summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorMatt Arsenault <Matthew.Arsenault@amd.com>2016-08-24 17:52:04 +0000
committerMatt Arsenault <Matthew.Arsenault@amd.com>2016-08-24 17:52:04 +0000
commit26d9c41ff6b863253e1e9b17bd5acffc30ead692 (patch)
tree565605f3bd27ad9829fdce42c49347719102bb18
parent769ec705bc95e58ab0812d8d0d44f494d2cd0b4b (diff)
downloadbcm5719-llvm-26d9c41ff6b863253e1e9b17bd5acffc30ead692.tar.gz
bcm5719-llvm-26d9c41ff6b863253e1e9b17bd5acffc30ead692.zip
amdgcn: Fix return type for get_global_size
llvm-svn: 279644
-rw-r--r--libclc/amdgcn/lib/SOURCES1
-rw-r--r--libclc/amdgcn/lib/workitem/get_global_size.ll21
-rw-r--r--libclc/amdgpu/lib/SOURCES1
-rw-r--r--libclc/r600/lib/SOURCES1
-rw-r--r--libclc/r600/lib/workitem/get_global_size.ll (renamed from libclc/amdgpu/lib/workitem/get_global_size.ll)2
5 files changed, 24 insertions, 2 deletions
diff --git a/libclc/amdgcn/lib/SOURCES b/libclc/amdgcn/lib/SOURCES
index 06ed228172b..f7e2671c1c9 100644
--- a/libclc/amdgcn/lib/SOURCES
+++ b/libclc/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/libclc/amdgcn/lib/workitem/get_global_size.ll b/libclc/amdgcn/lib/workitem/get_global_size.ll
new file mode 100644
index 00000000000..a5162557320
--- /dev/null
+++ b/libclc/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/libclc/amdgpu/lib/SOURCES b/libclc/amdgpu/lib/SOURCES
index efee251ed07..fac2fc3e73a 100644
--- a/libclc/amdgpu/lib/SOURCES
+++ b/libclc/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/libclc/r600/lib/SOURCES b/libclc/r600/lib/SOURCES
index 23da5028b30..d55ff30cbc0 100644
--- a/libclc/r600/lib/SOURCES
+++ b/libclc/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/libclc/amdgpu/lib/workitem/get_global_size.ll b/libclc/r600/lib/workitem/get_global_size.ll
index ac2d08d8ee1..d1c769cbc7e 100644
--- a/libclc/amdgpu/lib/workitem/get_global_size.ll
+++ b/libclc/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
}
OpenPOWER on IntegriCloud