r600g: implement buffer copying using CP DMA for R7xx, Evergreen, Cayman

R6xx doesn't work - the issue seems to be with flushing (sometimes
the destination buffer contains garbage). There are no hangs, so we're good.

R7xx doesn't seem to have any alignment restriction despite our initial
thinking. Everything just works.

Reviewed-by: Alex Deucher <[email protected]>

1 files changed, 34 insertions, 0 deletions

diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h
index 69bfd7a2f87..dd64aca3d51 100644
--- a/src/gallium/drivers/r600/r600d.h
+++ b/src/gallium/drivers/r600/r600d.h
@@ -159,6 +159,40 @@
 #define PKT3_PRED_S(x)               (((x) >> 0) & 0x1)
 #define PKT0(index, count) (PKT_TYPE_S(0) | PKT0_BASE_INDEX_S(index) | PKT_COUNT_S(count))
 
+#define PKT3_CP_DMA					0x41
+/* 1. header
+ * 2. SRC_ADDR_LO [31:0]
+ * 3. CP_SYNC [31] | SRC_ADDR_HI [7:0]
+ * 4. DST_ADDR_LO [31:0]
+ * 5. DST_ADDR_HI [7:0]
+ * 6. COMMAND [29:22] | BYTE_COUNT [20:0]
+ */
+#define PKT3_CP_DMA_CP_SYNC       (1 << 31)
+/* COMMAND */
+#define PKT3_CP_DMA_CMD_SRC_SWAP(x) ((x) << 23)
+/* 0 - none
+ * 1 - 8 in 16
+ * 2 - 8 in 32
+ * 3 - 8 in 64
+ */
+#define PKT3_CP_DMA_CMD_DST_SWAP(x) ((x) << 24)
+/* 0 - none
+ * 1 - 8 in 16
+ * 2 - 8 in 32
+ * 3 - 8 in 64
+ */
+#define PKT3_CP_DMA_CMD_SAS       (1 << 26)
+/* 0 - memory
+ * 1 - register
+ */
+#define PKT3_CP_DMA_CMD_DAS       (1 << 27)
+/* 0 - memory
+ * 1 - register
+ */
+#define PKT3_CP_DMA_CMD_SAIC      (1 << 28)
+#define PKT3_CP_DMA_CMD_DAIC      (1 << 29)
+
+
 /* Registers */
 #define R_008490_CP_STRMOUT_CNTL		     0x008490
 #define   S_008490_OFFSET_UPDATE_DONE(x)		(((x) & 0x1) << 0)