From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++) + for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0; diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15
#define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF
/* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22)
@@ -82,14 +73,4 @@
#define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING)
-/* - * Access flags to be used for CPU- and GPU- mappings. - * The idea is that the TTM synchronization mechanism will - * allow concurrent READ access and exclusive write access. - * Currently GPU- and CPU accesses are exclusive. - */ - -#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1) - #endif
From: Christian König christian.koenig@amd.com
Makes more sense to keep that together.
Signed-off-by: Christian König christian.koenig@amd.com --- include/drm/ttm/ttm_bo_api.h | 32 +------------------------------- include/drm/ttm/ttm_placement.h | 35 +++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+), 31 deletions(-)
diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h index 97aaf5c..d73c7c2 100644 --- a/include/drm/ttm/ttm_bo_api.h +++ b/include/drm/ttm/ttm_bo_api.h @@ -45,37 +45,7 @@ struct ttm_bo_device;
struct drm_mm_node;
-/** - * struct ttm_place - * - * @fpfn: first valid page frame number to put the object - * @lpfn: last valid page frame number to put the object - * @flags: memory domain and caching flags for the object - * - * Structure indicating a possible place to put an object. - */ -struct ttm_place { - unsigned fpfn; - unsigned lpfn; - uint32_t flags; -}; - -/** - * struct ttm_placement - * - * @num_placement: number of preferred placements - * @placement: preferred placements - * @num_busy_placement: number of preferred placements when need to evict buffer - * @busy_placement: preferred placements when need to evict buffer - * - * Structure indicating the placement you request for an object. - */ -struct ttm_placement { - unsigned num_placement; - const struct ttm_place *placement; - unsigned num_busy_placement; - const struct ttm_place *busy_placement; -}; +struct ttm_placement;
/** * struct ttm_bus_placement diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 20219d9..ff5195c 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -30,6 +30,9 @@
#ifndef _TTM_PLACEMENT_H_ #define _TTM_PLACEMENT_H_ + +#include <linux/types.h> + /* * Memory regions for data placement. */ @@ -73,4 +76,36 @@
#define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING)
+/** + * struct ttm_place + * + * @fpfn: first valid page frame number to put the object + * @lpfn: last valid page frame number to put the object + * @flags: memory domain and caching flags for the object + * + * Structure indicating a possible place to put an object. + */ +struct ttm_place { + unsigned fpfn; + unsigned lpfn; + uint32_t flags; +}; + +/** + * struct ttm_placement + * + * @num_placement: number of preferred placements + * @placement: preferred placements + * @num_busy_placement: number of preferred placements when need to evict buffer + * @busy_placement: preferred placements when need to evict buffer + * + * Structure indicating the placement you request for an object. + */ +struct ttm_placement { + unsigned num_placement; + const struct ttm_place *placement; + unsigned num_busy_placement; + const struct ttm_place *busy_placement; +}; + #endif
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Makes more sense to keep that together.
Signed-off-by: Christian König christian.koenig@amd.com
Reviewed-by: Chunming Zhou david1.zhou@amd.com
include/drm/ttm/ttm_bo_api.h | 32 +------------------------------- include/drm/ttm/ttm_placement.h | 35 +++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+), 31 deletions(-)
diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h index 97aaf5c..d73c7c2 100644 --- a/include/drm/ttm/ttm_bo_api.h +++ b/include/drm/ttm/ttm_bo_api.h @@ -45,37 +45,7 @@ struct ttm_bo_device;
struct drm_mm_node;
-/**
- struct ttm_place
- @fpfn: first valid page frame number to put the object
- @lpfn: last valid page frame number to put the object
- @flags: memory domain and caching flags for the object
- Structure indicating a possible place to put an object.
- */
-struct ttm_place {
- unsigned fpfn;
- unsigned lpfn;
- uint32_t flags;
-};
-/**
- struct ttm_placement
- @num_placement: number of preferred placements
- @placement: preferred placements
- @num_busy_placement: number of preferred placements when need to evict buffer
- @busy_placement: preferred placements when need to evict buffer
- Structure indicating the placement you request for an object.
- */
-struct ttm_placement {
- unsigned num_placement;
- const struct ttm_place *placement;
- unsigned num_busy_placement;
- const struct ttm_place *busy_placement;
-}; +struct ttm_placement;
/**
- struct ttm_bus_placement
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 20219d9..ff5195c 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -30,6 +30,9 @@
#ifndef _TTM_PLACEMENT_H_ #define _TTM_PLACEMENT_H_
+#include <linux/types.h>
- /*
*/
- Memory regions for data placement.
@@ -73,4 +76,36 @@
#define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING)
+/**
- struct ttm_place
- @fpfn: first valid page frame number to put the object
- @lpfn: last valid page frame number to put the object
- @flags: memory domain and caching flags for the object
- Structure indicating a possible place to put an object.
- */
+struct ttm_place {
- unsigned fpfn;
- unsigned lpfn;
- uint32_t flags;
+};
+/**
- struct ttm_placement
- @num_placement: number of preferred placements
- @placement: preferred placements
- @num_busy_placement: number of preferred placements when need to evict buffer
- @busy_placement: preferred placements when need to evict buffer
- Structure indicating the placement you request for an object.
- */
+struct ttm_placement {
- unsigned num_placement;
- const struct ttm_place *placement;
- unsigned num_busy_placement;
- const struct ttm_place *busy_placement;
+};
- #endif
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15
#define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF
/* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22)
@@ -82,14 +73,4 @@
#define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING)
-/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING) -/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
#endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
In this case please just add them back in your tree. That should be a two liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING) -/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Chris,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement. There is no guarantee that amdgpu would never introduce new memory domain in future. In such case, I'd like to vote for keeping these flags instead of adding them back when amdgpu need to add new memory domain.
As you mentioned that it just a two liner. And there is actually no functionality break with these flags. Then how about keep these flags?
Regards, Hawking
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of Christian K?nig Sent: Friday, September 09, 2016 17:07 To: Cui, Flora Flora.Cui@amd.com; Zhou, David(ChunMing) David1.Zhou@amd.com Cc: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
In this case please just add them back in your tree. That should be a two liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING) -/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Chris,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement. There is no guarantee that amdgpu would never introduce new memory domain in future. In such case, I'd like to vote for keeping these flags instead of adding them back when amdgpu need to add new memory domain.
As you mentioned that it just a two liner. And there is actually no functionality break with these flags. Then how about keep these flags?
Regards, Hawking
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of Christian K?nig Sent: Friday, September 09, 2016 17:07 To: Cui, Flora Flora.Cui@amd.com; Zhou, David(ChunMing) David1.Zhou@amd.com Cc: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
In this case please just add them back in your tree. That should be a two liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING) -/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement.
And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory domain in future.
Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Regards, Christian.
Am 09.09.2016 um 12:35 schrieb Zhang, Hawking:
Hi Chris,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement. There is no guarantee that amdgpu would never introduce new memory domain in future. In such case, I'd like to vote for keeping these flags instead of adding them back when amdgpu need to add new memory domain.
As you mentioned that it just a two liner. And there is actually no functionality break with these flags. Then how about keep these flags?
Regards, Hawking
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of Christian K?nig Sent: Friday, September 09, 2016 17:07 To: Cui, Flora Flora.Cui@amd.com; Zhou, David(ChunMing) David1.Zhou@amd.com Cc: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
In this case please just add them back in your tree. That should be a two liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, { int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM | TTM_PL_MASK_CACHING) -/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx _______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of Christian König Sent: Friday, September 09, 2016 7:24 AM To: Zhang, Hawking; Koenig, Christian; Cui, Flora; Zhou, David(ChunMing); amd-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the
corresponding placement and thus have impact on mem region create and bo movement. And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory
domain in future. Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
We need it for the hybrid driver in the near feature and the open driver may use it in the future depending on the use cases. Removing it just makes our lives more difficult for supporting dkms and distro integration for very minimal gain.
Alex
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Regards, Christian.
Am 09.09.2016 um 12:35 schrieb Zhang, Hawking:
Hi Chris,
Removing the flag will make ttm_mem_type_from_place skip counting the
corresponding placement and thus have impact on mem region create and bo movement. There is no guarantee that amdgpu would never introduce new memory domain in future. In such case, I'd like to vote for keeping these flags instead of adding them back when amdgpu need to add new memory domain.
As you mentioned that it just a two liner. And there is actually no
functionality break with these flags. Then how about keep these flags?
Regards, Hawking
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf
Of Christian K?nig
Sent: Friday, September 09, 2016 17:07 To: Cui, Flora Flora.Cui@amd.com; Zhou, David(ChunMing)
Cc: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
In this case please just add them back in your tree. That should be a two
liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const
struct ttm_place *place,
{ int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM |
TTM_PL_MASK_CACHING)
-/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx _______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Am 09.09.2016 um 15:41 schrieb Deucher, Alexander:
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of Christian König Sent: Friday, September 09, 2016 7:24 AM To: Zhang, Hawking; Koenig, Christian; Cui, Flora; Zhou, David(ChunMing); amd-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the
corresponding placement and thus have impact on mem region create and bo movement. And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory
domain in future. Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
We need it for the hybrid driver in the near feature and the open driver may use it in the future depending on the use cases. Removing it just makes our lives more difficult for supporting dkms and distro integration for very minimal gain.
Yeah, agree. The problem is I already did so without knowing that.
I will send out a V2 only removing the stuff we explicitly don't need.
Hopefully nobody will notice and question further, Christian.
Alex
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Regards, Christian.
Am 09.09.2016 um 12:35 schrieb Zhang, Hawking:
Hi Chris,
Removing the flag will make ttm_mem_type_from_place skip counting the
corresponding placement and thus have impact on mem region create and bo movement. There is no guarantee that amdgpu would never introduce new memory domain in future. In such case, I'd like to vote for keeping these flags instead of adding them back when amdgpu need to add new memory domain.
As you mentioned that it just a two liner. And there is actually no
functionality break with these flags. Then how about keep these flags?
Regards, Hawking
-----Original Message----- From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf
Of Christian K?nig
Sent: Friday, September 09, 2016 17:07 To: Cui, Flora Flora.Cui@amd.com; Zhou, David(ChunMing)
Cc: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: Re: [PATCH 1/2] drm/ttm: remove unused placement flags
In this case please just add them back in your tree. That should be a two
liner.
For upstream it certainly doesn't make sense to keep them.
Regards, Christian.
Am 09.09.2016 um 09:01 schrieb Flora Cui:
yes. please don't do this, I need them.
On Fri, Sep 09, 2016 at 02:41:16PM +0800, zhoucm1 wrote:
On 2016年09月08日 21:41, Christian König wrote:
From: Christian König christian.koenig@amd.com
Either never used or not used in quite a while.
No, I remember Flora's Direct GMA is using them like GDS use PRIV0-2. And you cannot make sure there isn't no one using them in other closed projects, right? If you removed now, that obviously will break her implementation and brings her many troubles.
Regards, David Zhou
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/ttm/ttm_bo.c | 2 +- include/drm/ttm/ttm_placement.h | 19 ------------------- 2 files changed, 1 insertion(+), 20 deletions(-)diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index bc37f02..4d2e8f2 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -59,7 +59,7 @@ static inline int ttm_mem_type_from_place(const
struct ttm_place *place,
{ int i;
- for (i = 0; i <= TTM_PL_PRIV5; i++)
- for (i = 0; i <= TTM_PL_PRIV2; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0;
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h index 8ed44f9..20219d9 100644 --- a/include/drm/ttm/ttm_placement.h +++ b/include/drm/ttm/ttm_placement.h @@ -40,10 +40,6 @@ #define TTM_PL_PRIV0 3 #define TTM_PL_PRIV1 4 #define TTM_PL_PRIV2 5 -#define TTM_PL_PRIV3 6 -#define TTM_PL_PRIV4 7 -#define TTM_PL_PRIV5 8 -#define TTM_PL_SWAPPED 15 #define TTM_PL_FLAG_SYSTEM (1 << TTM_PL_SYSTEM) #define TTM_PL_FLAG_TT (1 << TTM_PL_TT) @@ -51,10 +47,6 @@ #define TTM_PL_FLAG_PRIV0 (1 << TTM_PL_PRIV0) #define TTM_PL_FLAG_PRIV1 (1 << TTM_PL_PRIV1) #define TTM_PL_FLAG_PRIV2 (1 << TTM_PL_PRIV2) -#define TTM_PL_FLAG_PRIV3 (1 << TTM_PL_PRIV3) -#define TTM_PL_FLAG_PRIV4 (1 << TTM_PL_PRIV4) -#define TTM_PL_FLAG_PRIV5 (1 << TTM_PL_PRIV5) -#define TTM_PL_FLAG_SWAPPED (1 << TTM_PL_SWAPPED) #define TTM_PL_MASK_MEM 0x0000FFFF /* @@ -72,7 +64,6 @@ #define TTM_PL_FLAG_CACHED (1 << 16) #define TTM_PL_FLAG_UNCACHED (1 << 17) #define TTM_PL_FLAG_WC (1 << 18) -#define TTM_PL_FLAG_SHARED (1 << 20) #define TTM_PL_FLAG_NO_EVICT (1 << 21) #define TTM_PL_FLAG_TOPDOWN (1 << 22) @@ -82,14 +73,4 @@ #define TTM_PL_MASK_MEMTYPE (TTM_PL_MASK_MEM |
TTM_PL_MASK_CACHING)
-/*
- Access flags to be used for CPU- and GPU- mappings.
- The idea is that the TTM synchronization mechanism will
- allow concurrent READ access and exclusive write access.
- Currently GPU- and CPU accesses are exclusive.
- */
-#define TTM_ACCESS_READ (1 << 0) -#define TTM_ACCESS_WRITE (1 << 1)
- #endif
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx _______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
On 9 September 2016 at 12:24, Christian König deathsimple@vodafone.de wrote:
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement.
And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory domain in future.
Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Fwiw I second Christian here. If they are unused in open-source drivers there's no reason to keep them. If/as that changes the (newly introduced) user can add back the relevant code.
If closed-source driver(s) use them, then they can keep it as part of their blob. Upstream kernel does not cater for closed-source drivers, period. I realise that's not the answer some are hoping for, so if you want to question it take it up with Linus and co.
Regards, Emil
Am 09.09.2016 um 15:54 schrieb Emil Velikov:
On 9 September 2016 at 12:24, Christian König deathsimple@vodafone.de wrote:
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement.
And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory domain in future.
Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Fwiw I second Christian here. If they are unused in open-source drivers there's no reason to keep them. If/as that changes the (newly introduced) user can add back the relevant code.
Crap to late :( I was about to send a V2 of the patch to keep the PRIV flags.
If closed-source driver(s) use them, then they can keep it as part of their blob. Upstream kernel does not cater for closed-source drivers, period. I realise that's not the answer some are hoping for, so if you want to question it take it up with Linus and co.
It's not an issue between closed vs. open, but rather additional work of rebasing the open code when we start to use additional domains.
But on the other hand I still haven't seen a good reason for using those. As far as I know we have covered all resource in the current and next hardware generation with the existing flags.
Regards, Christian.
Regards, Emil
On 9 September 2016 at 15:30, Christian König deathsimple@vodafone.de wrote:
Am 09.09.2016 um 15:54 schrieb Emil Velikov:
On 9 September 2016 at 12:24, Christian König deathsimple@vodafone.de wrote:
Hi Hawking,
Removing the flag will make ttm_mem_type_from_place skip counting the corresponding placement and thus have impact on mem region create and bo movement.
And that is exactly the reason why I want to remove the unused flags.
There is no guarantee that amdgpu would never introduce new memory domain in future.
Irrelevant, if any driver wants to use additional domains it should add them when they are used.
BTW: Why would we want to add another TTM domain? I really don't see any need for that.
Then how about keep these flags?
Actually we used to have automated scanners which complain about unused code. I'm wondering why they don't detected that earlier.
Anyway any code which isn't used in a while should be removed.
Fwiw I second Christian here. If they are unused in open-source drivers there's no reason to keep them. If/as that changes the (newly introduced) user can add back the relevant code.
Crap to late :( I was about to send a V2 of the patch to keep the PRIV flags.
Oops, sorry :-)
If closed-source driver(s) use them, then they can keep it as part of their blob. Upstream kernel does not cater for closed-source drivers, period. I realise that's not the answer some are hoping for, so if you want to question it take it up with Linus and co.
It's not an issue between closed vs. open, but rather additional work of rebasing the open code when we start to use additional domains.
This should serve as a greater initiative to develop and upstream things in a gradual manner, no ? We all get carried away sometimes creating a massive branch which just cannot go in at once.
But on the other hand I still haven't seen a good reason for using those. As far as I know we have covered all resource in the current and next hardware generation with the existing flags.
This in itself is a pretty good point as well, considering you know the hardware fairly well and you've worked in the kernel for quite a while now.
Regards, Emil
dri-devel@lists.freedesktop.org
-
Christian König -
Christian König -
Deucher, Alexander -
Emil Velikov -
Flora Cui -
Zhang, Hawking -
zhoucm1