Mesa (main): amd/rgp: Use VGH clocks for RGP workaround.

GitLab Mirror gitlab-mirror at kemper.freedesktop.org
Wed Oct 27 22:17:50 UTC 2021


Module: Mesa
Branch: main
Commit: 2c43fd4c41f32479441a5845b46f844e66aeb3cc
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=2c43fd4c41f32479441a5845b46f844e66aeb3cc

Author: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
Date:   Wed Oct 27 01:49:23 2021 +0200

amd/rgp: Use VGH clocks for RGP workaround.

Hear that it matters for RGP. This is the most likely scenario where
we would hit this workaround, given the tooling for profiling on the
deck will set profile_peak as workaround for hangs.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13534>

---

 src/amd/common/ac_rgp.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/amd/common/ac_rgp.c b/src/amd/common/ac_rgp.c
index 5abe0c8b59a..f82bd59b5b7 100644
--- a/src/amd/common/ac_rgp.c
+++ b/src/amd/common/ac_rgp.c
@@ -453,12 +453,12 @@ static void ac_sqtt_fill_asic_info(struct radeon_info *rad_info,
    chunk->trace_shader_core_clock = rad_info->max_shader_clock * 1000000;
    chunk->trace_memory_clock = rad_info->max_memory_clock * 1000000;
 
-   /* RGP gets very confused if these clocks are 0. The 1 GHz clocks are not necessarily correct,
-    * but the resulting traces are at least somewhat useful. */
+   /* RGP gets very confused if these clocks are 0. The numbers here are for profile_peak on
+    * VGH since that is the chips where we've seen the need for this workaround. */
    if (!chunk->trace_shader_core_clock)
-      chunk->trace_shader_core_clock = 1e9;
+      chunk->trace_shader_core_clock = 1300000000;
    if (!chunk->trace_memory_clock)
-      chunk->trace_memory_clock = 1e9;
+      chunk->trace_memory_clock = 687000000;
 
    chunk->device_id = rad_info->pci_id;
    chunk->device_revision_id = rad_info->pci_rev_id;



More information about the mesa-commit mailing list