clang-misexpect: Profile Guided Validation of Performance Annotations in LLVM

This patch contains the basic functionality for reporting potentially
incorrect usage of __builtin_expect() by comparing the developer's
annotation against a collected PGO profile. A more detailed proposal and
discussion appears on the CFE-dev mailing list
(http://lists.llvm.org/pipermail/cfe-dev/2019-July/062971.html) and a
prototype of the initial frontend changes appear here in D65300

We revised the work in D65300 by moving the misexpect check into the
LLVM backend, and adding support for IR and sampling based profiles, in
addition to frontend instrumentation.

We add new misexpect metadata tags to those instructions directly
influenced by the llvm.expect intrinsic (branch, switch, and select)
when lowering the intrinsics. The misexpect metadata contains
information about the expected target of the intrinsic so that we can
check against the correct PGO counter when emitting diagnostics, and the
compiler's values for the LikelyBranchWeight and UnlikelyBranchWeight.
We use these branch weight values to determine when to emit the
diagnostic to the user.

A future patch should address the comment at the top of
LowerExpectIntrisic.cpp to hoist the LikelyBranchWeight and
UnlikelyBranchWeight values into a shared space that can be accessed
outside of the LowerExpectIntrinsic pass. Once that is done, the
misexpect metadata can be updated to be smaller.

In the long term, it is possible to reconstruct portions of the
misexpect metadata from the existing profile data. However, we have
avoided this to keep the code simple, and because some kind of metadata
tag will be required to identify which branch/switch/select instructions
are influenced by the use of llvm.expect

Patch By: paulkirth
Differential Revision: https://reviews.llvm.org/D66324

llvm-svn: 371584

14 files changed, 340 insertions, 0 deletions

diff --git a/clang/test/Profile/Inputs/misexpect-branch-nonconst-expect-arg.proftext b/clang/test/Profile/Inputs/misexpect-branch-nonconst-expect-arg.proftext
new file mode 100644
index 00000000000..a99351c06e5
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-branch-nonconst-expect-arg.proftext
@@ -0,0 +1,9 @@
+bar
+# Func Hash:
+11262309464
+# Num Counters:
+2
+# Counter Values:
+200000
+2
+
diff --git a/clang/test/Profile/Inputs/misexpect-branch.proftext b/clang/test/Profile/Inputs/misexpect-branch.proftext
new file mode 100644
index 00000000000..b809c1491fb
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-branch.proftext
@@ -0,0 +1,9 @@
+bar
+# Func Hash:
+45795613684824
+# Num Counters:
+2
+# Counter Values:
+200000
+0
+
diff --git a/clang/test/Profile/Inputs/misexpect-switch-default-only.proftext b/clang/test/Profile/Inputs/misexpect-switch-default-only.proftext
new file mode 100644
index 00000000000..ac59378e768
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-switch-default-only.proftext
@@ -0,0 +1,12 @@
+main
+# Func Hash:
+79676873694057560
+# Num Counters:
+5
+# Counter Values:
+1
+20
+20000
+20000
+20000
+
diff --git a/clang/test/Profile/Inputs/misexpect-switch-default.proftext b/clang/test/Profile/Inputs/misexpect-switch-default.proftext
new file mode 100644
index 00000000000..7b2d59781a1
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-switch-default.proftext
@@ -0,0 +1,16 @@
+main
+# Func Hash:
+8712453512413296413
+# Num Counters:
+9
+# Counter Values:
+1
+20000
+20000
+4066
+11889
+0
+0
+4045
+0
+
diff --git a/clang/test/Profile/Inputs/misexpect-switch-nonconst.proftext b/clang/test/Profile/Inputs/misexpect-switch-nonconst.proftext
new file mode 100644
index 00000000000..52b7b70cab9
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-switch-nonconst.proftext
@@ -0,0 +1,17 @@
+main
+# Func Hash:
+1965403898329309329
+# Num Counters:
+10
+# Counter Values:
+1
+20
+20000
+20000
+1
+0
+0
+0
+19999
+0
+
diff --git a/clang/test/Profile/Inputs/misexpect-switch.proftext b/clang/test/Profile/Inputs/misexpect-switch.proftext
new file mode 100644
index 00000000000..ce4c96b3e3a
--- /dev/null
+++ b/clang/test/Profile/Inputs/misexpect-switch.proftext
@@ -0,0 +1,16 @@
+main
+# Func Hash:
+1965403898329309329
+# Num Counters:
+9
+# Counter Values:
+1
+20
+20000
+20000
+12
+26
+0
+0
+19962
+
diff --git a/clang/test/Profile/misexpect-branch-cold.c b/clang/test/Profile/misexpect-branch-cold.c
new file mode 100644
index 00000000000..6d34f92a254
--- /dev/null
+++ b/clang/test/Profile/misexpect-branch-cold.c
@@ -0,0 +1,26 @@
+// Test that misexpect emits no warning when prediction is correct
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-branch.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -disable-llvm-passes -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect
+
+// expected-no-diagnostics
+#define likely(x) __builtin_expect(!!(x), 1)
+#define unlikely(x) __builtin_expect(!!(x), 0)
+
+int foo(int);
+int baz(int);
+int buzz();
+
+const int inner_loop = 100;
+const int outer_loop = 2000;
+
+int bar() {
+  int rando = buzz();
+  int x = 0;
+  if (unlikely(rando % (outer_loop * inner_loop) == 0)) {
+    x = baz(rando);
+  } else {
+    x = foo(50);
+  }
+  return x;
+}
diff --git a/clang/test/Profile/misexpect-branch-nonconst-expected-val.c b/clang/test/Profile/misexpect-branch-nonconst-expected-val.c
new file mode 100644
index 00000000000..c5167b9a2a0
--- /dev/null
+++ b/clang/test/Profile/misexpect-branch-nonconst-expected-val.c
@@ -0,0 +1,23 @@
+// Test that misexpect emits no warning when condition is not a compile-time constant
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-branch-nonconst-expect-arg.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -disable-llvm-passes -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect
+
+// expected-no-diagnostics
+int foo(int);
+int baz(int);
+int buzz();
+
+const int inner_loop = 100;
+const int outer_loop = 2000;
+
+int bar() {
+  int rando = buzz();
+  int x = 0;
+  if (__builtin_expect(rando % (outer_loop * inner_loop) == 0, buzz())) {
+    x = baz(rando);
+  } else {
+    x = foo(50);
+  }
+  return x;
+}
diff --git a/clang/test/Profile/misexpect-branch-unpredictable.c b/clang/test/Profile/misexpect-branch-unpredictable.c
new file mode 100644
index 00000000000..6c4f90146a2
--- /dev/null
+++ b/clang/test/Profile/misexpect-branch-unpredictable.c
@@ -0,0 +1,25 @@
+// Test that misexpect emits no warning when prediction is correct
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-branch.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -disable-llvm-passes -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect
+
+// expected-no-diagnostics
+#define unpredictable(x) __builtin_unpredictable(!!(x))
+
+int foo(int);
+int baz(int);
+int buzz();
+
+const int inner_loop = 100;
+const int outer_loop = 2000;
+
+int bar() {
+  int rando = buzz();
+  int x = 0;
+  if (unpredictable(rando % (outer_loop * inner_loop) == 0)) {
+    x = baz(rando);
+  } else {
+    x = foo(50);
+  }
+  return x;
+}
diff --git a/clang/test/Profile/misexpect-branch.c b/clang/test/Profile/misexpect-branch.c
new file mode 100644
index 00000000000..503aa7398e9
--- /dev/null
+++ b/clang/test/Profile/misexpect-branch.c
@@ -0,0 +1,28 @@
+// Test that misexpect detects mis-annotated branches
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-branch.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify=imprecise -Wmisexpect
+// RUN: %clang_cc1 %s -O2 -o - -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify=exact -Wmisexpect -debug-info-kind=line-tables-only
+// RUN: %clang_cc1 %s -O2 -o - -disable-llvm-passes -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify=foo
+
+// foo-no-diagnostics
+#define likely(x) __builtin_expect(!!(x), 1)
+#define unlikely(x) __builtin_expect(!!(x), 0)
+
+int foo(int);
+int baz(int);
+int buzz();
+
+const int inner_loop = 100;
+const int outer_loop = 2000;
+
+int bar() { // imprecise-warning-re {{Potential performance regression from use of __builtin_expect(): Annotation was correct on {{.+}}% ({{[0-9]+ / [0-9]+}}) of profiled executions.}}
+  int rando = buzz();
+  int x = 0;
+  if (likely(rando % (outer_loop * inner_loop) == 0)) { // exact-warning-re {{Potential performance regression from use of __builtin_expect(): Annotation was correct on {{.+}}% ({{[0-9]+ / [0-9]+}}) of profiled executions.}}
+    x = baz(rando);
+  } else {
+    x = foo(50);
+  }
+  return x;
+}
diff --git a/clang/test/Profile/misexpect-switch-default.c b/clang/test/Profile/misexpect-switch-default.c
new file mode 100644
index 00000000000..3d1079d79f9
--- /dev/null
+++ b/clang/test/Profile/misexpect-switch-default.c
@@ -0,0 +1,40 @@
+// Test that misexpect detects mis-annotated switch statements for default case
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-switch-default.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect -debug-info-kind=line-tables-only
+
+int sum(int *buff, int size);
+int random_sample(int *buff, int size);
+int rand();
+void init_arry();
+
+const int inner_loop = 1000;
+const int outer_loop = 20;
+const int arry_size = 25;
+
+int arry[arry_size] = {0};
+
+int main() {
+  init_arry();
+  int val = 0;
+  int j;
+  for (j = 0; j < outer_loop * inner_loop; ++j) {
+    unsigned condition = rand() % 5;
+    switch (__builtin_expect(condition, 6)) { // expected-warning-re {{Potential performance regression from use of __builtin_expect(): Annotation was correct on {{.+}}% ({{[0-9]+ / [0-9]+}}) of profiled executions.}}
+    case 0:
+      val += sum(arry, arry_size);
+      break;
+    case 1:
+    case 2:
+    case 3:
+      break;
+    case 4:
+      val += random_sample(arry, arry_size);
+      break;
+    default:
+      __builtin_unreachable();
+    } // end switch
+  }   // end outer_loop
+
+  return 0;
+}
diff --git a/clang/test/Profile/misexpect-switch-nonconst.c b/clang/test/Profile/misexpect-switch-nonconst.c
new file mode 100644
index 00000000000..fb719c5a2d1
--- /dev/null
+++ b/clang/test/Profile/misexpect-switch-nonconst.c
@@ -0,0 +1,43 @@
+// Test that misexpect emits no warning when switch condition is non-const
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-switch-nonconst.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -disable-llvm-passes -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify
+
+// expected-no-diagnostics
+int sum(int *buff, int size);
+int random_sample(int *buff, int size);
+int rand();
+void init_arry();
+
+const int inner_loop = 1000;
+const int outer_loop = 20;
+const int arry_size = 25;
+
+int arry[arry_size] = {0};
+
+int main() {
+  init_arry();
+  int val = 0;
+
+  int j, k;
+  for (j = 0; j < outer_loop; ++j) {
+    for (k = 0; k < inner_loop; ++k) {
+      unsigned condition = rand() % 10000;
+      switch (__builtin_expect(condition, rand())) {
+      case 0:
+        val += sum(arry, arry_size);
+        break;
+      case 1:
+      case 2:
+      case 3:
+      case 4:
+        val += random_sample(arry, arry_size);
+        break;
+      default:
+        __builtin_unreachable();
+      } // end switch
+    }   // end inner_loop
+  }     // end outer_loop
+
+  return 0;
+}
diff --git a/clang/test/Profile/misexpect-switch-only-default-case.c b/clang/test/Profile/misexpect-switch-only-default-case.c
new file mode 100644
index 00000000000..3886472e2b4
--- /dev/null
+++ b/clang/test/Profile/misexpect-switch-only-default-case.c
@@ -0,0 +1,35 @@
+// Test that misexpect emits no warning when there is only one switch case
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-switch-default-only.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect -debug-info-kind=line-tables-only
+
+// expected-no-diagnostics
+int sum(int *buff, int size);
+int random_sample(int *buff, int size);
+int rand();
+void init_arry();
+
+const int inner_loop = 1000;
+const int outer_loop = 20;
+const int arry_size = 25;
+
+int arry[arry_size] = {0};
+
+int main() {
+  init_arry();
+  int val = 0;
+
+  int j, k;
+  for (j = 0; j < outer_loop; ++j) {
+    for (k = 0; k < inner_loop; ++k) {
+      unsigned condition = rand() % 10000;
+      switch (__builtin_expect(condition, 0)) {
+      default:
+        val += random_sample(arry, arry_size);
+        break;
+      }; // end switch
+    }    // end inner_loop
+  }      // end outer_loop
+
+  return 0;
+}
diff --git a/clang/test/Profile/misexpect-switch.c b/clang/test/Profile/misexpect-switch.c
new file mode 100644
index 00000000000..a7f01bcc998
--- /dev/null
+++ b/clang/test/Profile/misexpect-switch.c
@@ -0,0 +1,41 @@
+// Test that misexpect detects mis-annotated switch statements
+
+// RUN: llvm-profdata merge %S/Inputs/misexpect-switch.proftext -o %t.profdata
+// RUN: %clang_cc1 %s -O2 -o - -emit-llvm -fprofile-instrument-use-path=%t.profdata -verify -Wmisexpect -debug-info-kind=line-tables-only
+
+int sum(int *buff, int size);
+int random_sample(int *buff, int size);
+int rand();
+void init_arry();
+
+const int inner_loop = 1000;
+const int outer_loop = 20;
+const int arry_size = 25;
+
+int arry[arry_size] = {0};
+
+int main() {
+  init_arry();
+  int val = 0;
+
+  int j, k;
+  for (j = 0; j < outer_loop; ++j) {
+    for (k = 0; k < inner_loop; ++k) {
+      unsigned condition = rand() % 10000;
+      switch (__builtin_expect(condition, 0)) { // expected-warning-re {{Potential performance regression from use of __builtin_expect(): Annotation was correct on {{.+}}% ({{[0-9]+ / [0-9]+}}) of profiled executions.}}
+      case 0:
+        val += sum(arry, arry_size);
+        break;
+      case 1:
+      case 2:
+      case 3:
+        break;
+      default:
+        val += random_sample(arry, arry_size);
+        break;
+      } // end switch
+    }   // end inner_loop
+  }     // end outer_loop
+
+  return 0;
+}