Mesa (master): r600: don't emit tes samplers/views when tes isn't active

[Roland Scheidegger]

Module: Mesa
Branch: master
Commit: ea227f4322debd68380feaad1de44a2feaf3d2a9
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=ea227f4322debd68380feaad1de44a2feaf3d2a9

Author: Roland Scheidegger <sroland at vmware.com>
Date:   Wed Jan  3 03:23:13 2018 +0100

r600: don't emit tes samplers/views when tes isn't active

Similar to const buffers. The driver must not emit any tes-related state if tes
is disabled, since the hw slots are all shared by VS, therefore it would
overwrite them (the mesa state tracker might not do this, but it would be
perfectly legal to do so).
Nevertheless I think the dirty state tracking logic in the driver is
fundamentally flawed when tes is disabled/enabled, since it looks to me like
the VS (and TES) state would not get reemitted to the correct slots (if it's
not dirty anyway). Unless I'm missing something...
Theoretically, the overwrite problem could be solved by using non-overlapping
resource slots for TES and VS (since we're not even close to using half the
resource slots), but it wouldn't work for constant buffers nor samplers, and
for VS would still need to propagate changes to both LS and VS, so probably
not a useful idea.
Unfortunately there's zero coverage of this with piglit, since all tessellation
shader tests are just shader_runner tests, which are unsuitable for testing
any kind of state dependency tracking issues (so I can't even quickly hack
something up to proove it and fix it...).
TCS otoh is just fine - like GS it has its own hw slots.

Tested-by: Konstantin Kharlamov <hi-angel at yandex.ru>
Reviewed-by: Dave Airlie <airlied at redhat.com>

---

 src/gallium/drivers/r600/evergreen_state.c   |  4 ++++
 src/gallium/drivers/r600/r600_state_common.c | 15 +++++++++++++++
 2 files changed, 19 insertions(+)

diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 4cc48dfa11..fb1de9cbf4 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -2334,6 +2334,8 @@ static void evergreen_emit_tcs_sampler_views(struct r600_context *rctx, struct r
 
 static void evergreen_emit_tes_sampler_views(struct r600_context *rctx, struct r600_atom *atom)
 {
+	if (!rctx->tes_shader)
+		return;
 	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_TESS_EVAL].views,
 	                             EG_FETCH_CONSTANTS_OFFSET_VS + R600_MAX_CONST_BUFFERS, 0);
 }
@@ -2404,6 +2406,8 @@ static void evergreen_emit_tcs_sampler_states(struct r600_context *rctx, struct
 
 static void evergreen_emit_tes_sampler_states(struct r600_context *rctx, struct r600_atom *atom)
 {
+	if (!rctx->tes_shader)
+		return;
 	evergreen_emit_sampler_states(rctx, &rctx->samplers[PIPE_SHADER_TESS_EVAL], 18,
 				      R_00A414_TD_VS_SAMPLER0_BORDER_INDEX, 0);
 }
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 7f4d9f3e33..b49b05608d 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1724,6 +1724,21 @@ static bool r600_update_derived_state(struct r600_context *rctx)
 		}
 	}
 
+	/*
+	 * XXX: I believe there's some fatal flaw in the dirty state logic when
+	 * enabling/disabling tes.
+	 * VS/ES share all buffer/resource/sampler slots. If TES is enabled,
+	 * it will therefore overwrite the VS slots. If it now gets disabled,
+	 * the VS needs to rebind all buffer/resource/sampler slots - not only
+	 * has TES overwritten the corresponding slots, but when the VS was
+	 * operating as LS the things with correpsonding dirty bits got bound
+	 * to LS slots and won't reflect what is dirty as VS stage even if the
+	 * TES didn't overwrite it. The story for re-enabled TES is similar.
+	 * In any case, we're not allowed to submit any TES state when
+	 * TES is disabled (the state tracker may not do this but this looks
+	 * like an optimization to me, not something which can be relied on).
+	 */
+
 	/* Update clip misc state. */
 	if (clip_so_current) {
 		r600_update_clip_state(rctx, clip_so_current);