diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 21 |
1 files changed, 12 insertions, 9 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index a699134a1e8c..cbd593f7d553 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -40,10 +40,10 @@ #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1 /* - * Align VRAM allocations to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB + * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB * BO chunk */ -#define VRAM_ALLOCATION_ALIGN (1 << 21) +#define VRAM_AVAILABLITY_ALIGN (1 << 21) /* Impose limit on how much memory KFD can use */ static struct { @@ -149,7 +149,7 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, * to avoid fragmentation caused by 4K allocations in the tail * 2M BO chunk. */ - vram_needed = ALIGN(size, VRAM_ALLOCATION_ALIGN); + vram_needed = size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { system_mem_needed = size; } else if (!(alloc_flag & @@ -182,8 +182,10 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, */ WARN_ONCE(vram_needed && !adev, "adev reference can't be null when vram is used"); - if (adev) + if (adev) { adev->kfd.vram_used += vram_needed; + adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN); + } kfd_mem_limit.system_mem_used += system_mem_needed; kfd_mem_limit.ttm_mem_used += ttm_mem_needed; @@ -203,8 +205,10 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev, } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { WARN_ONCE(!adev, "adev reference can't be null when alloc mem flags vram is set"); - if (adev) - adev->kfd.vram_used -= ALIGN(size, VRAM_ALLOCATION_ALIGN); + if (adev) { + adev->kfd.vram_used -= size; + adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN); + } } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { kfd_mem_limit.system_mem_used -= size; } else if (!(alloc_flag & @@ -1608,15 +1612,14 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev) uint64_t reserved_for_pt = ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size); size_t available; - spin_lock(&kfd_mem_limit.mem_limit_lock); available = adev->gmc.real_vram_size - - adev->kfd.vram_used + - adev->kfd.vram_used_aligned - atomic64_read(&adev->vram_pin_size) - reserved_for_pt; spin_unlock(&kfd_mem_limit.mem_limit_lock); - return ALIGN_DOWN(available, VRAM_ALLOCATION_ALIGN); + return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN); } int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( |