linux-toradex.git/drivers/gpu/drm/radeon/evergreen_cs.c, branch v4.9.41

drm/radeon: allow PACKET3_PFP_SYNC_ME on evergreen

2016-07-07T18:50:59+00:00

Reviewed-by: Christian König 
Signed-off-by: Edmondo Tommasina 
Signed-off-by: Alex Deucher

drm/radeon: add cayman VM support for append packet.

2016-05-05T00:30:16+00:00

This adds support for SET_APPEND_CNT packet3 to the VM paths.

Signed-off-by: Dave Airlie 
Signed-off-by: Alex Deucher

drm/radeon: add support for SET_APPEND_CNT packet3 (v2)

2016-05-05T00:19:49+00:00

This adds support to the command parser for the set append counter
packet3, this is required to support atomic counters on
evergreen/cayman GPUs.

v2: fixup some of the hardcoded numbers with real register names
(Christian)

Reviewed-by: Christian König 
Signed-off-by: Dave Airlie 
Signed-off-by: Alex Deucher

drm/radeon: fix indentation.

2016-03-16T22:08:06+00:00

I hate doing this but it hurts my eyes to go over code that does not
comply with indentation rules. Only thing that is not only space change
is in atom.c all other files are space indentation issues.

Acked-by: Christian König 
Signed-off-by: Jérôme Glisse 
Cc: Alex Deucher 
Signed-off-by: Alex Deucher

drm/radeon: remove volatile qualifier

2015-10-02T20:08:30+00:00

There doesn't seem to be any need to have 'ib' volatile, the code is
not even consistent with it and some places already miss it. As it is
now it's just making gcc produce worse code. If there are special
requirements for that memory, then proper primitives like memory
barriers or accessor functions should be used, but it doesn't look
like that is needed here.
While at it, change the type to match the one in radeon_ib structure.

Reviewed-by: Dave Airlie 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher

drm/radeon: refactor register check loop

2015-10-02T20:08:29+00:00

After this patch the register check loop does the same thing as before,
except that now gcc does better job optimizing it: it now sees that
end_reg was already checked against PACKET3_SET_CONTEXT_REG_END and can
optimize REG_SAFE_BM_SIZE comparison out of evergreen_is_safe_reg()
as (PACKET3_SET_CONTEXT_REG_END >> 7) < REG_SAFE_BM_SIZE.

Reviewed-by: Dave Airlie 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher

drm/radeon: split evergreen_cs_check_reg

2015-10-02T20:08:27+00:00

evergreen_cs_check_reg() is a large function and gcc doesn't want to
inline it. It has a quick check for reg_safe_bm[] to see if register
needs special handling, which often results in early exit. However
because the function is large, it has a long prologue/epilogue to
save/restore all the callee-save registers which according to perf is
taking significant amount of time. To avoid this, we can reuse
evergreen_is_safe_reg() to do the early check directly in register loop.

Reviewed-by: Dave Airlie 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher

drm/radeon: simplify register checker

2015-10-02T20:08:26+00:00

To avoid having to distinguish between CAYMAN or older on every register
check, place a pointer in evergreen_cs_track and use it unconditionally.
Also make use of the fact that both reg_safe_bm[] arrays are of the same
length to remove another CAYMAN check.

Reviewed-by: Dave Airlie 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher

drm/radeon: evergreen/cayman indirect draw support (v2)

2015-01-22T15:38:46+00:00

Add the necessary set of commands to support OpenGL
indirect draw calls on evergreen/cayman devices that
do not have VM.

v2: agd5f: fix warning on 32-bit

Reviewed-by: Marek Olšák 
Signed-off-by: Glenn Kennard 
Signed-off-by: Alex Deucher

drm/radeon: use pointers instead of indexes for CS chunks

2014-12-03T23:26:53+00:00

Nobody is interested at which index the chunk is. What's needed is
a pointer to the chunk. Remove unused chunk_id field as well.

Signed-off-by: Christian König 
Signed-off-by: Alex Deucher