From: "Michel Dänzer" <michel@daenzer.net>
To: dri-devel@lists.freedesktop.org, mesa-dev@lists.freedesktop.org
Subject: [PATCH] r600g, radeonsi: Inform the kernel if a BO will likely be accessed by the CPU
Date: Thu, 28 Aug 2014 15:56:01 +0900 [thread overview]
Message-ID: <1409208961-7322-2-git-send-email-michel@daenzer.net> (raw)
In-Reply-To: <1409208961-7322-1-git-send-email-michel@daenzer.net>
From: Michel Dänzer <michel.daenzer@amd.com>
This allows the kernel to prevent such BOs from ever being stored in the
CPU inaccessible part of VRAM.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
---
src/gallium/drivers/radeon/r600_buffer_common.c | 23 ++++++++++++++---------
src/gallium/winsys/radeon/drm/radeon_drm_bo.c | 8 +++++++-
src/gallium/winsys/radeon/drm/radeon_winsys.h | 3 ++-
3 files changed, 23 insertions(+), 11 deletions(-)
diff --git a/src/gallium/drivers/radeon/r600_buffer_common.c b/src/gallium/drivers/radeon/r600_buffer_common.c
index acdabc0..1a6e97d 100644
--- a/src/gallium/drivers/radeon/r600_buffer_common.c
+++ b/src/gallium/drivers/radeon/r600_buffer_common.c
@@ -126,6 +126,7 @@ bool r600_init_resource(struct r600_common_screen *rscreen,
flags = RADEON_FLAG_GTT_WC;
break;
}
+ flags = RADEON_FLAG_CPU_ACCESS;
/* fall through */
case PIPE_USAGE_DEFAULT:
case PIPE_USAGE_IMMUTABLE:
@@ -136,23 +137,27 @@ bool r600_init_resource(struct r600_common_screen *rscreen,
break;
}
- /* Use GTT for all persistent mappings with older kernels, because they
- * didn't always flush the HDP cache before CS execution.
- *
- * Write-combined CPU mappings are fine, the kernel ensures all CPU
- * writes finish before the GPU executes a command stream.
- */
- if (rscreen->info.drm_minor < 40 &&
- res->b.b.target == PIPE_BUFFER &&
+ if (res->b.b.target == PIPE_BUFFER &&
res->b.b.flags & (PIPE_RESOURCE_FLAG_MAP_PERSISTENT |
PIPE_RESOURCE_FLAG_MAP_COHERENT)) {
- res->domains = RADEON_DOMAIN_GTT;
+ /* Use GTT for all persistent mappings with older kernels,
+ * because they didn't always flush the HDP cache before CS
+ * execution.
+ *
+ * Write-combined CPU mappings are fine, the kernel ensures all CPU
+ * writes finish before the GPU executes a command stream.
+ */
+ if (rscreen->info.drm_minor < 40)
+ res->domains = RADEON_DOMAIN_GTT;
+ else if (res->domains & RADEON_DOMAIN_VRAM)
+ flags |= RADEON_FLAG_CPU_ACCESS;
}
/* Tiled textures are unmappable. Always put them in VRAM. */
if (res->b.b.target != PIPE_BUFFER &&
rtex->surface.level[0].mode >= RADEON_SURF_MODE_1D) {
res->domains = RADEON_DOMAIN_VRAM;
+ flags &= ~RADEON_FLAG_CPU_ACCESS;
}
/* Allocate a new resource. */
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_bo.c b/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
index 73f8d38..03b9b1d 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
@@ -478,7 +478,11 @@ const struct pb_vtbl radeon_bo_vtbl = {
};
#ifndef RADEON_GEM_GTT_WC
-#define RADEON_GEM_GTT_WC (1 << 2)
+#define RADEON_GEM_GTT_WC (1 << 2)
+#endif
+#ifndef RADEON_GTM_CPU_ACCESS
+/* BO is expected to be accessed by the CPU */
+#define RADEON_GEM_CPU_ACCESS (1 << 3)
#endif
static struct pb_buffer *radeon_bomgr_create_bo(struct pb_manager *_mgr,
@@ -505,6 +509,8 @@ static struct pb_buffer *radeon_bomgr_create_bo(struct pb_manager *_mgr,
if (rdesc->flags & RADEON_FLAG_GTT_WC)
args.flags |= RADEON_GEM_GTT_WC;
+ if (rdesc->flags & RADEON_FLAG_CPU_ACCESS)
+ args.flags |= RADEON_GEM_CPU_ACCESS;
if (drmCommandWriteRead(rws->fd, DRM_RADEON_GEM_CREATE,
&args, sizeof(args))) {
diff --git a/src/gallium/winsys/radeon/drm/radeon_winsys.h b/src/gallium/winsys/radeon/drm/radeon_winsys.h
index dbd58f1..69bf6ed 100644
--- a/src/gallium/winsys/radeon/drm/radeon_winsys.h
+++ b/src/gallium/winsys/radeon/drm/radeon_winsys.h
@@ -66,7 +66,8 @@ enum radeon_bo_domain { /* bitfield */
};
enum radeon_bo_flag { /* bitfield */
- RADEON_FLAG_GTT_WC = (1 << 0)
+ RADEON_FLAG_GTT_WC = (1 << 0),
+ RADEON_FLAG_CPU_ACCESS = (1 << 1),
};
enum radeon_bo_usage { /* bitfield */
--
2.1.0
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
next prev parent reply other threads:[~2014-08-28 6:56 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-08-28 6:56 [PATCH] drm/radeon: Add RADEON_GEM_CPU_ACCESS BO creation flag Michel Dänzer
2014-08-28 6:56 ` Michel Dänzer [this message]
2014-09-01 10:21 ` [Mesa-dev] [PATCH] r600g, radeonsi: Inform the kernel if a BO will likely be accessed by the CPU Marek Olšák
2014-08-28 8:57 ` [PATCH] drm/radeon: Add RADEON_GEM_CPU_ACCESS BO creation flag Christian König
2014-08-28 15:01 ` Alex Deucher
2014-08-29 1:46 ` [Mesa-dev] " Michel Dänzer
2014-09-08 17:36 ` Alex Deucher
2014-09-09 0:47 ` [Mesa-dev] " Michel Dänzer
2014-09-09 1:15 ` Michel Dänzer
2014-09-09 16:28 ` Alex Deucher
2014-09-10 4:03 ` Michel Dänzer
2014-09-10 13:40 ` Alex Deucher
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=1409208961-7322-2-git-send-email-michel@daenzer.net \
--to=michel@daenzer.net \
--cc=dri-devel@lists.freedesktop.org \
--cc=mesa-dev@lists.freedesktop.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.