[PATCH 1/1] Add test for hsaKmtAvailableMemory available memory inquiry

Felix Kuehling felix.kuehling at amd.com
Mon Jan 10 23:46:52 UTC 2022


On 2022-01-10 4:48 p.m., Daniel Phillips wrote:
> Basic test for the new hsaKmtAvailableMemory library call. This is
> a standalone test, does not modify any of the other tests just to
> be on the safe side. More elaborate tests coming soon.
>
> Change-Id: I738600d4b74cc5dba6b857e4c793f6b14b7d2283
> Signed-off-by: Daniel Phillips <daniel.phillips at amd.com>
> ---
>   tests/kfdtest/src/KFDMemoryTest.cpp | 17 +++++++++++++++++
>   1 file changed, 17 insertions(+)
>
> diff --git a/tests/kfdtest/src/KFDMemoryTest.cpp b/tests/kfdtest/src/KFDMemoryTest.cpp
> index 9f62727..1f93928 100644
> --- a/tests/kfdtest/src/KFDMemoryTest.cpp
> +++ b/tests/kfdtest/src/KFDMemoryTest.cpp
> @@ -595,6 +595,23 @@ TEST_F(KFDMemoryTest, MemoryAlloc) {
>       TEST_END
>   }
>   
> +// Basic test for hsaKmtAllocMemory
> +TEST_F(KFDMemoryTest, MemoryAllocAll) {
> +    TEST_START(TESTPROFILE_RUNALL)
> +
> +    unsigned int* pBig = NULL;
> +    unsigned int* pSmall = NULL;
> +    m_MemoryFlags.ui32.NoNUMABind = 1;
> +    HSAuint64 available;
> +    EXPECT_SUCCESS(hsaKmtAvailableMemory(0 /* system */, &available));
I don't think you've even implemented this API for system memory. The 
system memory node doesn't have a valid GPUID, so the ioctl will fail.

I'd expect this test to work only for VRAM.


> +    EXPECT_SUCCESS(hsaKmtAllocMemory(0 /* system */, available, m_MemoryFlags, reinterpret_cast<void**>(&pBig)));
> +    EXPECT_NE(HSAKMT_STATUS_SUCCESS, hsaKmtAllocMemory(0 /* system */, PAGE_SIZE, m_MemoryFlags, reinterpret_cast<void**>(&pSmall)));
> +    EXPECT_SUCCESS(hsaKmtFreeMemory(pBig, available));
> +    EXPECT_SUCCESS(hsaKmtAllocMemory(0 /* system */, PAGE_SIZE, m_MemoryFlags, reinterpret_cast<void**>(&pSmall)));

You're leaking pSmall here.

Regards,
   Felix


> +
> +    TEST_END
> +}
> +
>   TEST_F(KFDMemoryTest, AccessPPRMem) {
>       TEST_START(TESTPROFILE_RUNALL)
>   


More information about the amd-gfx mailing list