intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-24 20:39:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor: Correct logic for SIMD1

Browse Source 
- For calculating number of threads per workgroup, for SIMD 1, return
local work size (each software thread should be mapped into a whole hardware
thread).
- Correct logic of calculating space for per thread data for SIMD 1.
- Minor: unit tests refactor.
- Corrected naming.
Related-To: NEO-8261
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
This commit is contained in:
Kacper Nowak
2023-09-26 14:07:13 +00:00
committed by Compute-Runtime-Automation
parent 3c2ef4fbbf
commit 160303924d
4 changed files with 71 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
shared/test/unit_test/helpers/local_id_tests.cpp
View File

@@ -23,22 +23,27 @@ using LocalIdTests = ::testing::Test;
HWTEST_F(LocalIdTests, GivenSimd8WhenGettingGrfsPerThreadThenOneIsReturned) {
uint32_t simd = 8;
EXPECT_EQ(1u, getGRFsPerThread(simd, 32));
EXPECT_EQ(1u, getNumGrfsPerLocalIdCoordinate(simd, 32));
}
HWTEST_F(LocalIdTests, GivenSimd16WhenGettingGrfsPerThreadThenOneIsReturned) {
uint32_t simd = 16;
EXPECT_EQ(1u, getGRFsPerThread(simd, 32));
EXPECT_EQ(1u, getNumGrfsPerLocalIdCoordinate(simd, 32));
}
HWTEST_F(LocalIdTests, GivenSimd32WhenGettingGrfsPerThreadThenTwoIsReturned) {
uint32_t simd = 32;
EXPECT_EQ(2u, getGRFsPerThread(simd, 32));
EXPECT_EQ(2u, getNumGrfsPerLocalIdCoordinate(simd, 32));
}
HWTEST_F(LocalIdTests, GivenSimd1WhenGettingGrfsPerThreadThenOneIsReturned) {
uint32_t simd = 1;
EXPECT_EQ(1u, getNumGrfsPerLocalIdCoordinate(simd, 32));
}
HWTEST_F(LocalIdTests, GivenSimd32AndNon32GrfSizeWhenGettingGrfsPerThreadThenTwoIsReturned) {
uint32_t simd = 32;
EXPECT_EQ(1u, getGRFsPerThread(simd, 33));
EXPECT_EQ(1u, getNumGrfsPerLocalIdCoordinate(simd, 33));
}
TEST(LocalID, GivenSimd32AndLws33WhenGettingThreadsPerWorkgroupThenTwoIsReturned) {
@@ -78,6 +83,28 @@ TEST(LocalID, GivenSimd1WhenGettingPerThreadSizeLocalIdsThenValueIsEqualGrfSize)
EXPECT_EQ(grfSize, getPerThreadSizeLocalIDs(simd, grfSize));
}
TEST(LocalID, WhenThreadsPerWgAreGeneratedThenCalculationsAreCorrect) {
const auto lws = 33u;
for (const auto &simd : {1u, 8u, 16u, 32u}) {
switch (simd) {
case 1u:
EXPECT_EQ(lws, getThreadsPerWG(simd, lws));
break;
case 32u:
EXPECT_EQ((lws + std::max(32u, simd) - 1) >> 5, getThreadsPerWG(simd, lws));
break;
case 8u:
EXPECT_EQ((lws + simd - 1) >> 3, getThreadsPerWG(simd, lws));
break;
case 16u:
EXPECT_EQ((lws + simd - 1) >> 4, getThreadsPerWG(simd, lws));
break;
default:
break;
}
}
}
TEST(LocalIdTest, givenVariadicGrfSizeWhenLocalSizesAreEmittedThenUseFullRowSize) {
auto localIdsPtr = allocateAlignedMemory(3 * 64u, MemoryConstants::cacheLineSize);