From: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>
To: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>,
intel-xe@lists.freedesktop.org
Cc: Matthew Auld <matthew.auld@intel.com>
Subject: Re: [PATCH v7 07/10] drm/xe/xe2: Update emit_pte to use compression enabled PAT index
Date: Tue, 12 Dec 2023 13:28:51 +0100 [thread overview]
Message-ID: <5460aee0-2dc5-77f8-5884-d71dbc1e032a@linux.intel.com> (raw)
In-Reply-To: <20231211134356.1645973-8-himal.prasad.ghimiray@intel.com>
On 12/11/23 14:43, Himal Prasad Ghimiray wrote:
> For indirect accessed buffer use compression enabled PAT index.
>
> v2:
> - Fix parameter name.
>
> v3:
> - use a relevant define instead of fix number.
>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Matthew Auld <matthew.auld@intel.com>
> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> ---
> drivers/gpu/drm/xe/tests/xe_migrate.c | 2 +-
> drivers/gpu/drm/xe/xe_migrate.c | 21 +++++++++++++++------
> drivers/gpu/drm/xe/xe_pat.c | 1 +
> drivers/gpu/drm/xe/xe_pt_types.h | 1 +
> 4 files changed, 18 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/tests/xe_migrate.c b/drivers/gpu/drm/xe/tests/xe_migrate.c
> index 47fcd6e6b777..d6c23441632a 100644
> --- a/drivers/gpu/drm/xe/tests/xe_migrate.c
> +++ b/drivers/gpu/drm/xe/tests/xe_migrate.c
> @@ -330,7 +330,7 @@ static void xe_migrate_sanity_test(struct xe_migrate *m, struct kunit *test)
> else
> xe_res_first_sg(xe_bo_sg(pt), 0, pt->size, &src_it);
>
> - emit_pte(m, bb, NUM_KERNEL_PDE - 1, xe_bo_is_vram(pt),
> + emit_pte(m, bb, NUM_KERNEL_PDE - 1, xe_bo_is_vram(pt), false,
> &src_it, XE_PAGE_SIZE, pt);
>
> run_sanity_job(m, xe, bb, bb->len, "Writing PTE for our fake PT", test);
> diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c
> index 9698986eab06..1ecf2274c7ba 100644
> --- a/drivers/gpu/drm/xe/xe_migrate.c
> +++ b/drivers/gpu/drm/xe/xe_migrate.c
> @@ -422,15 +422,24 @@ static u32 pte_update_size(struct xe_migrate *m,
>
> static void emit_pte(struct xe_migrate *m,
> struct xe_bb *bb, u32 at_pt,
> - bool is_vram,
> + bool is_vram, bool is_comp_pte,
> struct xe_res_cursor *cur,
> u32 size, struct xe_bo *bo)
> {
> - u16 pat_index = tile_to_xe(m->tile)->pat.idx[XE_CACHE_WB];
> + struct xe_device *xe = tile_to_xe(m->tile);
> +
> + u16 pat_index;
> u32 ptes;
> u64 ofs = at_pt * XE_PAGE_SIZE;
> u64 cur_ofs;
>
> + /* Indirect access needs compression enabled uncached PAT index */
> + if (GRAPHICS_VERx100(xe) >= 2000)
> + pat_index = is_comp_pte ? xe->pat.idx[XE_CACHE_NONE_COMPRESSION] :
> + xe->pat.idx[XE_CACHE_NONE];
> + else
> + pat_index = xe->pat.idx[XE_CACHE_WB];
> +
> /*
> * FIXME: Emitting VRAM PTEs to L0 PTs is forbidden. Currently
> * we're only emitting VRAM PTEs during sanity tests, so when
> @@ -717,19 +726,19 @@ struct dma_fence *xe_migrate_copy(struct xe_migrate *m,
> }
>
> if (!src_is_vram)
> - emit_pte(m, bb, src_L0_pt, src_is_vram, &src_it, src_L0,
> + emit_pte(m, bb, src_L0_pt, src_is_vram, true, &src_it, src_L0,
> src_bo);
> else
> xe_res_next(&src_it, src_L0);
>
> if (!dst_is_vram)
> - emit_pte(m, bb, dst_L0_pt, dst_is_vram, &dst_it, src_L0,
> + emit_pte(m, bb, dst_L0_pt, dst_is_vram, true, &dst_it, src_L0,
> dst_bo);
> else
> xe_res_next(&dst_it, src_L0);
>
> if (copy_system_ccs)
> - emit_pte(m, bb, ccs_pt, false, &ccs_it, ccs_size, src_bo);
> + emit_pte(m, bb, ccs_pt, false, false, &ccs_it, ccs_size, src_bo);
>
> bb->cs[bb->len++] = MI_BATCH_BUFFER_END;
> update_idx = bb->len;
> @@ -962,7 +971,7 @@ struct dma_fence *xe_migrate_clear(struct xe_migrate *m,
>
> /* Preemption is enabled again by the ring ops. */
> if (!clear_vram) {
> - emit_pte(m, bb, clear_L0_pt, clear_vram, &src_it, clear_L0,
> + emit_pte(m, bb, clear_L0_pt, clear_vram, true, &src_it, clear_L0,
> bo);
> } else {
> xe_res_next(&src_it, clear_L0);
> diff --git a/drivers/gpu/drm/xe/xe_pat.c b/drivers/gpu/drm/xe/xe_pat.c
> index 1892ff81086f..1ff6bc79e7d4 100644
> --- a/drivers/gpu/drm/xe/xe_pat.c
> +++ b/drivers/gpu/drm/xe/xe_pat.c
> @@ -387,6 +387,7 @@ void xe_pat_init_early(struct xe_device *xe)
> xe->pat.idx[XE_CACHE_NONE] = 3;
> xe->pat.idx[XE_CACHE_WT] = 15;
> xe->pat.idx[XE_CACHE_WB] = 2;
> + xe->pat.idx[XE_CACHE_NONE_COMPRESSION] = 12; /*Applicable on xe2 and beyond */
> } else if (xe->info.platform == XE_METEORLAKE) {
> xe->pat.ops = &xelpg_pat_ops;
> xe->pat.table = xelpg_pat_table;
> diff --git a/drivers/gpu/drm/xe/xe_pt_types.h b/drivers/gpu/drm/xe/xe_pt_types.h
> index 82cbf1ef8e57..cee70cb0f014 100644
> --- a/drivers/gpu/drm/xe/xe_pt_types.h
> +++ b/drivers/gpu/drm/xe/xe_pt_types.h
> @@ -18,6 +18,7 @@ enum xe_cache_level {
> XE_CACHE_NONE,
> XE_CACHE_WT,
> XE_CACHE_WB,
> + XE_CACHE_NONE_COMPRESSION, /*UC + COH_NONE + COMPRESSION */
> __XE_CACHE_LEVEL_COUNT,
> };
>
next prev parent reply other threads:[~2023-12-12 12:29 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-11 13:43 [PATCH v7 00/10] Enable compression handling on LNL Himal Prasad Ghimiray
2023-12-11 13:43 ` [PATCH v7 01/10] drm/xe/xe2: Determine bios enablement for flat ccs on igfx Himal Prasad Ghimiray
2023-12-11 23:12 ` Matt Roper
2023-12-12 12:23 ` Thomas Hellström
2023-12-11 13:43 ` [PATCH v7 02/10] drm/xe/xe2: Modify main memory to ccs memory ratio Himal Prasad Ghimiray
2023-12-11 23:15 ` Matt Roper
2023-12-11 13:43 ` [PATCH v7 03/10] drm/xe/xe2: Allocate extra pages for ccs during bo create Himal Prasad Ghimiray
2023-12-12 0:41 ` Matt Roper
2023-12-12 9:00 ` Ghimiray, Himal Prasad
2023-12-11 13:43 ` [PATCH v7 04/10] drm/xe/xe2: Updates on XY_CTRL_SURF_COPY_BLT Himal Prasad Ghimiray
2023-12-12 0:45 ` Matt Roper
2023-12-11 13:43 ` [PATCH v7 05/10] drm/xe/xe_migrate: Use NULL 1G PTE mapped at 255GiB VA for ccs clear Himal Prasad Ghimiray
2023-12-11 13:43 ` [PATCH v7 06/10] drm/xe/xe2: Update chunk size for each iteration of ccs copy Himal Prasad Ghimiray
2023-12-12 12:27 ` Thomas Hellström
2023-12-11 13:43 ` [PATCH v7 07/10] drm/xe/xe2: Update emit_pte to use compression enabled PAT index Himal Prasad Ghimiray
2023-12-12 12:28 ` Thomas Hellström [this message]
2023-12-11 13:43 ` [PATCH v7 08/10] drm/xe/xe2: Handle flat ccs move for igfx Himal Prasad Ghimiray
2023-12-12 12:31 ` Thomas Hellström
2023-12-11 13:43 ` [PATCH v7 09/10] drm/xe/xe2: Modify xe_bo_test for system memory Himal Prasad Ghimiray
2023-12-11 13:43 ` [PATCH v7 10/10] drm/xe/xe2: Support flat ccs Himal Prasad Ghimiray
2023-12-12 12:33 ` Thomas Hellström
2023-12-11 14:25 ` ✓ CI.Patch_applied: success for Enable compression handling on LNL. (rev8) Patchwork
2023-12-11 14:25 ` ✗ CI.checkpatch: warning " Patchwork
2023-12-11 14:26 ` ✓ CI.KUnit: success " Patchwork
2023-12-11 14:34 ` ✓ CI.Build: " Patchwork
2023-12-11 14:34 ` ✓ CI.Hooks: " Patchwork
2023-12-11 14:35 ` ✓ CI.checksparse: " Patchwork
2023-12-11 15:10 ` ✓ CI.BAT: " Patchwork
-- strict thread matches above, loose matches on Subject: below --
2023-12-11 13:41 [PATCH v7 00/10] *Enable compression handling on LNL Himal Prasad Ghimiray
2023-12-11 13:41 ` [PATCH v7 07/10] drm/xe/xe2: Update emit_pte to use compression enabled PAT index Himal Prasad Ghimiray
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5460aee0-2dc5-77f8-5884-d71dbc1e032a@linux.intel.com \
--to=thomas.hellstrom@linux.intel.com \
--cc=himal.prasad.ghimiray@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=matthew.auld@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox