Skip to contentSkip to conversation

drm/i915/execlists: Move WA_TAIL_DWORDS to callee

PATCH In review Dave Gordon 3 participants 4 messages Aug 19, 2016

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v7 1/1 drm/i915/execlists: Move WA_TAIL_DWORDS to callee

Currently the execlist-specific emit-request functions start writing to the ring and reserve space for a workaround to be emitted later whilst submitting the request. It is easier to read if the caller only allocates sufficient space for its own accesses (then the reader can quickly verify that the ring begin allocates the exact space for the number of dwords emitted) and closes the access to the ring. During submit, if we need to add the workaround, we can reacquire ring access, in the assurance that we reserved space for ourselves when beginning the request.

v3:
Reinstated #define for WA_TAIL_DWORDS so we could accommodate GPUs that required different amounts of padding, generalised NOOP fill [Rodrigo Vivi], added W/A space to reserved amount and updated code comments [Dave Gordon],

v4:
Made #define for WA_TAIL_DWORDS a function-type macro in case we want different values on different platforms (or engines), to address comment by [Rodrigo Vivi]. But the current value is still the constant (2) on all platforms, so this doesn't add any overhead.

v5:
Eliminated local variable & multiple indirections. Added long essay comment about WaIdleLiteRestore.

v6:
Renamed #define for amount of space used by add_request() - it's the *maximum* length of opcodes emitted by that function, not a minimum, and made it (hypothetically) platform- or engine-dependent. But it's still the same literal-constant on all platforms for now.

v7:
Rebased

Cc Rodrigo Vivi <[email protected]>
Cc Chris Wilson <[email protected]>
---
 drivers/gpu/drm/i915/i915_gem_request.c |  2 +-
 drivers/gpu/drm/i915/intel_lrc.c        | 92 ++++++++++++++++++++++++---------
 drivers/gpu/drm/i915/intel_ringbuffer.h | 19 ++++---
 3 files changed, 81 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/i915_gem_request.c +1 -1
index 1a21532..f5b38dd 100644
@@ -411,7 +411,7 @@ struct drm_i915_gem_request *
 	 * to be redone if the request is not actually submitted straight
 	 * away, e.g. because a GPU scheduler has deferred it.
 	 */
-	req->reserved_space = MIN_SPACE_FOR_ADD_REQUEST;
+	req->reserved_space = sizeof(u32) * MAX_ADD_REQUEST_DWORDS(req);
 
 	if (i915.enable_execlists)
 		ret = intel_logical_ring_alloc_request_extras(req);
drivers/gpu/drm/i915/intel_lrc.c +67 -25

146 lines not on this page · raw message

drivers/gpu/drm/i915/intel_ringbuffer.h +13 -6
index 86612d5..9c4175c 100644
@@ -509,13 +509,20 @@ static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine)
 int init_workarounds_ring(struct intel_engine_cs *engine);
 
 /*
- * Arbitrary size for largest possible 'add request' sequence. The code paths
- * are complex and variable. Empirical measurement shows that the worst case
- * is BDW at 192 bytes (6 + 6 + 36 dwords), then ILK at 136 bytes. However,
- * we need to allocate double the largest single packet within that emission
- * to account for tail wraparound (so 6 + 6 + 72 dwords for BDW).
+ * Size for the longest possible 'add_request' sequence, for the given request.
+ *
+ * In theory, this could depend on the platform, or the engine, or many other
+ * things; hence the macro takes the request as an argument on which to base
+ * this calculation.  In practive, the code paths are complex and variable,
+ * and it's simpler just to give an upper bound for all possibilities,
+ * ignoring the details of the actual request.
+ *
+ * Empirical measurement shows that the worst case is BDW at 192 bytes
+ * (6 + 6 + 36 dwords), then ILK at 136 bytes. However, we need to allocate
+ * double the largest single packet within that emissionto account for tail
+ * wraparound (so 6 + 6 + 72 dwords for BDW).
  */
-#define MIN_SPACE_FOR_ADD_REQUEST 336
+#define MAX_ADD_REQUEST_DWORDS(request)	(6+6+72)
 
 static inline u32 intel_hws_seqno_address(struct intel_engine_cs *engine)
 {
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

drivers/gpu/drm/i915/intel_lrc.c int intel_logical_ring_alloc_request_extras(struct drm_i915_gem_request *request · line 689

Quoted 11 lines · Dave Gordon
drivers/gpu/drm/i915/intel_lrc.c +7 -0
@@ -654,6 +680,14 @@ int intel_logical_ring_alloc_request_extras(struct drm_i915_gem_request *request
 	 */
 	request->reserved_space += EXECLISTS_REQUEST_SIZE;
 
+	/*
+	 * WA_TAIL_DWORDS is specific to the execlist submission mechanism,
+	 * to accommodate some NOOPs at the end of each request, to be used
+	 * by a workaround for not being allowed to do lite restore with
+	 * HEAD==TAIL (WaIdleLiteRestore). See intel_logical_ring_submit()
+	 */
+	request->reserved_space += sizeof(u32) * WA_TAIL_DWORDS(request);

We already have the define that accommodates the tail. Whilst this remains a fixed size, let's use it appropriately. And when it is dynamic, we store it in the engine (or context).
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
Quoted 5 lines · Chris Wilson
Quoted 4 lines · Dave Gordon
drivers/gpu/drm/i915/intel_lrc.c +0 -0
@@ -654,6 +680,14 @@ int intel_logical_ring_alloc_request_extras(struct drm_i915_gem_request *request
 	 */
 	request->reserved_space += EXECLISTS_REQUEST_SIZE;
 
Quoted 12 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_logical_ring_alloc_request_extras(struct drm_i915_gem_request *request · line 689

Quoted 7 lines
drivers/gpu/drm/i915/intel_lrc.c +7 -0
+	/*
+	 * WA_TAIL_DWORDS is specific to the execlist submission mechanism,
+	 * to accommodate some NOOPs at the end of each request, to be used
+	 * by a workaround for not being allowed to do lite restore with
+	 * HEAD==TAIL (WaIdleLiteRestore). See intel_logical_ring_submit()
+	 */
+	request->reserved_space += sizeof(u32) * WA_TAIL_DWORDS(request);

We already have the define that accommodates the tail. Whilst this remains a fixed size, let's use it appropriately. And when it is dynamic, we store it in the engine (or context).
-Chris

It might not depend on the engine or context, but on the specifics of the request. For the other cases, we could find the engine or context from the request but not vice-versa. Hence we should (pretend to) pass the request here, because that provides the most generality.

We do have a future use for this, so this is really part of the prepwork. And Rodrigo specifically asked for the possibility of variable amounts of padding, as documented in the commit message.

But the main point of this patch is to get rid of the "+WA_TAIL_DWORDS" in the *_emit_request() functions and instead claim-and-use the space inside intel_logical_ring_advance(), thus keeping all begin()/advance() pairs matched and local, as opposed to the current scheme of the space being claimed in _emit_request() but *used* by _logical_ring_advance().

.Dave.
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw

✗ Ro.CI.BAT: failure for drm/i915/execlists: Move WA_TAIL_DWORDS to callee (rev3)

In reply to Dave Gordon
== Series Details ==

== Series Details ==

Series: drm/i915/execlists: Move WA_TAIL_DWORDS to callee (rev3)
URL : https://patchwork.freedesktop.org/series/3133/
State : failure

== Summary ==

Series 3133v3 drm/i915/execlists: Move WA_TAIL_DWORDS to callee
http://patchwork.freedesktop.org/api/1.0/series/3133/revisions/3/mbox

Test kms_cursor_legacy:

        Subgroup basic-flip-vs-cursor-legacy:
                pass       -> FAIL       (ro-byt-n2820)
                fail       -> PASS       (ro-bdw-i5-5250u)
        Subgroup basic-flip-vs-cursor-varying-size:
                pass       -> FAIL       (ro-byt-n2820)
                fail       -> PASS       (fi-hsw-i7-4770k)
                fail       -> PASS       (ro-bdw-i5-5250u)
Test kms_pipe_crc_basic:
        Subgroup suspend-read-crc-pipe-a:
                dmesg-warn -> SKIP       (ro-bdw-i5-5250u)
        Subgroup suspend-read-crc-pipe-b:
                dmesg-warn -> PASS       (ro-bdw-i7-5600u)
fi-hsw-i7-4770k  total:244  pass:222  dwarn:0   dfail:0   fail:0   skip:22 
fi-kbl-qkkr      total:244  pass:186  dwarn:28  dfail:0   fail:3   skip:27 
fi-skl-i7-6700k  total:244  pass:208  dwarn:4   dfail:2   fail:2   skip:28 
fi-snb-i7-2600   total:244  pass:202  dwarn:0   dfail:0   fail:0   skip:42 
ro-bdw-i5-5250u  total:240  pass:220  dwarn:1   dfail:0   fail:0   skip:19 
ro-bdw-i7-5557U  total:240  pass:219  dwarn:2   dfail:0   fail:1   skip:18 
ro-bdw-i7-5600u  total:240  pass:206  dwarn:0   dfail:0   fail:2   skip:32 
ro-bsw-n3050     total:240  pass:195  dwarn:0   dfail:0   fail:3   skip:42 
ro-byt-n2820     total:240  pass:196  dwarn:0   dfail:0   fail:4   skip:40 
ro-hsw-i3-4010u  total:240  pass:213  dwarn:0   dfail:0   fail:1   skip:26 
ro-hsw-i7-4770r  total:240  pass:185  dwarn:0   dfail:0   fail:0   skip:55 
ro-ilk1-i5-650   total:235  pass:174  dwarn:0   dfail:0   fail:2   skip:59 
ro-ivb-i7-3770   total:240  pass:204  dwarn:0   dfail:0   fail:1   skip:35 
ro-ivb2-i7-3770  total:240  pass:208  dwarn:0   dfail:0   fail:1   skip:31 
ro-skl3-i5-6260u total:240  pass:223  dwarn:0   dfail:0   fail:3   skip:14 

Results at /archive/results/CI_IGT_test/RO_Patchwork_1942/

21defee drm-intel-nightly: 2016y-08m-19d-09h-06m-36s UTC integration manifest fc5328f drm/i915/execlists: Move WA_TAIL_DWORDS to callee

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx