nvc0: add support for indirect compute on Fermi

When indirect compute is used, the size of the grid (in blocks) is
stored as three integers inside a buffer. This requires a macro to
set up GRIDDIM_YX and GRIDDIM_Z.

Changes from v2:
 - do not launch the grid if the number of groups for a dimension is 0

Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>

3 files changed, 44 insertions, 1 deletions

diff --git a/src/gallium/drivers/nouveau/nvc0/mme/Makefile b/src/gallium/drivers/nouveau/nvc0/mme/Makefile
index 1c0f5835973..52fb0a54812 100644
--- a/src/gallium/drivers/nouveau/nvc0/mme/Makefile
+++ b/src/gallium/drivers/nouveau/nvc0/mme/Makefile
@@ -1,5 +1,5 @@
 ENVYAS?=envyas
-TARGETS=com9097.mme.h
+TARGETS=com9097.mme.h com90c0.mme.h
 
 all: $(TARGETS)
 
diff --git a/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme b/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme
new file mode 100644
index 00000000000..a3f1bdeebcd
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme
@@ -0,0 +1,24 @@
+/* NVC0_COMPUTE_MACRO_LAUNCH_GRID_INDIRECT
+ *
+ * arg     = num_groups_x
+ * parm[0] = num_groups_y
+ * parm[1] = num_groups_z
+ */
+.section #mme90c0_launch_grid_indirect
+   parm $r2 maddr 0x108e /* GRIDDIM_YX */
+   braz $r1 #fail
+   parm $r3
+   braz annul $r2 #fail
+   braz annul $r3 #fail
+   send (extrinsrt $r1 $r2 0x0 0x10 0x10) /* num_groups_y << 16 | num_groups_x */
+   send $r3
+   maddrsend 0xa7 /* COMPUTE_BEGIN */
+   maddrsend 0x282 /* UNKA08 */
+   maddr 0xda /* LAUNCH */
+   send 0x1000
+   maddrsend 0x281 /* COMPUTE_END */
+   exit maddr 0xd8 /* UNK360 */
+   send 0x1
+fail:
+   exit
+   nop
diff --git a/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme.h b/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme.h
new file mode 100644
index 00000000000..1dc06e5e690
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nvc0/mme/com90c0.mme.h
@@ -0,0 +1,19 @@
+uint32_t mme90c0_launch_grid_indirect[] = {
+	0x04238251,
+	0x00034807,
+	0x00000301,
+/* 0x000e: fail */
+	0x0002d027,
+	0x00029827,
+	0x84008842,
+	0x00001841,
+	0x0029c071,
+	0x00a08071,
+	0x00368021,
+	0x04000041,
+	0x00a04071,
+	0x003600a1,
+	0x00004041,
+	0x00000091,
+	0x00000011,
+};