leandrof/linux
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

drm/amdgpu: reserve at least 4MB of VRAM for page tables v2

Browse Source 
This hopefully helps reduce the contention for page tables.

v2: adjust maximum reported VRAM size as well

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Christian König
2019-08-30 14:38:37 +02:00
committed by Alex Deucher
parent 629be20395
commit 9d1b3c7805
3 changed files with 22 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
View File

@@ -619,9 +619,12 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
struct drm_amdgpu_info_vram_gtt vram_gtt; struct drm_amdgpu_info_vram_gtt vram_gtt;
vram_gtt.vram_size = adev->gmc.real_vram_size - vram_gtt.vram_size = adev->gmc.real_vram_size -
atomic64_read(&adev->vram_pin_size); atomic64_read(&adev->vram_pin_size) -
vram_gtt.vram_cpu_accessible_size = adev->gmc.visible_vram_size - AMDGPU_VM_RESERVED_VRAM;
atomic64_read(&adev->visible_pin_size); vram_gtt.vram_cpu_accessible_size =
min(adev->gmc.visible_vram_size -
atomic64_read(&adev->visible_pin_size),
vram_gtt.vram_size);
vram_gtt.gtt_size = adev->mman.bdev.man[TTM_PL_TT].size; vram_gtt.gtt_size = adev->mman.bdev.man[TTM_PL_TT].size;
vram_gtt.gtt_size *= PAGE_SIZE; vram_gtt.gtt_size *= PAGE_SIZE;
vram_gtt.gtt_size -= atomic64_read(&adev->gart_pin_size); vram_gtt.gtt_size -= atomic64_read(&adev->gart_pin_size);
@@ -634,15 +637,18 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
memset(&mem, 0, sizeof(mem)); memset(&mem, 0, sizeof(mem));
mem.vram.total_heap_size = adev->gmc.real_vram_size; mem.vram.total_heap_size = adev->gmc.real_vram_size;
mem.vram.usable_heap_size = adev->gmc.real_vram_size - mem.vram.usable_heap_size = adev->gmc.real_vram_size -
atomic64_read(&adev->vram_pin_size); atomic64_read(&adev->vram_pin_size) -
AMDGPU_VM_RESERVED_VRAM;
mem.vram.heap_usage = mem.vram.heap_usage =
amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]);
mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4; mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4;
mem.cpu_accessible_vram.total_heap_size = mem.cpu_accessible_vram.total_heap_size =
adev->gmc.visible_vram_size; adev->gmc.visible_vram_size;
mem.cpu_accessible_vram.usable_heap_size = adev->gmc.visible_vram_size - mem.cpu_accessible_vram.usable_heap_size =
atomic64_read(&adev->visible_pin_size); min(adev->gmc.visible_vram_size -
atomic64_read(&adev->visible_pin_size),
mem.vram.usable_heap_size);
mem.cpu_accessible_vram.heap_usage = mem.cpu_accessible_vram.heap_usage =
amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]);
mem.cpu_accessible_vram.max_allocation = mem.cpu_accessible_vram.max_allocation =

3
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
View File

@@ -99,6 +99,9 @@ struct amdgpu_bo_list_entry;
#define AMDGPU_VM_FAULT_STOP_FIRST 1 #define AMDGPU_VM_FAULT_STOP_FIRST 1
#define AMDGPU_VM_FAULT_STOP_ALWAYS 2 #define AMDGPU_VM_FAULT_STOP_ALWAYS 2
/* Reserve 4MB VRAM for page tables */
#define AMDGPU_VM_RESERVED_VRAM (4ULL << 20)
/* max number of VMHUB */ /* max number of VMHUB */
#define AMDGPU_MAX_VMHUBS 3 #define AMDGPU_MAX_VMHUBS 3
#define AMDGPU_GFXHUB_0 0 #define AMDGPU_GFXHUB_0 0

9
drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
View File

@@ -23,6 +23,7 @@
*/ */
#include "amdgpu.h" #include "amdgpu.h"
#include "amdgpu_vm.h"
struct amdgpu_vram_mgr { struct amdgpu_vram_mgr {
struct drm_mm mm; struct drm_mm mm;
@@ -275,7 +276,7 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man,
struct drm_mm_node *nodes; struct drm_mm_node *nodes;
enum drm_mm_insert_mode mode; enum drm_mm_insert_mode mode;
unsigned long lpfn, num_nodes, pages_per_node, pages_left; unsigned long lpfn, num_nodes, pages_per_node, pages_left;
uint64_t vis_usage = 0, mem_bytes; uint64_t vis_usage = 0, mem_bytes, max_bytes;
unsigned i; unsigned i;
int r; int r;
@@ -283,9 +284,13 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man,
if (!lpfn) if (!lpfn)
lpfn = man->size; lpfn = man->size;
max_bytes = adev->gmc.mc_vram_size;
if (tbo->type != ttm_bo_type_kernel)
max_bytes -= AMDGPU_VM_RESERVED_VRAM;
/* bail out quickly if there's likely not enough VRAM for this BO */ /* bail out quickly if there's likely not enough VRAM for this BO */
mem_bytes = (u64)mem->num_pages << PAGE_SHIFT; mem_bytes = (u64)mem->num_pages << PAGE_SHIFT;
if (atomic64_add_return(mem_bytes, &mgr->usage) > adev->gmc.mc_vram_size) { if (atomic64_add_return(mem_bytes, &mgr->usage) > max_bytes) {
atomic64_sub(mem_bytes, &mgr->usage); atomic64_sub(mem_bytes, &mgr->usage);
mem->mm_node = NULL; mem->mm_node = NULL;
return 0; return 0;