1 // REQUIRES: x86-registered-target 2 // REQUIRES: nvptx-registered-target 3 4 // RUN: %clang_cc1 -std=c++11 -fcuda-is-device -triple nvptx64-nvidia-cuda -emit-llvm -o - %s | FileCheck --check-prefix=DEVICE %s 5 // RUN: echo "GPU binary would be here" > %t 6 // RUN: %clang_cc1 -std=c++11 -triple x86_64-unknown-linux-gnu -target-sdk-version=8.0 -fcuda-include-gpubinary %t -emit-llvm -o - %s | FileCheck --check-prefix=HOST %s 7 8 struct textureReference { 9 int desc; 10 }; 11 12 enum ReadMode { 13 ElementType = 0, 14 NormalizedFloat = 1 15 }; 16 17 template <typename T, int dim = 1, enum ReadMode mode = ElementType> 18 struct __attribute__((device_builtin_texture_type)) texture : public textureReference { 19 }; 20 21 // On the device side, texture references are represented as `i64` handles. 22 // DEVICE: @tex ={{.*}} addrspace(1) externally_initialized global i64 undef, align 4 23 // DEVICE: @norm ={{.*}} addrspace(1) externally_initialized global i64 undef, align 4 24 // On the host side, they remain in the original type. 25 // HOST: @tex = internal global %struct.texture 26 // HOST: @norm = internal global %struct.texture 27 // HOST: @0 = private unnamed_addr constant [4 x i8] c"tex\00" 28 // HOST: @1 = private unnamed_addr constant [5 x i8] c"norm\00" 29 texture<float, 2, ElementType> tex; 30 texture<float, 2, NormalizedFloat> norm; 31 32 struct v4f { 33 float x, y, z, w; 34 }; 35 36 __attribute__((device)) v4f tex2d_ld(texture<float, 2, ElementType>, float, float) asm("llvm.nvvm.tex.unified.2d.v4f32.f32"); 37 __attribute__((device)) v4f tex2d_ld(texture<float, 2, NormalizedFloat>, int, int) asm("llvm.nvvm.tex.unified.2d.v4f32.s32"); 38 39 // DEVICE-LABEL: float @_Z3fooff(float noundef %x, float noundef %y) 40 // DEVICE: call i64 @llvm.nvvm.texsurf.handle.internal.p1i64(i64 addrspace(1)* @tex) 41 // DEVICE: call %struct.v4f @llvm.nvvm.tex.unified.2d.v4f32.f32(i64 %{{.*}}, float noundef %{{.*}}, float noundef %{{.*}}) 42 // DEVICE: call i64 @llvm.nvvm.texsurf.handle.internal.p1i64(i64 addrspace(1)* @norm) 43 // DEVICE: call %struct.v4f @llvm.nvvm.tex.unified.2d.v4f32.s32(i64 %{{.*}}, i32 noundef %{{.*}}, i32 noundef %{{.*}}) 44 __attribute__((device)) float foo(float x, float y) { 45 return tex2d_ld(tex, x, y).x + tex2d_ld(norm, int(x), int(y)).x; 46 } 47 48 // HOST: define internal void @[[PREFIX:__cuda]]_register_globals 49 // Texture references need registering with correct arguments. 50 // HOST: call void @[[PREFIX]]RegisterTexture(i8** %0, i8*{{.*}}({{.*}}@tex{{.*}}), i8*{{.*}}({{.*}}@0{{.*}}), i8*{{.*}}({{.*}}@0{{.*}}), i32 2, i32 0, i32 0) 51 // HOST: call void @[[PREFIX]]RegisterTexture(i8** %0, i8*{{.*}}({{.*}}@norm{{.*}}), i8*{{.*}}({{.*}}@1{{.*}}), i8*{{.*}}({{.*}}@1{{.*}}), i32 2, i32 1, i32 0) 52 53 // They also need annotating in metadata. 54 // DEVICE: !0 = !{i64 addrspace(1)* @tex, !"texture", i32 1} 55 // DEVICE: !1 = !{i64 addrspace(1)* @norm, !"texture", i32 1} 56