Mesa (main): radeonsi/gfx11: synchronize correctly before setting SPI_ATTRIBUTE_RING_*

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 56d4e0be86b50c5f6bddeef22c92332b8549d833
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=56d4e0be86b50c5f6bddeef22c92332b8549d833

Author: Marek Olšák <marek.olsak at amd.com>
Date:   Thu May 19 20:30:33 2022 -0400

radeonsi/gfx11: synchronize correctly before setting SPI_ATTRIBUTE_RING_*

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer at amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16990>

---

 src/gallium/drivers/radeonsi/si_state.c | 32 ++++++++++++++++++++++++++------
 1 file changed, 26 insertions(+), 6 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index ddbf393f87d..fb31de37ac3 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -5900,12 +5900,32 @@ void si_init_cs_preamble_state(struct si_context *sctx, bool uses_reg_shadowing)
       si_pm4_set_reg(pm4, R_028620_PA_RATE_CNTL,
                      S_028620_VERTEX_RATE(2) | S_028620_PRIM_RATE(1));
 
-      /* We must wait for idle before changing the SPI attribute ring registers. */
-      /* TODO: Find a more reliable way to wait for idle. */
-      for (unsigned i = 0; i < 4; i++) {
-         si_pm4_cmd_add(pm4, PKT3(PKT3_EVENT_WRITE, 0, 0));
-         si_pm4_cmd_add(pm4, EVENT_TYPE(V_028A90_PS_PARTIAL_FLUSH) | EVENT_INDEX(4));
-      }
+      /* We must wait for idle using an EOP event before changing the attribute ring registers.
+       * Use the bottom-of-pipe EOP event, but increment the PWS counter instead of writing memory.
+       */
+      si_pm4_cmd_add(pm4, PKT3(PKT3_RELEASE_MEM, 6, 0));
+      si_pm4_cmd_add(pm4, S_490_EVENT_TYPE(V_028A90_BOTTOM_OF_PIPE_TS) |
+                          S_490_EVENT_INDEX(5) |
+                          S_490_PWS_ENABLE(1));
+      si_pm4_cmd_add(pm4, 0); /* DST_SEL, INT_SEL, DATA_SEL */
+      si_pm4_cmd_add(pm4, 0); /* ADDRESS_LO */
+      si_pm4_cmd_add(pm4, 0); /* ADDRESS_HI */
+      si_pm4_cmd_add(pm4, 0); /* DATA_LO */
+      si_pm4_cmd_add(pm4, 0); /* DATA_HI */
+      si_pm4_cmd_add(pm4, 0); /* INT_CTXID */
+
+      /* Wait for the PWS counter. */
+      si_pm4_cmd_add(pm4, PKT3(PKT3_ACQUIRE_MEM, 6, 0));
+      si_pm4_cmd_add(pm4, S_580_PWS_STAGE_SEL(V_580_CP_ME) |
+                          S_580_PWS_COUNTER_SEL(V_580_TS_SELECT) |
+                          S_580_PWS_ENA2(1) |
+                          S_580_PWS_COUNT(0));
+      si_pm4_cmd_add(pm4, 0xffffffff); /* GCR_SIZE */
+      si_pm4_cmd_add(pm4, 0x01ffffff); /* GCR_SIZE_HI */
+      si_pm4_cmd_add(pm4, 0); /* GCR_BASE_LO */
+      si_pm4_cmd_add(pm4, 0); /* GCR_BASE_HI */
+      si_pm4_cmd_add(pm4, S_585_PWS_ENA(1));
+      si_pm4_cmd_add(pm4, 0); /* GCR_CNTL */
 
       si_pm4_set_reg(pm4, R_031110_SPI_GS_THROTTLE_CNTL1, 0x12355123);
       si_pm4_set_reg(pm4, R_031114_SPI_GS_THROTTLE_CNTL2, 0x1544D);