[funcattrs] Infer writeonly argument attribute

This change extends the current logic for inferring readonly and readnone argument attributes to also infer writeonly.

This change is deliberately minimal; there's a couple of areas for follow up.
* I left out all call handling and thus any benefit from the SCC walk. When examining the test changes, I realized the existing code is imprecise, and am going to fix that in it's own revision before adding in the writeonly handling. (Mostly because updating the tests is hard when I, the human, can't figure out whether the result is correct.)
* I left out handling for storing a value (as opposed to storing to a pointer). This should benefit readonly/readnone as well, and applies to a bunch of other instructions. Seemed worth having as a separate review.

Differential Revision: https://reviews.llvm.org/D114963

diff --git a/clang/test/CodeGen/SystemZ/systemz-inline-asm.c b/clang/test/CodeGen/SystemZ/systemz-inline-asm.c
index 0994b11825fe8a3667c78f06e77756729cf5bc4a..2a656eaf4a2a876c7febdfa7cd6ceac96846c779 100644 (file)
--- a/clang/test/CodeGen/SystemZ/systemz-inline-asm.c
+++ b/clang/test/CodeGen/SystemZ/systemz-inline-asm.c
@@ -123,7 +123,7 @@ double test_f64(double f, double g) {
 long double test_f128(long double f, long double g) {
   asm("axbr %0, %2" : "=f" (f) : "0" (f), "f" (g));
   return f;
-// CHECK: define{{.*}} void @test_f128(fp128* noalias nocapture sret(fp128) align 8 [[DEST:%.*]], fp128* nocapture readonly %0, fp128* nocapture readonly %1)
+// CHECK: define{{.*}} void @test_f128(fp128* noalias nocapture writeonly sret(fp128) align 8 [[DEST:%.*]], fp128* nocapture readonly %0, fp128* nocapture readonly %1)
 // CHECK: %f = load fp128, fp128* %0
 // CHECK: %g = load fp128, fp128* %1
 // CHECK: [[RESULT:%.*]] = tail call fp128 asm "axbr $0, $2", "=f,0,f"(fp128 %f, fp128 %g)

diff --git a/clang/test/CodeGen/aarch64-sve-acle-__ARM_FEATURE_SVE_VECTOR_OPERATORS.c b/clang/test/CodeGen/aarch64-sve-acle-__ARM_FEATURE_SVE_VECTOR_OPERATORS.c
index 071932a27a97bd1906cf0750a627285640632003..4fa72c666bf9800b5357797c1197784451be66bd 100644 (file)
--- a/clang/test/CodeGen/aarch64-sve-acle-__ARM_FEATURE_SVE_VECTOR_OPERATORS.c
+++ b/clang/test/CodeGen/aarch64-sve-acle-__ARM_FEATURE_SVE_VECTOR_OPERATORS.c
@@ -59,7 +59,7 @@ typedef int8_t vec_int8 __attribute__((vector_size(N / 8)));
 // CHECK128-NEXT:    ret <16 x i8> [[CASTFIXEDSVE]]
 
 // CHECK-LABEL: define{{.*}} void @f2(
-// CHECK-SAME:   <[[#div(VBITS,8)]] x i8>* noalias nocapture sret(<[[#div(VBITS,8)]] x i8>) align 16 %agg.result, <[[#div(VBITS,8)]] x i8>* nocapture readonly %0)
+// CHECK-SAME:   <[[#div(VBITS,8)]] x i8>* noalias nocapture writeonly sret(<[[#div(VBITS,8)]] x i8>) align 16 %agg.result, <[[#div(VBITS,8)]] x i8>* nocapture readonly %0)
 // CHECK-NEXT: entry:
 // CHECK-NEXT:   [[X:%.*]] = load <[[#div(VBITS,8)]] x i8>, <[[#div(VBITS,8)]] x i8>* [[TMP0:%.*]], align 16, [[TBAA6:!tbaa !.*]]
 // CHECK-NEXT:   [[TMP1:%.*]] = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)

diff --git a/clang/test/CodeGen/arm-vfp16-arguments2.cpp b/clang/test/CodeGen/arm-vfp16-arguments2.cpp
index 35b716230a096f8e8fd4333f6ab70f1c62e5dc20..b1b466c48dd315fd39937817c5338e32b7c81fc5 100644 (file)
--- a/clang/test/CodeGen/arm-vfp16-arguments2.cpp
+++ b/clang/test/CodeGen/arm-vfp16-arguments2.cpp
@@ -37,27 +37,27 @@ struct S5 : B1 {
   B1 M[1];
 };
 
-// CHECK-SOFT: define{{.*}} void @_Z2f12S1(%struct.S1* noalias nocapture sret(%struct.S1) align 8 %agg.result, [2 x i64] %s1.coerce)
+// CHECK-SOFT: define{{.*}} void @_Z2f12S1(%struct.S1* noalias nocapture writeonly sret(%struct.S1) align 8 %agg.result, [2 x i64] %s1.coerce)
 // CHECK-HARD: define{{.*}} arm_aapcs_vfpcc [2 x <2 x i32>] @_Z2f12S1([2 x <2 x i32>] returned %s1.coerce)
 // CHECK-FULL: define{{.*}} arm_aapcs_vfpcc %struct.S1 @_Z2f12S1(%struct.S1 returned %s1.coerce)
 struct S1 f1(struct S1 s1) { return s1; }
 
-// CHECK-SOFT: define{{.*}} void @_Z2f22S2(%struct.S2* noalias nocapture sret(%struct.S2) align 8 %agg.result, [4 x i32] %s2.coerce)
+// CHECK-SOFT: define{{.*}} void @_Z2f22S2(%struct.S2* noalias nocapture writeonly sret(%struct.S2) align 8 %agg.result, [4 x i32] %s2.coerce)
 // CHECK-HARD: define{{.*}} arm_aapcs_vfpcc [2 x <2 x i32>] @_Z2f22S2([2 x <2 x i32>] returned %s2.coerce)
 // CHECK-FULL: define{{.*}} arm_aapcs_vfpcc %struct.S2 @_Z2f22S2(%struct.S2 returned %s2.coerce)
 struct S2 f2(struct S2 s2) { return s2; }
 
-// CHECK-SOFT: define{{.*}} void @_Z2f32S3(%struct.S3* noalias nocapture sret(%struct.S3) align 8 %agg.result, [2 x i64] %s3.coerce)
+// CHECK-SOFT: define{{.*}} void @_Z2f32S3(%struct.S3* noalias nocapture writeonly sret(%struct.S3) align 8 %agg.result, [2 x i64] %s3.coerce)
 // CHECK-HARD: define{{.*}} arm_aapcs_vfpcc [2 x <2 x i32>] @_Z2f32S3([2 x <2 x i32>] returned %s3.coerce)
 // CHECK-FULL: define{{.*}} arm_aapcs_vfpcc %struct.S3 @_Z2f32S3(%struct.S3 returned %s3.coerce)
 struct S3 f3(struct S3 s3) { return s3; }
 
-// CHECK-SOFT: define{{.*}} void @_Z2f42S4(%struct.S4* noalias nocapture sret(%struct.S4) align 8 %agg.result, [2 x i64] %s4.coerce)
+// CHECK-SOFT: define{{.*}} void @_Z2f42S4(%struct.S4* noalias nocapture writeonly sret(%struct.S4) align 8 %agg.result, [2 x i64] %s4.coerce)
 // CHECK-HARD: define{{.*}} arm_aapcs_vfpcc [2 x <2 x i32>] @_Z2f42S4([2 x <2 x i32>] returned %s4.coerce)
 // CHECK-FULL: define{{.*}} arm_aapcs_vfpcc %struct.S4 @_Z2f42S4(%struct.S4 returned %s4.coerce)
 struct S4 f4(struct S4 s4) { return s4; }
 
-// CHECK-SOFT: define{{.*}} void @_Z2f52S5(%struct.S5* noalias nocapture sret(%struct.S5) align 8 %agg.result, [2 x i64] %s5.coerce)
+// CHECK-SOFT: define{{.*}} void @_Z2f52S5(%struct.S5* noalias nocapture writeonly sret(%struct.S5) align 8 %agg.result, [2 x i64] %s5.coerce)
 // CHECK-HARD: define{{.*}} arm_aapcs_vfpcc %struct.S5 @_Z2f52S5(%struct.S5 returned %s5.coerce)
 // CHECK-FULL: define{{.*}} arm_aapcs_vfpcc %struct.S5 @_Z2f52S5(%struct.S5 returned %s5.coerce)
 struct S5 f5(struct S5 s5) { return s5; }

diff --git a/clang/test/CodeGen/mips-vector-return.c b/clang/test/CodeGen/mips-vector-return.c
index dd3c400c0197289a6307429cba37f130a13daa4e..dfb0ca304dbe776b48ea5a8a7aadfbed1544e494 100644 (file)
--- a/clang/test/CodeGen/mips-vector-return.c
+++ b/clang/test/CodeGen/mips-vector-return.c
@@ -8,14 +8,14 @@ typedef float  v4sf __attribute__ ((__vector_size__ (16)));
 typedef double v4df __attribute__ ((__vector_size__ (32)));
 typedef int v4i32 __attribute__ ((__vector_size__ (16)));
 
-// O32-LABEL: define dso_local void @test_v4sf(<4 x float>* noalias nocapture sret
+// O32-LABEL: define dso_local void @test_v4sf(<4 x float>* noalias nocapture writeonly sret
 // N64: define inreg { i64, i64 } @test_v4sf
 v4sf test_v4sf(float a) {
   return (v4sf){0.0f, a, 0.0f, 0.0f};
 }
 
-// O32-LABEL: define dso_local void @test_v4df(<4 x double>* noalias nocapture sret
-// N64-LABEL: define void @test_v4df(<4 x double>* noalias nocapture sret
+// O32-LABEL: define dso_local void @test_v4df(<4 x double>* noalias nocapture writeonly sret
+// N64-LABEL: define void @test_v4df(<4 x double>* noalias nocapture writeonly sret
 v4df test_v4df(double a) {
   return (v4df){0.0, a, 0.0, 0.0};
 }

diff --git a/clang/test/CodeGen/mips64-nontrivial-return.cpp b/clang/test/CodeGen/mips64-nontrivial-return.cpp
index 2164b20c184a86f7b90a486ffcd8b1b70eabaac5..31a3c1e35fcf5a4418369a70fe9fe39e0a7ac30d 100644 (file)
--- a/clang/test/CodeGen/mips64-nontrivial-return.cpp
+++ b/clang/test/CodeGen/mips64-nontrivial-return.cpp
@@ -10,7 +10,7 @@ class D : public B {
 
 extern D gd0;
 
-// CHECK: _Z4foo1v(%class.D* noalias nocapture sret
+// CHECK: _Z4foo1v(%class.D* noalias nocapture writeonly sret
 
 D foo1(void) {
   return gd0;

diff --git a/clang/test/CodeGen/ms-mixed-ptr-sizes.c b/clang/test/CodeGen/ms-mixed-ptr-sizes.c
index ececa42a4c4dd8f4331805f322b49611ee1e30b5..294a8910e13e3061a55114361776cd6d0176db98 100644 (file)
--- a/clang/test/CodeGen/ms-mixed-ptr-sizes.c
+++ b/clang/test/CodeGen/ms-mixed-ptr-sizes.c
@@ -7,32 +7,32 @@ struct Foo {
 };
 void use_foo(struct Foo *f);
 void test_sign_ext(struct Foo *f, int * __ptr32 __sptr i) {
-// X64-LABEL: define dso_local void @test_sign_ext({{.*}}i32 addrspace(270)* %i)
-// X86-LABEL: define dso_local void @test_sign_ext(%struct.Foo* %f, i32* %i)
+// X64-LABEL: define dso_local void @test_sign_ext({{.*}}i32 addrspace(270)* writeonly %i)
+// X86-LABEL: define dso_local void @test_sign_ext(%struct.Foo* %f, i32* writeonly %i)
 // X64: %{{.+}} = addrspacecast i32 addrspace(270)* %i to i32*
 // X86: %{{.+}} = addrspacecast i32* %i to i32 addrspace(272)*
   f->p64 = i;
   use_foo(f);
 }
 void test_zero_ext(struct Foo *f, int * __ptr32 __uptr i) {
-// X64-LABEL: define dso_local void @test_zero_ext({{.*}}i32 addrspace(271)* %i)
-// X86-LABEL: define dso_local void @test_zero_ext({{.*}}i32 addrspace(271)* %i)
+// X64-LABEL: define dso_local void @test_zero_ext({{.*}}i32 addrspace(271)* writeonly %i)
+// X86-LABEL: define dso_local void @test_zero_ext({{.*}}i32 addrspace(271)* writeonly %i)
 // X64: %{{.+}} = addrspacecast i32 addrspace(271)* %i to i32*
 // X86: %{{.+}} = addrspacecast i32 addrspace(271)* %i to i32 addrspace(272)*
   f->p64 = i;
   use_foo(f);
 }
 void test_trunc(struct Foo *f, int * __ptr64 i) {
-// X64-LABEL: define dso_local void @test_trunc(%struct.Foo* %f, i32* %i)
-// X86-LABEL: define dso_local void @test_trunc({{.*}}i32 addrspace(272)* %i)
+// X64-LABEL: define dso_local void @test_trunc(%struct.Foo* %f, i32* writeonly %i)
+// X86-LABEL: define dso_local void @test_trunc({{.*}}i32 addrspace(272)* writeonly %i)
 // X64: %{{.+}} = addrspacecast i32* %i to i32 addrspace(270)*
 // X86: %{{.+}} = addrspacecast i32 addrspace(272)* %i to i32*
   f->p32 = i;
   use_foo(f);
 }
 void test_noop(struct Foo *f, int * __ptr32 i) {
-// X64-LABEL: define dso_local void @test_noop({{.*}}i32 addrspace(270)* %i)
-// X86-LABEL: define dso_local void @test_noop({{.*}}i32* %i)
+// X64-LABEL: define dso_local void @test_noop({{.*}}i32 addrspace(270)* writeonly %i)
+// X86-LABEL: define dso_local void @test_noop({{.*}}i32* writeonly %i)
 // X64-NOT: addrspacecast
 // X86-NOT: addrspacecast
   f->p32 = i;
@@ -40,8 +40,8 @@ void test_noop(struct Foo *f, int * __ptr32 i) {
 }
 
 void test_other(struct Foo *f, __attribute__((address_space(10))) int *i) {
-// X64-LABEL: define dso_local void @test_other({{.*}}i32 addrspace(10)* %i)
-// X86-LABEL: define dso_local void @test_other({{.*}}i32 addrspace(10)* %i)
+// X64-LABEL: define dso_local void @test_other({{.*}}i32 addrspace(10)* writeonly %i)
+// X86-LABEL: define dso_local void @test_other({{.*}}i32 addrspace(10)* writeonly %i)
 // X64: %{{.+}} = addrspacecast i32 addrspace(10)* %i to i32 addrspace(270)*
 // X86: %{{.+}} = addrspacecast i32 addrspace(10)* %i to i32*
   f->p32 = (int * __ptr32)i;

diff --git a/clang/test/CodeGenOpenCL/amdgpu-abi-struct-coerce.cl b/clang/test/CodeGenOpenCL/amdgpu-abi-struct-coerce.cl
index 0a8abf9a958e8d0a76051a526c3d070d2338e39c..17333cc80e141039bfa0d9025f2a82cff640a740 100644 (file)
--- a/clang/test/CodeGenOpenCL/amdgpu-abi-struct-coerce.cl
+++ b/clang/test/CodeGenOpenCL/amdgpu-abi-struct-coerce.cl
@@ -439,7 +439,7 @@ different_size_type_pair func_different_size_type_pair_ret()
   return s;
 }
 
-// CHECK: define{{.*}} void @func_flexible_array_ret(%struct.flexible_array addrspace(5)* noalias nocapture sret(%struct.flexible_array) align 4 %agg.result)
+// CHECK: define{{.*}} void @func_flexible_array_ret(%struct.flexible_array addrspace(5)* noalias nocapture writeonly sret(%struct.flexible_array) align 4 %agg.result)
 flexible_array func_flexible_array_ret()
 {
   flexible_array s = { 0 };

diff --git a/clang/test/CodeGenOpenCL/amdgpu-call-kernel.cl b/clang/test/CodeGenOpenCL/amdgpu-call-kernel.cl
index 9ecdeeb32c5b2c4d0a4e0c23d133c25037d36297..7bce2f9454a54b8537cf1b88eb9a1a38bf2ec68e 100755 (executable)
--- a/clang/test/CodeGenOpenCL/amdgpu-call-kernel.cl
+++ b/clang/test/CodeGenOpenCL/amdgpu-call-kernel.cl
@@ -1,6 +1,6 @@
 // REQUIRES: amdgpu-registered-target
 // RUN: %clang_cc1 -triple amdgcn-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
-// CHECK: define{{.*}} amdgpu_kernel void @test_call_kernel(i32 addrspace(1)* nocapture %out)
+// CHECK: define{{.*}} amdgpu_kernel void @test_call_kernel(i32 addrspace(1)* nocapture writeonly %out)
 // CHECK: store i32 4, i32 addrspace(1)* %out, align 4
 
 kernel void test_kernel(global int *out)

diff --git a/clang/test/CodeGenOpenCL/kernels-have-spir-cc-by-default.cl b/clang/test/CodeGenOpenCL/kernels-have-spir-cc-by-default.cl
index 3b5a784d7cdf0de5b0ec98fb50ece5caa3b305ee..641e63c183a105f261d472848b929ac1a36a7570 100644 (file)
--- a/clang/test/CodeGenOpenCL/kernels-have-spir-cc-by-default.cl
+++ b/clang/test/CodeGenOpenCL/kernels-have-spir-cc-by-default.cl
@@ -28,7 +28,7 @@ kernel void test_single(int_single input, global int* output) {
 // CHECK: spir_kernel
 // AMDGCN: define{{.*}} amdgpu_kernel void @test_single
 // CHECK: struct.int_single* nocapture {{.*}} byval(%struct.int_single)
-// CHECK: i32* nocapture %output
+// CHECK: i32* nocapture writeonly %output
  output[0] = input.a;
 }
 
@@ -36,7 +36,7 @@ kernel void test_pair(int_pair input, global int* output) {
 // CHECK: spir_kernel
 // AMDGCN: define{{.*}} amdgpu_kernel void @test_pair
 // CHECK: struct.int_pair* nocapture {{.*}} byval(%struct.int_pair)
-// CHECK: i32* nocapture %output
+// CHECK: i32* nocapture writeonly %output
  output[0] = (int)input.a;
  output[1] = (int)input.b;
 }
@@ -45,7 +45,7 @@ kernel void test_kernel(test_struct input, global int* output) {
 // CHECK: spir_kernel
 // AMDGCN: define{{.*}} amdgpu_kernel void @test_kernel
 // CHECK: struct.test_struct* nocapture {{.*}} byval(%struct.test_struct)
-// CHECK: i32* nocapture %output
+// CHECK: i32* nocapture writeonly %output
  output[0] = input.elementA;
  output[1] = input.elementB;
  output[2] = (int)input.elementC;
@@ -59,7 +59,7 @@ kernel void test_kernel(test_struct input, global int* output) {
 void test_function(int_pair input, global int* output) {
 // CHECK-NOT: spir_kernel
 // AMDGCN-NOT: define{{.*}} amdgpu_kernel void @test_function
-// CHECK: i64 %input.coerce0, i64 %input.coerce1, i32* nocapture %output
+// CHECK: i64 %input.coerce0, i64 %input.coerce1, i32* nocapture writeonly %output
  output[0] = (int)input.a;
  output[1] = (int)input.b;
 }

diff --git a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
index cde78713b5541e8a524880a5a4883890da26ab71..1ad0055b563824cefde7e66fb3033a02c6fc5b58 100644 (file)
--- a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
+++ b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
@@ -76,6 +76,7 @@ STATISTIC(NumNoCapture, "Number of arguments marked nocapture");
 STATISTIC(NumReturned, "Number of arguments marked returned");
 STATISTIC(NumReadNoneArg, "Number of arguments marked readnone");
 STATISTIC(NumReadOnlyArg, "Number of arguments marked readonly");
+STATISTIC(NumWriteOnlyArg, "Number of arguments marked writeonly");
 STATISTIC(NumNoAlias, "Number of function returns marked noalias");
 STATISTIC(NumNonNullReturn, "Number of function returns marked nonnull");
 STATISTIC(NumNoRecurse, "Number of functions marked as norecurse");
@@ -649,8 +650,8 @@ struct GraphTraits<ArgumentGraph *> : public GraphTraits<ArgumentGraphNode *> {
 
 /// Returns Attribute::None, Attribute::ReadOnly or Attribute::ReadNone.
 static Attribute::AttrKind
-determinePointerReadAttrs(Argument *A,
-                          const SmallPtrSet<Argument *, 8> &SCCNodes) {
+determinePointerAccessAttrs(Argument *A,
+                            const SmallPtrSet<Argument *, 8> &SCCNodes) {
   SmallVector<Use *, 32> Worklist;
   SmallPtrSet<Use *, 32> Visited;
 
@@ -659,7 +660,7 @@ determinePointerReadAttrs(Argument *A,
     return Attribute::None;
 
   bool IsRead = false;
-  // We don't need to track IsWritten. If A is written to, return immediately.
+  bool IsWrite = false;
 
   for (Use &U : A->uses()) {
     Visited.insert(&U);
@@ -667,6 +668,10 @@ determinePointerReadAttrs(Argument *A,
   }
 
   while (!Worklist.empty()) {
+    if (IsWrite && IsRead)
+      // No point in searching further..
+      return Attribute::None;
+
     Use *U = Worklist.pop_back_val();
     Instruction *I = cast<Instruction>(U->getUser());
 
@@ -763,6 +768,15 @@ determinePointerReadAttrs(Argument *A,
       IsRead = true;
       break;
 
+    case Instruction::Store:
+      // A volatile store has side effects beyond what writeonly can be relied
+      // upon.
+      if (cast<StoreInst>(I)->isVolatile())
+        return Attribute::None;
+
+      IsWrite = true;
+      break;
+
     case Instruction::ICmp:
     case Instruction::Ret:
       break;
@@ -772,7 +786,14 @@ determinePointerReadAttrs(Argument *A,
     }
   }
 
-  return IsRead ? Attribute::ReadOnly : Attribute::ReadNone;
+  if (IsWrite && IsRead)
+    return Attribute::None;
+  else if (IsRead)
+    return Attribute::ReadOnly;
+  else if (IsWrite)
+    return Attribute::WriteOnly;
+  else
+    return Attribute::ReadNone;
 }
 
 /// Deduce returned attributes for the SCC.
@@ -865,9 +886,10 @@ static bool addArgumentAttrsFromCallsites(Function &F) {
   return Changed;
 }
 
-static bool addReadAttr(Argument *A, Attribute::AttrKind R) {
-  assert((R == Attribute::ReadOnly || R == Attribute::ReadNone)
-         && "Must be a Read attribute.");
+static bool addAccessAttr(Argument *A, Attribute::AttrKind R) {
+  assert((R == Attribute::ReadOnly || R == Attribute::ReadNone ||
+          R == Attribute::WriteOnly)
+         && "Must be an access attribute.");
   assert(A && "Argument must not be null.");
 
   // If the argument already has the attribute, nothing needs to be done.
@@ -880,7 +902,12 @@ static bool addReadAttr(Argument *A, Attribute::AttrKind R) {
   A->removeAttr(Attribute::ReadOnly);
   A->removeAttr(Attribute::ReadNone);
   A->addAttr(R);
-  R == Attribute::ReadOnly ? ++NumReadOnlyArg : ++NumReadNoneArg;
+  if (R == Attribute::ReadOnly)
+    ++NumReadOnlyArg;
+  else if (R == Attribute::WriteOnly)
+    ++NumWriteOnlyArg;
+  else
+    ++NumReadNoneArg;
   return true;
 }
 
@@ -945,15 +972,15 @@ static void addArgumentAttrs(const SCCNodeSet &SCCNodes,
         // Otherwise, it's captured. Don't bother doing SCC analysis on it.
       }
       if (!HasNonLocalUses && !A->onlyReadsMemory()) {
-        // Can we determine that it's readonly/readnone without doing an SCC?
-        // Note that we don't allow any calls at all here, or else our result
-        // will be dependent on the iteration order through the functions in the
-        // SCC.
+        // Can we determine that it's readonly/readnone/writeonly without doing
+        // an SCC? Note that we don't allow any calls at all here, or else our
+        // result will be dependent on the iteration order through the
+        // functions in the SCC.
         SmallPtrSet<Argument *, 8> Self;
         Self.insert(&*A);
-        Attribute::AttrKind R = determinePointerReadAttrs(&*A, Self);
+        Attribute::AttrKind R = determinePointerAccessAttrs(&*A, Self);
         if (R != Attribute::None)
-          if (addReadAttr(A, R))
+          if (addAccessAttr(A, R))
             Changed.insert(F);
       }
     }
@@ -1023,10 +1050,10 @@ static void addArgumentAttrs(const SCCNodeSet &SCCNodes,
       Changed.insert(A->getParent());
     }
 
-    // We also want to compute readonly/readnone. With a small number of false
-    // negatives, we can assume that any pointer which is captured isn't going
-    // to be provably readonly or readnone, since by definition we can't
-    // analyze all uses of a captured pointer.
+    // We also want to compute readonly/readnone/writeonly. With a small number
+    // of false negatives, we can assume that any pointer which is captured
+    // isn't going to be provably readonly or readnone, since by definition
+    // we can't analyze all uses of a captured pointer.
     //
     // The false negatives happen when the pointer is captured by a function
     // that promises readonly/readnone behaviour on the pointer, then the
@@ -1034,24 +1061,28 @@ static void addArgumentAttrs(const SCCNodeSet &SCCNodes,
     // Also, a readonly/readnone pointer may be returned, but returning a
     // pointer is capturing it.
 
-    Attribute::AttrKind ReadAttr = Attribute::ReadNone;
-    for (unsigned i = 0, e = ArgumentSCC.size(); i != e; ++i) {
+    auto meetAccessAttr = [](Attribute::AttrKind A, Attribute::AttrKind B) {
+      if (A == B)
+        return A;
+      if (A == Attribute::ReadNone)
+        return B;
+      if (B == Attribute::ReadNone)
+        return A;
+      return Attribute::None;
+    };
+
+    Attribute::AttrKind AccessAttr = Attribute::ReadNone;
+    for (unsigned i = 0, e = ArgumentSCC.size();
+         i != e && AccessAttr != Attribute::None; ++i) {
       Argument *A = ArgumentSCC[i]->Definition;
-      Attribute::AttrKind K = determinePointerReadAttrs(A, ArgumentSCCNodes);
-      if (K == Attribute::ReadNone)
-        continue;
-      if (K == Attribute::ReadOnly) {
-        ReadAttr = Attribute::ReadOnly;
-        continue;
-      }
-      ReadAttr = K;
-      break;
+      Attribute::AttrKind K = determinePointerAccessAttrs(A, ArgumentSCCNodes);
+      AccessAttr = meetAccessAttr(AccessAttr, K);
     }
 
-    if (ReadAttr != Attribute::None) {
+    if (AccessAttr != Attribute::None) {
       for (unsigned i = 0, e = ArgumentSCC.size(); i != e; ++i) {
         Argument *A = ArgumentSCC[i]->Definition;
-        if (addReadAttr(A, ReadAttr))
+        if (addAccessAttr(A, AccessAttr))
           Changed.insert(A->getParent());
       }
     }

diff --git a/llvm/test/Analysis/TypeBasedAliasAnalysis/functionattrs.ll b/llvm/test/Analysis/TypeBasedAliasAnalysis/functionattrs.ll
index 7f1841f13f23189185114a409dee55faea59489f..65c0e27d5f468875250f4b066a46693160e754b2 100644 (file)
--- a/llvm/test/Analysis/TypeBasedAliasAnalysis/functionattrs.ll
+++ b/llvm/test/Analysis/TypeBasedAliasAnalysis/functionattrs.ll
@@ -15,7 +15,7 @@ define void @test0_yes(i32* %p) nounwind {
   ret void
 }
 
-; CHECK: define void @test0_no(i32* nocapture %p) #1 {
+; CHECK: define void @test0_no(i32* nocapture writeonly %p) #1 {
 define void @test0_no(i32* %p) nounwind {
   store i32 0, i32* %p, !tbaa !2
   ret void

diff --git a/llvm/test/Feature/OperandBundles/pr26510.ll b/llvm/test/Feature/OperandBundles/pr26510.ll
index 1877f35a4006787e6b58b7b9297325681d6dc2e9..878628627b056d5ff1f8e8b845ae1fbf60e1aa55 100644 (file)
--- a/llvm/test/Feature/OperandBundles/pr26510.ll
+++ b/llvm/test/Feature/OperandBundles/pr26510.ll
@@ -10,7 +10,7 @@
 
 declare void @foo() readnone
 
-; CHECK-LABEL: define i8* @test(i8* %p)
+; CHECK-LABEL: define i8* @test(i8* writeonly %p)
 ; CHECK:   %a = alloca i8*, align 8
 ; CHECK:   store i8* %p, i8** %a, align 8
 ; CHECK:   call void @foo() [ "abc"(i8** %a) ]

diff --git a/llvm/test/Transforms/Coroutines/coro-async.ll b/llvm/test/Transforms/Coroutines/coro-async.ll
index 151573fe229259e1057f6f4953497111ed839199..86d95c336654905887af2e03c034c8f8b9aab999 100644 (file)
--- a/llvm/test/Transforms/Coroutines/coro-async.ll
+++ b/llvm/test/Transforms/Coroutines/coro-async.ll
@@ -256,7 +256,7 @@ entry:
   unreachable
 }
 
-; CHECK-LABEL: define swiftcc void @my_async_function2(%async.task* %task, %async.actor* %actor, i8* %async.ctxt)
+; CHECK-LABEL: define swiftcc void @my_async_function2(%async.task* %task, %async.actor* %actor, i8* writeonly %async.ctxt)
 ; CHECK-SAME: #[[FRAMEPOINTER:[0-9]+]]
 ; CHECK-SAME: !dbg ![[SP3:[0-9]+]]
 ; CHECK: store i8* %async.ctxt,
@@ -269,7 +269,7 @@ entry:
 ; CHECK: tail call swiftcc void @asyncSuspend(i8* [[CALLEE_CTXT]], %async.task* %task, %async.actor* %actor)
 ; CHECK: ret void
 
-; CHECK-LABEL: define internal swiftcc void @my_async_function2.resume.0(i8* %0, i8* nocapture readnone %1, i8* nocapture readonly %2)
+; CHECK-LABEL: define internal swiftcc void @my_async_function2.resume.0(i8* writeonly %0, i8* nocapture readnone %1, i8* nocapture readonly %2)
 ; CHECK-SAME: #[[FRAMEPOINTER]]
 ; CHECK-SAME: !dbg ![[SP4:[0-9]+]]
 ; CHECK: [[CALLEE_CTXT_ADDR:%.*]] = bitcast i8* %2 to i8**

diff --git a/llvm/test/Transforms/FunctionAttrs/2009-01-02-LocalStores.ll b/llvm/test/Transforms/FunctionAttrs/2009-01-02-LocalStores.ll
index 435f7810fbde6c5a3537e2cc31bff97fae388333..e493787f93b3337dc054ce5093786228935b1225 100644 (file)
--- a/llvm/test/Transforms/FunctionAttrs/2009-01-02-LocalStores.ll
+++ b/llvm/test/Transforms/FunctionAttrs/2009-01-02-LocalStores.ll
@@ -7,7 +7,7 @@ define i32* @a(i32** %p) {
        ret i32* %tmp
 }
 
-; CHECK: define i32* @b(i32* %q)
+; CHECK: define i32* @b(i32* writeonly %q)
 define i32* @b(i32 *%q) {
        %mem = alloca i32*
        store i32* %q, i32** %mem

diff --git a/llvm/test/Transforms/FunctionAttrs/nocapture.ll b/llvm/test/Transforms/FunctionAttrs/nocapture.ll
index 370a74d4696001d85ce5e0c2dff137fff548e382..5fc33e5c8c1922fd95f43c538dee8418be576515 100644 (file)
--- a/llvm/test/Transforms/FunctionAttrs/nocapture.ll
+++ b/llvm/test/Transforms/FunctionAttrs/nocapture.ll
@@ -8,7 +8,7 @@ define i32* @c1(i32* %q) {
        ret i32* %q
 }
 
-; FNATTR: define void @c2(i32* %q)
+; FNATTR: define void @c2(i32* writeonly %q)
 ; It would also be acceptable to mark %q as readnone. Update @c3 too.
 define void @c2(i32* %q) {
        store i32* %q, i32** @g
@@ -259,7 +259,7 @@ define void @captureLaunder(i8* %p) {
   ret void
 }
 
-; FNATTR: @nocaptureStrip(i8* nocapture %p)
+; FNATTR: @nocaptureStrip(i8* nocapture writeonly %p)
 define void @nocaptureStrip(i8* %p) {
 entry:
   %b = call i8* @llvm.strip.invariant.group.p0i8(i8* %p)
@@ -268,7 +268,7 @@ entry:
 }
 
 @g3 = global i8* null
-; FNATTR: define void @captureStrip(i8* %p)
+; FNATTR: define void @captureStrip(i8* writeonly %p)
 define void @captureStrip(i8* %p) {
   %b = call i8* @llvm.strip.invariant.group.p0i8(i8* %p)
   store i8* %b, i8** @g3

diff --git a/llvm/test/Transforms/FunctionAttrs/readattrs.ll b/llvm/test/Transforms/FunctionAttrs/readattrs.ll
index 6585ee4d24482566adcafd2d1d6b5149902af154..e300b85cf9207bad7cd74dac60a5bc7b793cebe1 100644 (file)
--- a/llvm/test/Transforms/FunctionAttrs/readattrs.ll
+++ b/llvm/test/Transforms/FunctionAttrs/readattrs.ll
@@ -34,7 +34,7 @@ define void @test4_2(i8* %p) {
   ret void
 }
 
-; CHECK: define void @test5(i8** nocapture %p, i8* %q)
+; CHECK: define void @test5(i8** nocapture writeonly %p, i8* writeonly %q)
 ; Missed optz'n: we could make %q readnone, but don't break test6!
 define void @test5(i8** %p, i8* %q) {
   store i8* %q, i8** %p
@@ -42,7 +42,7 @@ define void @test5(i8** %p, i8* %q) {
 }
 
 declare void @test6_1()
-; CHECK: define void @test6_2(i8** nocapture %p, i8* %q)
+; CHECK: define void @test6_2(i8** nocapture writeonly %p, i8* writeonly %q)
 ; This is not a missed optz'n.
 define void @test6_2(i8** %p, i8* %q) {
   store i8* %q, i8** %p
@@ -68,7 +68,7 @@ entry:
   ret i32* %p
 }
 
-; CHECK: define void @test8_2(i32* %p)
+; CHECK: define void @test8_2(i32* writeonly %p)
 define void @test8_2(i32* %p) {
 entry:
   %call = call i32* @test8_1(i32* %p)

diff --git a/llvm/test/Transforms/FunctionAttrs/writeonly.ll b/llvm/test/Transforms/FunctionAttrs/writeonly.ll
index f3431ae42cb4cd105a6d8395bd164655baef5621..ac3b5ffcca51b36374883883169e2c72ee66c2a6 100644 (file)
--- a/llvm/test/Transforms/FunctionAttrs/writeonly.ll
+++ b/llvm/test/Transforms/FunctionAttrs/writeonly.ll
@@ -25,13 +25,13 @@ nouses-argworn-funwo_entry:
   ret void
 }
 
-; CHECK: define void @test_store(i8* nocapture %p)
+; CHECK: define void @test_store(i8* nocapture writeonly %p)
 define void @test_store(i8* %p) {
   store i8 0, i8* %p
   ret void
 }
 
-; CHECK: define void @test_addressing(i8* nocapture %p)
+; CHECK: define void @test_addressing(i8* nocapture writeonly %p)
 define void @test_addressing(i8* %p) {
   %gep = getelementptr i8, i8* %p, i64 8
   %bitcast = bitcast i8* %gep to i32*