nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-23 09:57:08 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: pd_cache enablement for >4k allocations in qnx

Browse Source 
Mapping of large buffers to GMMU end up needing many
pages for the PTE tables. Allocating these one by one
can end up being a performance bottleneck, particularly
in the virtualized case.

This is adding the following changes:

 - As the TLB invalidation doesn't have access to mem_off,
   allow top-level allocation by alloc_cache_direct().
 - Define NVGPU_PD_CACHE_SIZE, the allocation size for a new slab
   for the PD cache, effectively set to 64K bytes
 - Use the PD cache for any allocation < NVGPU_PD_CACHE_SIZE
   When freeing up cached entries, avoid prefetch errors by
   invalidating the entry (memset to 0).
 - Try to fall back to direct allocation of smaller chunk for
   contiguous allocation failures.
 - Unit test changes.

Bug 200649243

Change-Id: I0a667af0ba01d9147c703e64fc970880e52a8fbc
Signed-off-by: dt <dt@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2404371
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Peter Daifuku
2020-08-26 16:25:36 -07:00
committed by Alex Waterman
parent 94bc3a8135
commit a331fd4b3a
16 changed files with 122 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
userspace/units/mm/gmmu/pd_cache/pd_cache.c
View File

@@ -1,5 +1,5 @@
/*
* Copyright (c) 2018-2019, NVIDIA CORPORATION. All rights reserved.
* Copyright (c) 2018-2020, NVIDIA CORPORATION. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
@@ -321,7 +321,7 @@ int test_pd_free_empty_pd(struct unit_module *m, struct gk20a *g,
/* And now direct frees. */
memset(&pd, 0U, sizeof(pd));
err = nvgpu_pd_alloc(&vm, &pd, PAGE_SIZE);
err = nvgpu_pd_alloc(&vm, &pd, NVGPU_PD_CACHE_SIZE);
if (err != 0) {
unit_return_fail(m, "PD alloc failed");
}
@@ -610,7 +610,7 @@ static int do_test_pd_cache_packing_size(struct unit_module *m, struct gk20a *g,
{
int err;
u32 i;
u32 n = PAGE_SIZE / pd_size;
u32 n = NVGPU_PD_CACHE_SIZE / pd_size;
struct nvgpu_gmmu_pd pds[n], pd;
struct nvgpu_posix_fault_inj *dma_fi =
nvgpu_dma_alloc_get_fault_injection();
@@ -667,7 +667,7 @@ static int do_test_pd_reusability(struct unit_module *m, struct gk20a *g,
{
int err = UNIT_SUCCESS;
u32 i;
u32 n = PAGE_SIZE / pd_size;
u32 n = NVGPU_PD_CACHE_SIZE / pd_size;
struct nvgpu_gmmu_pd pds[n];
struct nvgpu_posix_fault_inj *dma_fi =
nvgpu_dma_alloc_get_fault_injection();