[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <90e62dcc-49f2-84d7-d845-1d05c9f3dd08@amd.com>
Date: Mon, 28 Nov 2022 22:39:39 +0530
From: Arunpravin Paneer Selvam <arunpravin.paneerselvam@....com>
To: xinhui pan <xinhui.pan@....com>, amd-gfx@...ts.freedesktop.org
Cc: linux-kernel@...r.kernel.org, dri-devel@...ts.freedesktop.org,
matthew.auld@...el.com, daniel@...ll.ch, christian.koenig@....com
Subject: Re: [PATCH v3] drm: Optimise for continuous memory allocation
Hi Xinhui,
On 11/28/2022 12:04 PM, xinhui pan wrote:
> Currently drm-buddy does not have full knowledge of continuous memory.
>
> Lets consider scenario below.
> order 1: L R
> order 0: LL LR RL RR
> for order 1 allocation, it can offer L or R or LR+RL.
>
> For now, we only implement L or R case for continuous memory allocation.
> So this patch aims to implement the LR+RL case.
>
> Signed-off-by: xinhui pan <xinhui.pan@....com>
> ---
> change from v2:
> search continuous block in nearby root if needed
>
> change from v1:
> implement top-down continuous allocation
> ---
> drivers/gpu/drm/drm_buddy.c | 78 +++++++++++++++++++++++++++++++++----
> 1 file changed, 71 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_buddy.c b/drivers/gpu/drm/drm_buddy.c
> index 11bb59399471..ff58eb3136d2 100644
> --- a/drivers/gpu/drm/drm_buddy.c
> +++ b/drivers/gpu/drm/drm_buddy.c
> @@ -386,6 +386,58 @@ alloc_range_bias(struct drm_buddy *mm,
> return ERR_PTR(err);
> }
>
> +static struct drm_buddy_block *
> +find_continuous_blocks(struct drm_buddy *mm,
> + int order,
> + unsigned long flags,
> + struct drm_buddy_block **rn)
> +{
> + struct list_head *head = &mm->free_list[order];
> + struct drm_buddy_block *node, *parent, *free_node, *max_node = NULL;
NIT: We usually name the variable as *block or ***_block for drm buddy
and we have *node or ***_node for drm mm manager.
> + int i;
> +
> + list_for_each_entry(free_node, head, link) {
> + if (max_node) {
> + if (!(flags & DRM_BUDDY_TOPDOWN_ALLOCATION))
> + break;
> +
> + if (drm_buddy_block_offset(free_node) <
> + drm_buddy_block_offset(max_node))
> + continue;
> + }
> +
> + parent = free_node;
> + do {
> + node = parent;
> + parent = parent->parent;
> + } while (parent && parent->right == node);
> +
> + if (!parent) {
> + for (i = 0; i < mm->n_roots - 1; i++)
> + if (mm->roots[i] == node)
> + break;
> + if (i == mm->n_roots - 1)
> + continue;
> + node = mm->roots[i + 1];
> + } else {
> + node = parent->right;
> + }
> +
> + while (drm_buddy_block_is_split(node))
> + node = node->left;
> +
> + if (drm_buddy_block_is_free(node) &&
> + drm_buddy_block_order(node) == order) {
> + *rn = node;
> + max_node = free_node;
> + BUG_ON(drm_buddy_block_offset(node) !=
> + drm_buddy_block_offset(max_node) +
> + drm_buddy_block_size(mm, max_node));
> + }
> + }
> + return max_node;
> +}
> +
> static struct drm_buddy_block *
> get_maxblock(struct list_head *head)
> {
> @@ -637,7 +689,7 @@ int drm_buddy_alloc_blocks(struct drm_buddy *mm,
> struct list_head *blocks,
> unsigned long flags)
> {
> - struct drm_buddy_block *block = NULL;
> + struct drm_buddy_block *block = NULL, *rblock = NULL;
> unsigned int min_order, order;
> unsigned long pages;
> LIST_HEAD(allocated);
> @@ -689,17 +741,29 @@ int drm_buddy_alloc_blocks(struct drm_buddy *mm,
> break;
>
> if (order-- == min_order) {
> + if (!(flags & DRM_BUDDY_RANGE_ALLOCATION) &&
> + min_order != 0 && pages == BIT(order + 1)) {
> + block = find_continuous_blocks(mm,
> + order,
> + flags,
> + &rblock);
> + if (block)
> + break;
> + }
> err = -ENOSPC;
> goto err_free;
> }
> } while (1);
>
> - mark_allocated(block);
> - mm->avail -= drm_buddy_block_size(mm, block);
> - kmemleak_update_trace(block);
> - list_add_tail(&block->link, &allocated);
> -
> - pages -= BIT(order);
> + do {
> + mark_allocated(block);
> + mm->avail -= drm_buddy_block_size(mm, block);
> + kmemleak_update_trace(block);
> + list_add_tail(&block->link, &allocated);
> + pages -= BIT(order);
> + block = rblock;
> + rblock = NULL;
> + } while (block);
I think with this approach, if we are lucky enough we may get contiguous
blocks in one order level down in RL
combination from the freelist?
Regards,
Arun
>
> if (!pages)
> break;
Powered by blists - more mailing lists