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

Simplify bit operations.

Browse Source 
Change-Id: If1401f32df5ebcb3abf614832152abf029a9c5e6
Signed-off-by: Piotr Fusik <piotr.fusik@intel.com>
This commit is contained in:
Piotr Fusik
2019-03-27 15:50:29 +01:00
committed by sys_ocldev
parent 6d2e92cf20
commit 8cf7cea1e9
5 changed files with 26 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
runtime/command_queue/local_work_size.cpp
View File

@@ -296,16 +296,16 @@ void computeWorkgroupSizeSquared(uint32_t maxWorkGroupSize, size_t workGroupSize
size_t itemsPowerOfTwoDivisors[3] = {1, 1, 1};
for (auto i = 0u; i < workDim; i++) {
uint32_t requiredWorkItemsCount = maxWorkGroupSize;
while (requiredWorkItemsCount > 1 && !(Math::isDivisableByPowerOfTwoDivisor(uint32_t(workItems[i]), requiredWorkItemsCount)))
requiredWorkItemsCount = requiredWorkItemsCount >> 1;
while (requiredWorkItemsCount > 1 && !(Math::isDivisibleByPowerOfTwoDivisor(uint32_t(workItems[i]), requiredWorkItemsCount)))
requiredWorkItemsCount >>= 1;
itemsPowerOfTwoDivisors[i] = requiredWorkItemsCount;
}
if (itemsPowerOfTwoDivisors[0] * itemsPowerOfTwoDivisors[1] >= maxWorkGroupSize) {
while (itemsPowerOfTwoDivisors[0] * itemsPowerOfTwoDivisors[1] > maxWorkGroupSize) {
if (itemsPowerOfTwoDivisors[0] > itemsPowerOfTwoDivisors[1])
itemsPowerOfTwoDivisors[0] = itemsPowerOfTwoDivisors[0] >> 1;
itemsPowerOfTwoDivisors[0] >>= 1;
else
itemsPowerOfTwoDivisors[1] = itemsPowerOfTwoDivisors[1] >> 1;
itemsPowerOfTwoDivisors[1] >>= 1;
}
for (auto i = 0u; i < 3; i++)
workGroupSize[i] = itemsPowerOfTwoDivisors[i];
@@ -338,8 +338,8 @@ void computeWorkgroupSizeND(WorkSizeInfo wsInfo, size_t workGroupSize[3], const
size_t itemsPowerOfTwoDivisors[3] = {1, 1, 1};
for (auto i = 0u; i < workDim; i++) {
uint32_t requiredWorkItemsCount = uint32_t(wsInfo.simdSize * optimalHardwareThreadCountGeneric[0]);
while (requiredWorkItemsCount > 1 && !(Math::isDivisableByPowerOfTwoDivisor(uint32_t(workItems[i]), requiredWorkItemsCount)))
requiredWorkItemsCount = requiredWorkItemsCount >> 1;
while (requiredWorkItemsCount > 1 && !(Math::isDivisibleByPowerOfTwoDivisor(uint32_t(workItems[i]), requiredWorkItemsCount)))
requiredWorkItemsCount >>= 1;
itemsPowerOfTwoDivisors[i] = requiredWorkItemsCount;
}