[libclc] 18/58: Split sources for amdgcn and r600
Andreas Boll
aboll-guest at moszumanska.debian.org
Thu Oct 6 08:16:29 UTC 2016
This is an automated email from the git hooks/post-receive script.
aboll-guest pushed a commit to branch master
in repository libclc.
commit ba011288a9659d57736956c47b8522eace520807
Author: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Sat Feb 13 01:01:59 2016 +0000
Split sources for amdgcn and r600
Most files remain in a common amdgpu directory.
Also switches barriers to to use convergent,
and use llvm.amdgcn.s.barrier.
This now requires 3.9/trunk to build amdgcn.
git-svn-id: https://llvm.org/svn/llvm-project/libclc/trunk@260777 91177308-0d34-0410-b5e6-96231b3b80d8
---
amdgcn/lib/OVERRIDES | 0
amdgcn/lib/SOURCES | 1 +
.../lib/synchronization/barrier_impl.ll | 17 +++++++++------
{r600 => amdgpu}/lib/OVERRIDES | 0
{r600 => amdgpu}/lib/SOURCES | 1 -
{r600 => amdgpu}/lib/atomic/atomic.cl | 0
.../lib/image/get_image_attributes_impl.ll | 0
.../lib/image/get_image_channel_data_type.cl | 0
.../lib/image/get_image_channel_order.cl | 0
{r600 => amdgpu}/lib/image/get_image_depth.cl | 0
{r600 => amdgpu}/lib/image/get_image_height.cl | 0
{r600 => amdgpu}/lib/image/get_image_width.cl | 0
{r600 => amdgpu}/lib/image/read_image_impl.ll | 0
{r600 => amdgpu}/lib/image/read_imagef.cl | 0
{r600 => amdgpu}/lib/image/read_imagei.cl | 0
{r600 => amdgpu}/lib/image/read_imageui.cl | 0
{r600 => amdgpu}/lib/image/write_image_impl.ll | 0
{r600 => amdgpu}/lib/image/write_imagef.cl | 0
{r600 => amdgpu}/lib/image/write_imagei.cl | 0
{r600 => amdgpu}/lib/image/write_imageui.cl | 0
{r600 => amdgpu}/lib/math/ldexp.cl | 0
{r600 => amdgpu}/lib/math/nextafter.cl | 0
{r600 => amdgpu}/lib/math/sqrt.cl | 0
{r600 => amdgpu}/lib/synchronization/barrier.cl | 0
{r600 => amdgpu}/lib/workitem/get_global_size.ll | 0
{r600 => amdgpu}/lib/workitem/get_group_id.ll | 0
{r600 => amdgpu}/lib/workitem/get_local_id.ll | 0
{r600 => amdgpu}/lib/workitem/get_local_size.ll | 0
{r600 => amdgpu}/lib/workitem/get_num_groups.ll | 0
{r600 => amdgpu}/lib/workitem/get_work_dim.ll | 0
configure.py | 8 +++----
r600/lib/OVERRIDES | 2 --
r600/lib/SOURCES | 25 ----------------------
r600/lib/synchronization/barrier_impl.ll | 18 ++++++++++------
34 files changed, 26 insertions(+), 46 deletions(-)
diff --git a/amdgcn/lib/OVERRIDES b/amdgcn/lib/OVERRIDES
new file mode 100644
index 0000000..e69de29
diff --git a/amdgcn/lib/SOURCES b/amdgcn/lib/SOURCES
new file mode 100644
index 0000000..c99f3fc
--- /dev/null
+++ b/amdgcn/lib/SOURCES
@@ -0,0 +1 @@
+synchronization/barrier_impl.ll
diff --git a/r600/lib/synchronization/barrier_impl.ll b/amdgcn/lib/synchronization/barrier_impl.ll
similarity index 54%
copy from r600/lib/synchronization/barrier_impl.ll
copy to amdgcn/lib/synchronization/barrier_impl.ll
index 3d8ee66..1809edd 100644
--- a/r600/lib/synchronization/barrier_impl.ll
+++ b/amdgcn/lib/synchronization/barrier_impl.ll
@@ -1,9 +1,8 @@
-declare i32 @__clc_clk_local_mem_fence() nounwind alwaysinline
-declare i32 @__clc_clk_global_mem_fence() nounwind alwaysinline
-declare void @llvm.AMDGPU.barrier.local() nounwind noduplicate
-declare void @llvm.AMDGPU.barrier.global() nounwind noduplicate
+declare i32 @__clc_clk_local_mem_fence() #1
+declare i32 @__clc_clk_global_mem_fence() #1
+declare void @llvm.amdgcn.s.barrier() #0
-define void @barrier(i32 %flags) nounwind noduplicate alwaysinline {
+define void @barrier(i32 %flags) #2 {
barrier_local_test:
%CLK_LOCAL_MEM_FENCE = call i32 @__clc_clk_local_mem_fence()
%0 = and i32 %flags, %CLK_LOCAL_MEM_FENCE
@@ -11,7 +10,7 @@ barrier_local_test:
br i1 %1, label %barrier_local, label %barrier_global_test
barrier_local:
- call void @llvm.AMDGPU.barrier.local() noduplicate
+ call void @llvm.amdgcn.s.barrier()
br label %barrier_global_test
barrier_global_test:
@@ -21,9 +20,13 @@ barrier_global_test:
br i1 %3, label %barrier_global, label %done
barrier_global:
- call void @llvm.AMDGPU.barrier.global() noduplicate
+ call void @llvm.amdgcn.s.barrier()
br label %done
done:
ret void
}
+
+attributes #0 = { nounwind convergent }
+attributes #1 = { nounwind alwaysinline }
+attributes #2 = { nounwind convergent alwaysinline }
diff --git a/r600/lib/OVERRIDES b/amdgpu/lib/OVERRIDES
similarity index 100%
copy from r600/lib/OVERRIDES
copy to amdgpu/lib/OVERRIDES
diff --git a/r600/lib/SOURCES b/amdgpu/lib/SOURCES
similarity index 95%
copy from r600/lib/SOURCES
copy to amdgpu/lib/SOURCES
index 029b22c..7505f3f 100644
--- a/r600/lib/SOURCES
+++ b/amdgpu/lib/SOURCES
@@ -9,7 +9,6 @@ workitem/get_local_id.ll
workitem/get_global_size.ll
workitem/get_work_dim.ll
synchronization/barrier.cl
-synchronization/barrier_impl.ll
image/get_image_width.cl
image/get_image_height.cl
image/get_image_depth.cl
diff --git a/r600/lib/atomic/atomic.cl b/amdgpu/lib/atomic/atomic.cl
similarity index 100%
rename from r600/lib/atomic/atomic.cl
rename to amdgpu/lib/atomic/atomic.cl
diff --git a/r600/lib/image/get_image_attributes_impl.ll b/amdgpu/lib/image/get_image_attributes_impl.ll
similarity index 100%
rename from r600/lib/image/get_image_attributes_impl.ll
rename to amdgpu/lib/image/get_image_attributes_impl.ll
diff --git a/r600/lib/image/get_image_channel_data_type.cl b/amdgpu/lib/image/get_image_channel_data_type.cl
similarity index 100%
rename from r600/lib/image/get_image_channel_data_type.cl
rename to amdgpu/lib/image/get_image_channel_data_type.cl
diff --git a/r600/lib/image/get_image_channel_order.cl b/amdgpu/lib/image/get_image_channel_order.cl
similarity index 100%
rename from r600/lib/image/get_image_channel_order.cl
rename to amdgpu/lib/image/get_image_channel_order.cl
diff --git a/r600/lib/image/get_image_depth.cl b/amdgpu/lib/image/get_image_depth.cl
similarity index 100%
rename from r600/lib/image/get_image_depth.cl
rename to amdgpu/lib/image/get_image_depth.cl
diff --git a/r600/lib/image/get_image_height.cl b/amdgpu/lib/image/get_image_height.cl
similarity index 100%
rename from r600/lib/image/get_image_height.cl
rename to amdgpu/lib/image/get_image_height.cl
diff --git a/r600/lib/image/get_image_width.cl b/amdgpu/lib/image/get_image_width.cl
similarity index 100%
rename from r600/lib/image/get_image_width.cl
rename to amdgpu/lib/image/get_image_width.cl
diff --git a/r600/lib/image/read_image_impl.ll b/amdgpu/lib/image/read_image_impl.ll
similarity index 100%
rename from r600/lib/image/read_image_impl.ll
rename to amdgpu/lib/image/read_image_impl.ll
diff --git a/r600/lib/image/read_imagef.cl b/amdgpu/lib/image/read_imagef.cl
similarity index 100%
rename from r600/lib/image/read_imagef.cl
rename to amdgpu/lib/image/read_imagef.cl
diff --git a/r600/lib/image/read_imagei.cl b/amdgpu/lib/image/read_imagei.cl
similarity index 100%
rename from r600/lib/image/read_imagei.cl
rename to amdgpu/lib/image/read_imagei.cl
diff --git a/r600/lib/image/read_imageui.cl b/amdgpu/lib/image/read_imageui.cl
similarity index 100%
rename from r600/lib/image/read_imageui.cl
rename to amdgpu/lib/image/read_imageui.cl
diff --git a/r600/lib/image/write_image_impl.ll b/amdgpu/lib/image/write_image_impl.ll
similarity index 100%
rename from r600/lib/image/write_image_impl.ll
rename to amdgpu/lib/image/write_image_impl.ll
diff --git a/r600/lib/image/write_imagef.cl b/amdgpu/lib/image/write_imagef.cl
similarity index 100%
rename from r600/lib/image/write_imagef.cl
rename to amdgpu/lib/image/write_imagef.cl
diff --git a/r600/lib/image/write_imagei.cl b/amdgpu/lib/image/write_imagei.cl
similarity index 100%
rename from r600/lib/image/write_imagei.cl
rename to amdgpu/lib/image/write_imagei.cl
diff --git a/r600/lib/image/write_imageui.cl b/amdgpu/lib/image/write_imageui.cl
similarity index 100%
rename from r600/lib/image/write_imageui.cl
rename to amdgpu/lib/image/write_imageui.cl
diff --git a/r600/lib/math/ldexp.cl b/amdgpu/lib/math/ldexp.cl
similarity index 100%
rename from r600/lib/math/ldexp.cl
rename to amdgpu/lib/math/ldexp.cl
diff --git a/r600/lib/math/nextafter.cl b/amdgpu/lib/math/nextafter.cl
similarity index 100%
rename from r600/lib/math/nextafter.cl
rename to amdgpu/lib/math/nextafter.cl
diff --git a/r600/lib/math/sqrt.cl b/amdgpu/lib/math/sqrt.cl
similarity index 100%
rename from r600/lib/math/sqrt.cl
rename to amdgpu/lib/math/sqrt.cl
diff --git a/r600/lib/synchronization/barrier.cl b/amdgpu/lib/synchronization/barrier.cl
similarity index 100%
rename from r600/lib/synchronization/barrier.cl
rename to amdgpu/lib/synchronization/barrier.cl
diff --git a/r600/lib/workitem/get_global_size.ll b/amdgpu/lib/workitem/get_global_size.ll
similarity index 100%
rename from r600/lib/workitem/get_global_size.ll
rename to amdgpu/lib/workitem/get_global_size.ll
diff --git a/r600/lib/workitem/get_group_id.ll b/amdgpu/lib/workitem/get_group_id.ll
similarity index 100%
rename from r600/lib/workitem/get_group_id.ll
rename to amdgpu/lib/workitem/get_group_id.ll
diff --git a/r600/lib/workitem/get_local_id.ll b/amdgpu/lib/workitem/get_local_id.ll
similarity index 100%
rename from r600/lib/workitem/get_local_id.ll
rename to amdgpu/lib/workitem/get_local_id.ll
diff --git a/r600/lib/workitem/get_local_size.ll b/amdgpu/lib/workitem/get_local_size.ll
similarity index 100%
rename from r600/lib/workitem/get_local_size.ll
rename to amdgpu/lib/workitem/get_local_size.ll
diff --git a/r600/lib/workitem/get_num_groups.ll b/amdgpu/lib/workitem/get_num_groups.ll
similarity index 100%
rename from r600/lib/workitem/get_num_groups.ll
rename to amdgpu/lib/workitem/get_num_groups.ll
diff --git a/r600/lib/workitem/get_work_dim.ll b/amdgpu/lib/workitem/get_work_dim.ll
similarity index 100%
rename from r600/lib/workitem/get_work_dim.ll
rename to amdgpu/lib/workitem/get_work_dim.ll
diff --git a/configure.py b/configure.py
index d591ef8..2663212 100755
--- a/configure.py
+++ b/configure.py
@@ -69,8 +69,8 @@ llvm_version = string.split(string.replace(llvm_config(['--version']), 'svn', ''
llvm_int_version = int(llvm_version[0]) * 100 + int(llvm_version[1]) * 10
llvm_string_version = 'LLVM' + llvm_version[0] + '.' + llvm_version[1]
-if llvm_int_version < 370:
- print "libclc requires LLVM >= 3.7"
+if llvm_int_version < 390:
+ print "libclc requires LLVM >= 3.9"
sys.exit(1)
llvm_system_libs = llvm_config(['--system-libs'])
@@ -175,8 +175,8 @@ for target in targets:
subdirs.append("%s-%s-%s" % (arch, t_vendor, t_os))
subdirs.append("%s-%s" % (arch, t_os))
subdirs.append(arch)
- if arch == 'amdgcn':
- subdirs.append('r600')
+ if arch == 'amdgcn' or arch == 'r600':
+ subdirs.append('amdgpu')
incdirs = filter(os.path.isdir,
[os.path.join(srcdir, subdir, 'include') for subdir in subdirs])
diff --git a/r600/lib/OVERRIDES b/r600/lib/OVERRIDES
index 3f941d8..e69de29 100644
--- a/r600/lib/OVERRIDES
+++ b/r600/lib/OVERRIDES
@@ -1,2 +0,0 @@
-workitem/get_group_id.cl
-workitem/get_global_size.cl
diff --git a/r600/lib/SOURCES b/r600/lib/SOURCES
index 029b22c..c99f3fc 100644
--- a/r600/lib/SOURCES
+++ b/r600/lib/SOURCES
@@ -1,26 +1 @@
-atomic/atomic.cl
-math/ldexp.cl
-math/nextafter.cl
-math/sqrt.cl
-workitem/get_num_groups.ll
-workitem/get_group_id.ll
-workitem/get_local_size.ll
-workitem/get_local_id.ll
-workitem/get_global_size.ll
-workitem/get_work_dim.ll
-synchronization/barrier.cl
synchronization/barrier_impl.ll
-image/get_image_width.cl
-image/get_image_height.cl
-image/get_image_depth.cl
-image/get_image_channel_data_type.cl
-image/get_image_channel_order.cl
-image/get_image_attributes_impl.ll
-image/read_imagef.cl
-image/read_imagei.cl
-image/read_imageui.cl
-image/read_image_impl.ll
-image/write_imagef.cl
-image/write_imagei.cl
-image/write_imageui.cl
-image/write_image_impl.ll
diff --git a/r600/lib/synchronization/barrier_impl.ll b/r600/lib/synchronization/barrier_impl.ll
index 3d8ee66..825b2eb 100644
--- a/r600/lib/synchronization/barrier_impl.ll
+++ b/r600/lib/synchronization/barrier_impl.ll
@@ -1,9 +1,9 @@
-declare i32 @__clc_clk_local_mem_fence() nounwind alwaysinline
-declare i32 @__clc_clk_global_mem_fence() nounwind alwaysinline
-declare void @llvm.AMDGPU.barrier.local() nounwind noduplicate
-declare void @llvm.AMDGPU.barrier.global() nounwind noduplicate
+declare i32 @__clc_clk_local_mem_fence() #1
+declare i32 @__clc_clk_global_mem_fence() #1
+declare void @llvm.AMDGPU.barrier.local() #0
+declare void @llvm.AMDGPU.barrier.global() #0
-define void @barrier(i32 %flags) nounwind noduplicate alwaysinline {
+define void @barrier(i32 %flags) #2 {
barrier_local_test:
%CLK_LOCAL_MEM_FENCE = call i32 @__clc_clk_local_mem_fence()
%0 = and i32 %flags, %CLK_LOCAL_MEM_FENCE
@@ -11,7 +11,7 @@ barrier_local_test:
br i1 %1, label %barrier_local, label %barrier_global_test
barrier_local:
- call void @llvm.AMDGPU.barrier.local() noduplicate
+ call void @llvm.AMDGPU.barrier.local()
br label %barrier_global_test
barrier_global_test:
@@ -21,9 +21,13 @@ barrier_global_test:
br i1 %3, label %barrier_global, label %done
barrier_global:
- call void @llvm.AMDGPU.barrier.global() noduplicate
+ call void @llvm.AMDGPU.barrier.global()
br label %done
done:
ret void
}
+
+attributes #0 = { nounwind convergent }
+attributes #1 = { nounwind alwaysinline }
+attributes #2 = { nounwind convergent alwaysinline }
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-opencl/libclc.git
More information about the Pkg-opencl-commits
mailing list