Skip to contentSkip to conversation

drm/i915: Add GPGPU_THREADS_DISPATCHED to the register whitelist

PATCH Applied Jordan Justen 3 participants 5 messages Dec 11, 2014 – Dec 16, 2014

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

PATCH 1/1 drm/i915: Add GPGPU_THREADS_DISPATCHED to the register whitelist

This will allow us to read the number of dispatched compute threads for GL_ARB_pipeline_statistics_query.

---
 drivers/gpu/drm/i915/i915_cmd_parser.c |  4 +++-
 drivers/gpu/drm/i915/i915_reg.h        | 23 ++++++++++++-----------
 2 files changed, 15 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_cmd_parser.c +3 -1
index 6e9eac4..82679c9 100644
@@ -401,6 +401,7 @@ static const struct drm_i915_cmd_table hsw_blt_ring_cmds[] = {
 #define REG64(addr) (addr), (addr + sizeof(u32))
 
 static const u32 gen7_render_regs[] = {
+	REG64(GPGPU_THREADS_DISPATCHED),
 	REG64(HS_INVOCATION_COUNT),
 	REG64(DS_INVOCATION_COUNT),
 	REG64(IA_VERTICES_COUNT),
@@ -1076,6 +1077,7 @@ int i915_cmd_parser_get_version(void)
 	 *    hardware parsing enabled (so does not allow new use cases).
 	 * 2. Allow access to the MI_PREDICATE_SRC0 and
 	 *    MI_PREDICATE_SRC1 registers.
+	 * 3. Allow access to the GPGPU_THREADS_DISPATCHED register.
 	 */
-	return 2;
+	return 3;
 }
drivers/gpu/drm/i915/i915_reg.h +12 -11
index 869e5ae..0e6e694 100644
@@ -460,17 +460,18 @@
  */
 #define BCS_SWCTRL 0x22200
 
-#define HS_INVOCATION_COUNT 0x2300
-#define DS_INVOCATION_COUNT 0x2308
-#define IA_VERTICES_COUNT   0x2310
-#define IA_PRIMITIVES_COUNT 0x2318
-#define VS_INVOCATION_COUNT 0x2320
-#define GS_INVOCATION_COUNT 0x2328
-#define GS_PRIMITIVES_COUNT 0x2330
-#define CL_INVOCATION_COUNT 0x2338
-#define CL_PRIMITIVES_COUNT 0x2340
-#define PS_INVOCATION_COUNT 0x2348
-#define PS_DEPTH_COUNT      0x2350
+#define GPGPU_THREADS_DISPATCHED        0x2290
+#define HS_INVOCATION_COUNT             0x2300
+#define DS_INVOCATION_COUNT             0x2308
+#define IA_VERTICES_COUNT               0x2310
+#define IA_PRIMITIVES_COUNT             0x2318
+#define VS_INVOCATION_COUNT             0x2320
+#define GS_INVOCATION_COUNT             0x2328
+#define GS_PRIMITIVES_COUNT             0x2330
+#define CL_INVOCATION_COUNT             0x2338
+#define CL_PRIMITIVES_COUNT             0x2340
+#define PS_INVOCATION_COUNT             0x2348
+#define PS_DEPTH_COUNT                  0x2350
 
 /* There are the 4 64-bit counter registers, one for each stream output */
 #define GEN7_SO_NUM_PRIMS_WRITTEN(n) (0x5200 + (n) * 8)
Signature · 6 lines
2.1.3

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
Quoted 3 lines · Ben Widawsky
Quoted 2 lines · Jordan Justen

This will allow us to read the number of dispatched compute threads for GL_ARB_pipeline_statistics_query.

Just so we have all the formally required pieces: Can you please supply a link to the mesa code for this? I'm a bit behind on reading mesa-dev.

Quoted 3 lines
Quoted 2 lines

Queued for -next, thanks for the patch.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw
Quoted 7 lines · Daniel Vetter
Quoted 3 lines · Ben Widawsky
Quoted 2 lines · Jordan Justen

This will allow us to read the number of dispatched compute threads for GL_ARB_pipeline_statistics_query.

Just so we have all the formally required pieces: Can you please supply a link to the mesa code for this? I'm a bit behind on reading mesa-dev.

This is an early hacked version of mesa that uses the register:
http://cgit.freedesktop.org/~jljusten/mesa/log/?h=cs%2bps

I veried reading the register with the modified mesa and a piglit test. I first tested with the cmd parser disabled, and then with this patch.

I'm waiting for the initial GL_ARB_pipeline_statistics_query extension (without CS support) to make it upstream in mesa before finishing and sending out the CS support.

If this is in next, can I proceed to upstream a mesa patch that depends on version 3 of the command parser? Or, should I wait until this makes it fully upstream?

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

#4 lore raw
Quoted 16 lines · Jordan Justen
Quoted 7 lines · Daniel Vetter
Quoted 3 lines · Ben Widawsky
Quoted 2 lines · Jordan Justen

This will allow us to read the number of dispatched compute threads for GL_ARB_pipeline_statistics_query.

Just so we have all the formally required pieces: Can you please supply a link to the mesa code for this? I'm a bit behind on reading mesa-dev.

This is an early hacked version of mesa that uses the register:
http://cgit.freedesktop.org/~jljusten/mesa/log/?h=cs%2bps

I veried reading the register with the modified mesa and a piglit test. I first tested with the cmd parser disabled, and then with this patch.

Quoted 7 lines

I'm waiting for the initial GL_ARB_pipeline_statistics_query extension (without CS support) to make it upstream in mesa before finishing and sending out the CS support.

If this is in next, can I proceed to upstream a mesa patch that depends on version 3 of the command parser? Or, should I wait until this makes it fully upstream?

If we go super strict with rule abi is locked down when I tag drm-intel-next at the end of this weeek. But that's hair-splitting. So yeah, as soon as all the mesa depencies have landed you can just go ahead with your patches.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx