[Openmp-commits] [PATCH] D103090: [libomptarget][nfc][amdgpu] Factor out setting upper bounds
Jon Chesterfield via Phabricator via Openmp-commits
openmp-commits at lists.llvm.org
Wed May 26 11:23:42 PDT 2021
JonChesterfield updated this revision to Diff 348032.
JonChesterfield added a comment.
- twist code around clang-tidy
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D103090/new/
https://reviews.llvm.org/D103090
Files:
openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
Index: openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
===================================================================
--- openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
+++ openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
@@ -708,6 +708,16 @@
return RequiresFlags;
}
+namespace {
+template <typename T> bool enforce_upper_bound(T *value, T upper) {
+ bool changed = *value > upper;
+ if (changed) {
+ *value = upper;
+ }
+ return changed;
+}
+} // namespace
+
int32_t __tgt_rtl_init_device(int device_id) {
hsa_status_t err;
@@ -769,11 +779,13 @@
DeviceInfo.ThreadsPerGroup[device_id] =
reinterpret_cast<uint32_t *>(&grid_max_dim)[0] /
DeviceInfo.GroupsPerDevice[device_id];
- if ((DeviceInfo.ThreadsPerGroup[device_id] >
- RTLDeviceInfoTy::Max_WG_Size) ||
- DeviceInfo.ThreadsPerGroup[device_id] == 0) {
- DP("Capped thread limit: %d\n", RTLDeviceInfoTy::Max_WG_Size);
+
+ if (DeviceInfo.ThreadsPerGroup[device_id] == 0) {
DeviceInfo.ThreadsPerGroup[device_id] = RTLDeviceInfoTy::Max_WG_Size;
+ DP("Default thread limit: %d\n", RTLDeviceInfoTy::Max_WG_Size);
+ } else if (enforce_upper_bound(&DeviceInfo.ThreadsPerGroup[device_id],
+ RTLDeviceInfoTy::Max_WG_Size)) {
+ DP("Capped thread limit: %d\n", RTLDeviceInfoTy::Max_WG_Size);
} else {
DP("Using ROCm Queried thread limit: %d\n",
DeviceInfo.ThreadsPerGroup[device_id]);
@@ -799,9 +811,10 @@
}
// Adjust teams to the env variables
+
if (DeviceInfo.EnvTeamLimit > 0 &&
- DeviceInfo.GroupsPerDevice[device_id] > DeviceInfo.EnvTeamLimit) {
- DeviceInfo.GroupsPerDevice[device_id] = DeviceInfo.EnvTeamLimit;
+ (enforce_upper_bound(&DeviceInfo.GroupsPerDevice[device_id],
+ DeviceInfo.EnvTeamLimit))) {
DP("Capping max groups per device to OMP_TEAM_LIMIT=%d\n",
DeviceInfo.EnvTeamLimit);
}
@@ -824,8 +837,8 @@
TeamsPerCU, DeviceInfo.ComputeUnits[device_id]);
}
- if (DeviceInfo.NumTeams[device_id] > DeviceInfo.GroupsPerDevice[device_id]) {
- DeviceInfo.NumTeams[device_id] = DeviceInfo.GroupsPerDevice[device_id];
+ if (enforce_upper_bound(&DeviceInfo.NumTeams[device_id],
+ DeviceInfo.GroupsPerDevice[device_id])) {
DP("Default number of teams exceeds device limit, capping at %d\n",
DeviceInfo.GroupsPerDevice[device_id]);
}
@@ -834,9 +847,8 @@
DeviceInfo.NumThreads[device_id] = RTLDeviceInfoTy::Default_WG_Size;
DP("Default number of threads set according to library's default %d\n",
RTLDeviceInfoTy::Default_WG_Size);
- if (DeviceInfo.NumThreads[device_id] >
- DeviceInfo.ThreadsPerGroup[device_id]) {
- DeviceInfo.NumThreads[device_id] = DeviceInfo.ThreadsPerGroup[device_id];
+ if (enforce_upper_bound(&DeviceInfo.NumThreads[device_id],
+ DeviceInfo.ThreadsPerGroup[device_id])) {
DP("Default number of threads exceeds device limit, capping at %d\n",
DeviceInfo.ThreadsPerGroup[device_id]);
}
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D103090.348032.patch
Type: text/x-patch
Size: 3085 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/openmp-commits/attachments/20210526/b6918dbf/attachment.bin>
More information about the Openmp-commits
mailing list