intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-07 21:27:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

Allow squared algorithm to work together with base one.

Browse Source 
Change-Id: I9087957bb427a422b1be632f6375c96b8f91a492
This commit is contained in:
Mrozek, Michal
2018-01-12 09:58:30 +01:00
committed by sys_ocldev
parent 031b537e2b
commit 7640201585
3 changed files with 56 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
unit_tests/command_queue/local_work_size_tests.cpp
View File

@@ -22,6 +22,7 @@
#include "runtime/command_queue/dispatch_walker.h"
#include "runtime/helpers/options.h"
#include "unit_tests/helpers/debug_manager_state_restore.h"
#include "gtest/gtest.h"
using namespace OCLRT;
@@ -95,8 +96,8 @@ TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual8WhenComputeCalledThenLoca
size_t workGroupSize[3];
OCLRT::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 128u);
EXPECT_EQ(workGroupSize[1], 2u);
EXPECT_EQ(workGroupSize[0], 16u);
EXPECT_EQ(workGroupSize[1], 16u);
EXPECT_EQ(workGroupSize[2], 1u);
workGroup[0] = 48;
@@ -115,6 +116,8 @@ TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual8WhenComputeCalledThenLoca
}
TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual32WhenComputeCalledThenLocalGroupComputed) {
DebugManagerStateRestore dbgRestore;
DebugManager.flags.EnableComputeWorkSizeSquared.set(false);
WorkSizeInfo wsInfo(256, 0u, 32, 0u, platformDevices[0]->pPlatform->eRenderCoreFamily, 32u, 0u, false, false);
uint32_t workDim = 2;
size_t workGroup[3] = {384, 96, 1};
@@ -232,8 +235,8 @@ TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual256WhenComputeCalledThenLo
size_t workGroupSize[3];
OCLRT::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 128u);
EXPECT_EQ(workGroupSize[1], 2u);
EXPECT_EQ(workGroupSize[0], 16u);
EXPECT_EQ(workGroupSize[1], 16u);
EXPECT_EQ(workGroupSize[2], 1u);
}
@@ -267,8 +270,8 @@ TEST(localWorkSizeTest, givenKernelWithTwoDimensionalGlobalSizesWhenLwsIsCompute
workGroup[0] = 1024;
workGroup[1] = 1024;
OCLRT::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 256u);
EXPECT_EQ(workGroupSize[1], 1u);
EXPECT_EQ(workGroupSize[0], 16u);
EXPECT_EQ(workGroupSize[1], 16u);
EXPECT_EQ(workGroupSize[2], 1u);
}
@@ -521,6 +524,8 @@ TEST(localWorkSizeTest, given2DimWorkWhenComputeSquaredCalledThenLocalGroupCompu
}
TEST(localWorkSizeTest, givenDeviceSupportingLws1024AndKernelCompiledInSimd8WhenGwsIs1024ThenLwsIsComputedAsMaxOptimalMultipliedBySimd) {
DebugManagerStateRestore dbgRestore;
DebugManager.flags.EnableComputeWorkSizeSquared.set(false);
WorkSizeInfo wsInfo(1024, 0u, 8, 0u, platformDevices[0]->pPlatform->eRenderCoreFamily, 32u, 0u, false, false);
uint32_t workDim = 2;
@@ -535,4 +540,7 @@ TEST(localWorkSizeTest, givenDeviceSupportingLws1024AndKernelCompiledInSimd8When
TEST(localWorkSizeTest, givenDebugVariableEnableComputeWorkSizeNDWhenCheckValueExpectTrue) {
EXPECT_TRUE(DebugManager.flags.EnableComputeWorkSizeND.get());
}
TEST(localWorkSizeTest, givenDefaultDebugVariablesWhenEnableComputeWorkSizeSquaredIsCheckdThenTrueIsReturned) {
EXPECT_TRUE(DebugManager.flags.EnableComputeWorkSizeSquared.get());
}