PTX: Emit global arrays with proper sizes

- Emit all arrays as type .b8 and proper sizes in bytes to conform
  to the output of nvcc

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@127584 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/PTX/PTXAsmPrinter.cpp b/lib/Target/PTX/PTXAsmPrinter.cpp
index 087d4f98c4869f8d518f44b09b1f397fa310d7a5..0b07f749b3e4eb90017ccbbb81395f960786174c 100644 (file)
--- a/lib/Target/PTX/PTXAsmPrinter.cpp
+++ b/lib/Target/PTX/PTXAsmPrinter.cpp
@@ -316,13 +316,42 @@ void PTXAsmPrinter::EmitVariableDeclaration(const GlobalVariable *gv) {
     decl += " ";
   }
 
-  decl += getTypeName(gv->getType());
-  decl += " ";
 
-  decl += gvsym->getName();
+  if (PointerType::classof(gv->getType())) {
+    const PointerType* pointerTy = dyn_cast<const PointerType>(gv->getType());
+    const Type* elementTy = pointerTy->getElementType();
+
+    assert(elementTy->isArrayTy() && "Only pointers to arrays are supported");
+
+    const ArrayType* arrayTy = dyn_cast<const ArrayType>(elementTy);
+    elementTy = arrayTy->getElementType();
+
+    // FIXME: isPrimitiveType() == false for i16?
+    assert(elementTy->isSingleValueType() &&
+           "Non-primitive types are not handled");
+
+    // Compute the size of the array, in bytes.
+    uint64_t arraySize = (elementTy->getPrimitiveSizeInBits() >> 3)
+                         * arrayTy->getNumElements();
 
-  if (ArrayType::classof(gv->getType()) || PointerType::classof(gv->getType()))
-    decl += "[]";
+    decl += ".b8 ";
+    decl += gvsym->getName();
+    decl += "[";
+    decl += utostr(arraySize);
+    decl += "]";
+  }
+  else {
+    // Note: this is currently the fall-through case and most likely generates
+    //       incorrect code.
+    decl += getTypeName(gv->getType());
+    decl += " ";
+
+    decl += gvsym->getName();
+
+    if (ArrayType::classof(gv->getType()) ||
+        PointerType::classof(gv->getType()))
+      decl += "[]";
+  }
 
   decl += ";";
 

diff --git a/test/CodeGen/PTX/ld.ll b/test/CodeGen/PTX/ld.ll
index e7cc92e3c9972965fac8230927afdec2b3a11aa0..6d120dcf7af8f32f1b5d98285a5aca588ab76ab4 100644 (file)
--- a/test/CodeGen/PTX/ld.ll
+++ b/test/CodeGen/PTX/ld.ll
@@ -1,63 +1,63 @@
 ; RUN: llc < %s -march=ptx | FileCheck %s
 
-;CHECK: .extern .global .u16 array_i16[];
+;CHECK: .extern .global .b8 array_i16[20];
 @array_i16 = external global [10 x i16]
 
-;CHECK: .extern .const .u16 array_constant_i16[];
+;CHECK: .extern .const .b8 array_constant_i16[20];
 @array_constant_i16 = external addrspace(1) constant [10 x i16]
 
-;CHECK: .extern .local .u16 array_local_i16[];
+;CHECK: .extern .local .b8 array_local_i16[20];
 @array_local_i16 = external addrspace(2) global [10 x i16]
 
-;CHECK: .extern .shared .u16 array_shared_i16[];
+;CHECK: .extern .shared .b8 array_shared_i16[20];
 @array_shared_i16 = external addrspace(4) global [10 x i16]
 
-;CHECK: .extern .global .u32 array_i32[];
+;CHECK: .extern .global .b8 array_i32[40];
 @array_i32 = external global [10 x i32]
 
-;CHECK: .extern .const .u32 array_constant_i32[];
+;CHECK: .extern .const .b8 array_constant_i32[40];
 @array_constant_i32 = external addrspace(1) constant [10 x i32]
 
-;CHECK: .extern .local .u32 array_local_i32[];
+;CHECK: .extern .local .b8 array_local_i32[40];
 @array_local_i32 = external addrspace(2) global [10 x i32]
 
-;CHECK: .extern .shared .u32 array_shared_i32[];
+;CHECK: .extern .shared .b8 array_shared_i32[40];
 @array_shared_i32 = external addrspace(4) global [10 x i32]
 
-;CHECK: .extern .global .u64 array_i64[];
+;CHECK: .extern .global .b8 array_i64[80];
 @array_i64 = external global [10 x i64]
 
-;CHECK: .extern .const .u64 array_constant_i64[];
+;CHECK: .extern .const .b8 array_constant_i64[80];
 @array_constant_i64 = external addrspace(1) constant [10 x i64]
 
-;CHECK: .extern .local .u64 array_local_i64[];
+;CHECK: .extern .local .b8 array_local_i64[80];
 @array_local_i64 = external addrspace(2) global [10 x i64]
 
-;CHECK: .extern .shared .u64 array_shared_i64[];
+;CHECK: .extern .shared .b8 array_shared_i64[80];
 @array_shared_i64 = external addrspace(4) global [10 x i64]
 
-;CHECK: .extern .global .f32 array_float[];
+;CHECK: .extern .global .b8 array_float[40];
 @array_float = external global [10 x float]
 
-;CHECK: .extern .const .f32 array_constant_float[];
+;CHECK: .extern .const .b8 array_constant_float[40];
 @array_constant_float = external addrspace(1) constant [10 x float]
 
-;CHECK: .extern .local .f32 array_local_float[];
+;CHECK: .extern .local .b8 array_local_float[40];
 @array_local_float = external addrspace(2) global [10 x float]
 
-;CHECK: .extern .shared .f32 array_shared_float[];
+;CHECK: .extern .shared .b8 array_shared_float[40];
 @array_shared_float = external addrspace(4) global [10 x float]
 
-;CHECK: .extern .global .f64 array_double[];
+;CHECK: .extern .global .b8 array_double[80];
 @array_double = external global [10 x double]
 
-;CHECK: .extern .const .f64 array_constant_double[];
+;CHECK: .extern .const .b8 array_constant_double[80];
 @array_constant_double = external addrspace(1) constant [10 x double]
 
-;CHECK: .extern .local .f64 array_local_double[];
+;CHECK: .extern .local .b8 array_local_double[80];
 @array_local_double = external addrspace(2) global [10 x double]
 
-;CHECK: .extern .shared .f64 array_shared_double[];
+;CHECK: .extern .shared .b8 array_shared_double[80];
 @array_shared_double = external addrspace(4) global [10 x double]
 
 

diff --git a/test/CodeGen/PTX/st.ll b/test/CodeGen/PTX/st.ll
index bbe89a106486b605d454e0bc5df72587dc1a502f..2b1f6e79b6ca7370195944bedaa4247a2900eb37 100644 (file)
--- a/test/CodeGen/PTX/st.ll
+++ b/test/CodeGen/PTX/st.ll
@@ -1,63 +1,63 @@
 ; RUN: llc < %s -march=ptx | FileCheck %s
 
-;CHECK: .extern .global .u16 array_i16[];
+;CHECK: .extern .global .b8 array_i16[20];
 @array_i16 = external global [10 x i16]
 
-;CHECK: .extern .const .u16 array_constant_i16[];
+;CHECK: .extern .const .b8 array_constant_i16[20];
 @array_constant_i16 = external addrspace(1) constant [10 x i16]
 
-;CHECK: .extern .local .u16 array_local_i16[];
+;CHECK: .extern .local .b8 array_local_i16[20];
 @array_local_i16 = external addrspace(2) global [10 x i16]
 
-;CHECK: .extern .shared .u16 array_shared_i16[];
+;CHECK: .extern .shared .b8 array_shared_i16[20];
 @array_shared_i16 = external addrspace(4) global [10 x i16]
 
-;CHECK: .extern .global .u32 array_i32[];
+;CHECK: .extern .global .b8 array_i32[40];
 @array_i32 = external global [10 x i32]
 
-;CHECK: .extern .const .u32 array_constant_i32[];
+;CHECK: .extern .const .b8 array_constant_i32[40];
 @array_constant_i32 = external addrspace(1) constant [10 x i32]
 
-;CHECK: .extern .local .u32 array_local_i32[];
+;CHECK: .extern .local .b8 array_local_i32[40];
 @array_local_i32 = external addrspace(2) global [10 x i32]
 
-;CHECK: .extern .shared .u32 array_shared_i32[];
+;CHECK: .extern .shared .b8 array_shared_i32[40];
 @array_shared_i32 = external addrspace(4) global [10 x i32]
 
-;CHECK: .extern .global .u64 array_i64[];
+;CHECK: .extern .global .b8 array_i64[80];
 @array_i64 = external global [10 x i64]
 
-;CHECK: .extern .const .u64 array_constant_i64[];
+;CHECK: .extern .const .b8 array_constant_i64[80];
 @array_constant_i64 = external addrspace(1) constant [10 x i64]
 
-;CHECK: .extern .local .u64 array_local_i64[];
+;CHECK: .extern .local .b8 array_local_i64[80];
 @array_local_i64 = external addrspace(2) global [10 x i64]
 
-;CHECK: .extern .shared .u64 array_shared_i64[];
+;CHECK: .extern .shared .b8 array_shared_i64[80];
 @array_shared_i64 = external addrspace(4) global [10 x i64]
 
-;CHECK: .extern .global .f32 array_float[];
+;CHECK: .extern .global .b8 array_float[40];
 @array_float = external global [10 x float]
 
-;CHECK: .extern .const .f32 array_constant_float[];
+;CHECK: .extern .const .b8 array_constant_float[40];
 @array_constant_float = external addrspace(1) constant [10 x float]
 
-;CHECK: .extern .local .f32 array_local_float[];
+;CHECK: .extern .local .b8 array_local_float[40];
 @array_local_float = external addrspace(2) global [10 x float]
 
-;CHECK: .extern .shared .f32 array_shared_float[];
+;CHECK: .extern .shared .b8 array_shared_float[40];
 @array_shared_float = external addrspace(4) global [10 x float]
 
-;CHECK: .extern .global .f64 array_double[];
+;CHECK: .extern .global .b8 array_double[80];
 @array_double = external global [10 x double]
 
-;CHECK: .extern .const .f64 array_constant_double[];
+;CHECK: .extern .const .b8 array_constant_double[80];
 @array_constant_double = external addrspace(1) constant [10 x double]
 
-;CHECK: .extern .local .f64 array_local_double[];
+;CHECK: .extern .local .b8 array_local_double[80];
 @array_local_double = external addrspace(2) global [10 x double]
 
-;CHECK: .extern .shared .f64 array_shared_double[];
+;CHECK: .extern .shared .b8 array_shared_double[80];
 @array_shared_double = external addrspace(4) global [10 x double]