs.ll
--- a/llvm/test/Transforms/Inline/noalias-cs.ll
+++ b/llvm/test/Transforms/Inline/noalias-cs.ll
@@ -65,8 +65,8 @@
 ; CHECK-LABEL: @caller(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[C:%.*]] = load float*, float** [[C_PTR:%.*]], align 8, !alias.scope !6
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]]) #[[ATTR2:[0-9]+]], !noalias !6
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]]) #[[ATTR2]], !noalias !6
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]]), !noalias !6
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]]), !noalias !6
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[C]], align 4, !noalias !14
 ; CHECK-NEXT:    [[ARRAYIDX_I_I:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 5
 ; CHECK-NEXT:    store float [[TMP0]], float* [[ARRAYIDX_I_I]], align 4, !alias.scope !9, !noalias !15
@@ -75,8 +75,8 @@
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[C]], align 4, !noalias !6
 ; CHECK-NEXT:    [[ARRAYIDX_I:%.*]] = getelementptr inbounds float, float* [[A]], i64 7
 ; CHECK-NEXT:    store float [[TMP1]], float* [[ARRAYIDX_I]], align 4, !noalias !6
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]]) #[[ATTR2]], !alias.scope !6
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]]) #[[ATTR2]], !alias.scope !6
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]]), !alias.scope !6
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]]), !alias.scope !6
 ; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[A]], align 4, !alias.scope !6, !noalias !22
 ; CHECK-NEXT:    [[ARRAYIDX_I_I1:%.*]] = getelementptr inbounds float, float* [[B]], i64 5
 ; CHECK-NEXT:    store float [[TMP2]], float* [[ARRAYIDX_I_I1]], align 4, !alias.scope !23, !noalias !20
diff --git a/llvm/test/Transforms/Inline/noalias2.ll b/llvm/test/Transforms/Inline/noalias2.ll
--- a/llvm/test/Transforms/Inline/noalias2.ll
+++ b/llvm/test/Transforms/Inline/noalias2.ll
@@ -71,8 +71,8 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
 ; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]]) #[[ATTR2:[0-9]+]]
-; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]]) #[[ATTR2]]
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[C]], align 4, !alias.scope !15, !noalias !16
 ; CHECK-NEXT:    [[ARRAYIDX_I_I:%.*]] = getelementptr inbounds float, float* [[A]], i64 5
 ; CHECK-NEXT:    store float [[TMP0]], float* [[ARRAYIDX_I_I]], align 4, !alias.scope !16, !noalias !15
diff --git a/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll b/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
--- a/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
+++ b/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
@@ -65,7 +65,7 @@
 ; CHECK-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP11]] to i32
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV2]], [[CONV]]
 ; CHECK-NEXT:    [[SHR:%.*]] = ashr i32 [[MUL]], 15
-; CHECK-NEXT:    [[TMP12:%.*]] = tail call i32 @llvm.smin.i32(i32 [[SHR]], i32 32767) #[[ATTR2:[0-9]+]]
+; CHECK-NEXT:    [[TMP12:%.*]] = tail call i32 @llvm.smin.i32(i32 [[SHR]], i32 32767)
 ; CHECK-NEXT:    [[CONV3:%.*]] = trunc i32 [[TMP12]] to i16
 ; CHECK-NEXT:    [[INCDEC_PTR4]] = getelementptr inbounds i16, i16* [[PDST_ADDR_05]], i32 1
 ; CHECK-NEXT:    store i16 [[CONV3]], i16* [[PDST_ADDR_05]], align 2
diff --git a/llvm/test/Transforms/PhaseOrdering/deletion-of-loops-that-became-side-effect-free.ll b/llvm/test/Transforms/PhaseOrdering/deletion-of-loops-that-became-side-effect-free.ll
--- a/llvm/test/Transforms/PhaseOrdering/deletion-of-loops-that-became-side-effect-free.ll
+++ b/llvm/test/Transforms/PhaseOrdering/deletion-of-loops-that-became-side-effect-free.ll
@@ -120,7 +120,7 @@
 ; O1-NEXT:    [[SIZE_06_I:%.*]] = phi i64 [ [[INC_I:%.*]], [[WHILE_BODY_I]] ], [ 0, [[ENTRY:%.*]] ]
 ; O1-NEXT:    [[P_ADDR_05_I:%.*]] = phi %struct.node* [ [[TMP0:%.*]], [[WHILE_BODY_I]] ], [ [[P]], [[ENTRY]] ]
 ; O1-NEXT:    [[CMP_I:%.*]] = icmp ne i64 [[SIZE_06_I]], -1
-; O1-NEXT:    call void @llvm.assume(i1 [[CMP_I]]) #[[ATTR3:[0-9]+]]
+; O1-NEXT:    call void @llvm.assume(i1 [[CMP_I]])
 ; O1-NEXT:    [[NEXT_I:%.*]] = getelementptr inbounds [[STRUCT_NODE:%.*]], %struct.node* [[P_ADDR_05_I]], i64 0, i32 0
 ; O1-NEXT:    [[TMP0]] = load %struct.node*, %struct.node** [[NEXT_I]], align 8
 ; O1-NEXT:    [[INC_I]] = add i64 [[SIZE_06_I]], 1