diff options
author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-08-24 17:52:04 +0000 |
---|---|---|
committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-08-24 17:52:04 +0000 |
commit | 0f2ddaf8df0094d51602a2509fbc40af90112c90 (patch) | |
tree | 936eda1e3cb605aaf6df5046d2becaadab313ff6 | |
parent | 3aa4eea26b7314fda84ed0612ef02986fe5e0846 (diff) |
amdgcn: Fix return type for get_global_size
git-svn-id: https://llvm.org/svn/llvm-project/libclc/trunk@279644 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r-- | amdgcn/lib/SOURCES | 1 | ||||
-rw-r--r-- | amdgcn/lib/workitem/get_global_size.ll | 21 | ||||
-rw-r--r-- | amdgpu/lib/SOURCES | 1 | ||||
-rw-r--r-- | r600/lib/SOURCES | 1 | ||||
-rw-r--r-- | r600/lib/workitem/get_global_size.ll (renamed from amdgpu/lib/workitem/get_global_size.ll) | 2 |
5 files changed, 24 insertions, 2 deletions
diff --git a/amdgcn/lib/SOURCES b/amdgcn/lib/SOURCES index 06ed228..f7e2671 100644 --- a/amdgcn/lib/SOURCES +++ b/amdgcn/lib/SOURCES @@ -2,6 +2,7 @@ math/ldexp.cl synchronization/barrier_impl.ll workitem/get_global_offset.cl workitem/get_group_id.cl +workitem/get_global_size.ll workitem/get_local_id.cl workitem/get_local_size.ll workitem/get_work_dim.cl diff --git a/amdgcn/lib/workitem/get_global_size.ll b/amdgcn/lib/workitem/get_global_size.ll new file mode 100644 index 0000000..a516255 --- /dev/null +++ b/amdgcn/lib/workitem/get_global_size.ll @@ -0,0 +1,21 @@ +declare i32 @llvm.r600.read.global.size.x() nounwind readnone +declare i32 @llvm.r600.read.global.size.y() nounwind readnone +declare i32 @llvm.r600.read.global.size.z() nounwind readnone + +define i64 @get_global_size(i32 %dim) nounwind readnone alwaysinline { + switch i32 %dim, label %default [i32 0, label %x_dim i32 1, label %y_dim i32 2, label %z_dim] +x_dim: + %x = call i32 @llvm.r600.read.global.size.x() + %x.ext = zext i32 %x to i64 + ret i64 %x.ext +y_dim: + %y = call i32 @llvm.r600.read.global.size.y() + %y.ext = zext i32 %y to i64 + ret i64 %y.ext +z_dim: + %z = call i32 @llvm.r600.read.global.size.z() + %z.ext = zext i32 %z to i64 + ret i64 %z.ext +default: + ret i64 1 +} diff --git a/amdgpu/lib/SOURCES b/amdgpu/lib/SOURCES index efee251..fac2fc3 100644 --- a/amdgpu/lib/SOURCES +++ b/amdgpu/lib/SOURCES @@ -17,4 +17,3 @@ image/write_imagei.cl image/write_imageui.cl image/write_image_impl.ll workitem/get_num_groups.ll -workitem/get_global_size.ll diff --git a/r600/lib/SOURCES b/r600/lib/SOURCES index 23da502..d55ff30 100644 --- a/r600/lib/SOURCES +++ b/r600/lib/SOURCES @@ -1,6 +1,7 @@ synchronization/barrier_impl.ll workitem/get_global_offset.cl workitem/get_group_id.cl +workitem/get_global_size.ll workitem/get_local_id.cl workitem/get_local_size.ll workitem/get_work_dim.cl diff --git a/amdgpu/lib/workitem/get_global_size.ll b/r600/lib/workitem/get_global_size.ll index ac2d08d..d1c769c 100644 --- a/amdgpu/lib/workitem/get_global_size.ll +++ b/r600/lib/workitem/get_global_size.ll @@ -14,5 +14,5 @@ z_dim: %z = call i32 @llvm.r600.read.global.size.z() nounwind readnone ret i32 %z default: - ret i32 0 + ret i32 1 } |