intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-29 17:13:29 +08:00
Code Issues Packages Projects Releases Wiki Activity

Make sure that local workgroup size is properly passed for IOH estimation.

Browse Source 
Change-Id: I0ad5da4fffd1575f64d44803ce8eb4a6a0ab1532
This commit is contained in:
Mrozek, Michal
2018-02-13 15:43:33 +01:00
committed by sys_ocldev
parent 8a5b0ee518
commit 2d0af9d4a4
7 changed files with 57 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
runtime/command_queue/dispatch_walker.h
View File

@@ -455,6 +455,14 @@ void dispatchWalker(
OCLRT::IndirectHeap *dsh = nullptr, *ish = nullptr, *ioh = nullptr, *ssh = nullptr;
bool executionModelKernel = multiDispatchInfo.begin()->getKernel()->isParentKernel;
for (auto &dispatchInfo : multiDispatchInfo) {
// Compute local workgroup sizes
if (dispatchInfo.getLocalWorkgroupSize().x == 0) {
const auto lws = generateWorkgroupSize(dispatchInfo);
const_cast<DispatchInfo &>(dispatchInfo).setLWS(lws);
}
}
// Allocate command stream and indirect heaps
size_t cmdQInstructionHeapReservedBlockSize = 0;
if (blockQueue) {
@@ -541,7 +549,7 @@ void dispatchWalker(
Vec3<size_t> swgs = dispatchInfo.getStartOfWorkgroups();
// Compute local workgroup sizes
Vec3<size_t> lws = (dispatchInfo.getLocalWorkgroupSize().x > 0) ? dispatchInfo.getLocalWorkgroupSize() : generateWorkgroupSize(dispatchInfo);
Vec3<size_t> lws = dispatchInfo.getLocalWorkgroupSize();
Vec3<size_t> elws = (dispatchInfo.getEnqueuedWorkgroupSize().x > 0) ? dispatchInfo.getEnqueuedWorkgroupSize() : lws;
// Compute number of work groups