Skip to contentSkip to conversation

winsys/amdgpu: align VA allocations to fragment size

PATCH In review Christian König 2 participants 2 messages May 23, 2017 1 message not in the archive

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 winsys/amdgpu: align VA allocations to fragment size

From: Christian König <[email protected]>

BOs larger than the minimum fragment size should have their VA alignet to at least the fragment size for optimal performance.

---
 src/amd/common/ac_gpu_info.c                  | 1 +
 src/amd/common/ac_gpu_info.h                  | 1 +
 src/gallium/winsys/amdgpu/drm/amdgpu_bo.c     | 2 ++
 src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h | 1 +
 4 files changed, 5 insertions(+)
src/amd/common/ac_gpu_info.c +1 -0
index cf5d6e1..0b4933e 100644
@@ -281,6 +281,7 @@ bool ac_query_gpu_info(int fd, amdgpu_device_handle dev,
 	memcpy(info->cik_macrotile_mode_array, amdinfo->gb_macro_tile_mode,
 		sizeof(amdinfo->gb_macro_tile_mode));
 
+	info->pte_fragment_size = alignment_info.size_local;
 	info->gart_page_size = alignment_info.size_remote;
 
 	if (info->chip_class == SI)
src/amd/common/ac_gpu_info.h +1 -0
index d8029ef..3785eb4 100644
@@ -45,6 +45,7 @@ struct radeon_info {
 	uint32_t                    pci_id;
 	enum radeon_family          family;
 	enum chip_class             chip_class;
+	uint32_t                    pte_fragment_size;
 	uint32_t                    gart_page_size;
 	uint64_t                    gart_size;
 	uint64_t                    vram_size;
src/gallium/winsys/amdgpu/drm/amdgpu_bo.c +2 -0
index 6bdcce5..4017411 100644
@@ -415,6 +415,8 @@ static struct amdgpu_winsys_bo *amdgpu_create_bo(struct amdgpu_winsys *ws,
    }
 
    va_gap_size = ws->check_vm ? MAX2(4 * alignment, 64 * 1024) : 0;
+   if (size > ws->info.pte_fragment_size)
+	   alignment = MAX2(alignment, ws->info.pte_fragment_size);
    r = amdgpu_va_range_alloc(ws->dev, amdgpu_gpu_va_range_general,
                              size + va_gap_size, alignment, 0, &va, &va_handle, 0);
    if (r)
src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h +1 -0
index 896a463..88975e2 100644
@@ -73,6 +73,7 @@ struct amdgpu_winsys {
 
    struct amdgpu_gpu_info amdinfo;
    ADDR_HANDLE addrlib;
+   struct amdgpu_buffer_size_alignments	alignments;
 
    bool check_vm;
 
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1495557810-2577-1-git-send-email-deathsimple-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org

Reviewed-by
#2 lore raw
Quoted 55 lines · Christian König

From: Christian König <[email protected]>

BOs larger than the minimum fragment size should have their VA alignet to at least the fragment size for optimal performance.

---
 src/amd/common/ac_gpu_info.c                  | 1 +
 src/amd/common/ac_gpu_info.h                  | 1 +
 src/gallium/winsys/amdgpu/drm/amdgpu_bo.c     | 2 ++
 src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h | 1 +
 4 files changed, 5 insertions(+)
src/amd/common/ac_gpu_info.c +1 -0
index cf5d6e1..0b4933e 100644
@@ -281,6 +281,7 @@ bool ac_query_gpu_info(int fd, amdgpu_device_handle dev,
        memcpy(info->cik_macrotile_mode_array, amdinfo->gb_macro_tile_mode,
                sizeof(amdinfo->gb_macro_tile_mode));
 
+       info->pte_fragment_size = alignment_info.size_local;
        info->gart_page_size = alignment_info.size_remote;
 
        if (info->chip_class == SI)
src/amd/common/ac_gpu_info.h +1 -0
index d8029ef..3785eb4 100644
@@ -45,6 +45,7 @@ struct radeon_info {
        uint32_t                    pci_id;
        enum radeon_family          family;
        enum chip_class             chip_class;
+       uint32_t                    pte_fragment_size;
        uint32_t                    gart_page_size;
        uint64_t                    gart_size;
        uint64_t                    vram_size;
src/gallium/winsys/amdgpu/drm/amdgpu_bo.c +2 -0
index 6bdcce5..4017411 100644
@@ -415,6 +415,8 @@ static struct amdgpu_winsys_bo *amdgpu_create_bo(struct amdgpu_winsys *ws,
    }
 
    va_gap_size = ws->check_vm ? MAX2(4 * alignment, 64 * 1024) : 0;
+   if (size > ws->info.pte_fragment_size)
+          alignment = MAX2(alignment, ws->info.pte_fragment_size);
    r = amdgpu_va_range_alloc(ws->dev, amdgpu_gpu_va_range_general,
                              size + va_gap_size, alignment, 0, &va, &va_handle, 0);
    if (r)
src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h +0 -0
index 896a463..88975e2 100644

src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h struct amdgpu_winsys { · line 78

Quoted 6 lines
src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h +1 -0
@@ -73,6 +73,7 @@ struct amdgpu_winsys {
 
    struct amdgpu_gpu_info amdinfo;
    ADDR_HANDLE addrlib;
+   struct amdgpu_buffer_size_alignments        alignments;
 
    bool check_vm;

This hunk seems unnecessary. If you remove it, the patch is:

Marek
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx