[PATCH] drm/amdgpu: Enable SMU 13.0.0 optimizations when ROCm is active (v2)
Zhang, Hawking
Hawking.Zhang at amd.com
Thu Oct 5 03:40:08 UTC 2023
[AMD Official Use Only - General]
Hmm... thinking about it more, will it override the profile mode/workload for 0xC8 or 0xCC SKU as well. In another words, does it mean the pmfw fix is general to all the 13_0_0 SKUs.
Other than that, the patch looks good to me.
Regards,
Hawking
-----Original Message-----
From: amd-gfx <amd-gfx-bounces at lists.freedesktop.org> On Behalf Of Zhang, Hawking
Sent: Thursday, October 5, 2023 11:32
To: Deucher, Alexander <Alexander.Deucher at amd.com>; amd-gfx at lists.freedesktop.org
Cc: Deucher, Alexander <Alexander.Deucher at amd.com>; Liu, Kun <Kun.Liu2 at amd.com>
Subject: RE: [PATCH] drm/amdgpu: Enable SMU 13.0.0 optimizations when ROCm is active (v2)
[AMD Official Use Only - General]
[AMD Official Use Only - General]
Reviewed-by: Hawking Zhang <Hawking.Zhang at amd.com>
Regards,
Hawking
-----Original Message-----
From: amd-gfx <amd-gfx-bounces at lists.freedesktop.org> On Behalf Of Alex Deucher
Sent: Wednesday, October 4, 2023 23:34
To: amd-gfx at lists.freedesktop.org
Cc: Deucher, Alexander <Alexander.Deucher at amd.com>; Liu, Kun <Kun.Liu2 at amd.com>
Subject: [PATCH] drm/amdgpu: Enable SMU 13.0.0 optimizations when ROCm is active (v2)
From: Kun Liu <Kun.Liu2 at amd.com>
When ROCm is active enable additional SMU 13.0.0 optimizations.
This reuses the unused powersave profile on PMFW.
v2: move to the swsmu code since we need both bits active in
the workload mask.
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 17 ++++++++++++++++-
1 file changed, 16 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c
index 684b4e01fac2..83035fb1839a 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c
@@ -2447,6 +2447,7 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu,
DpmActivityMonitorCoeffInt_t *activity_monitor =
&(activity_monitor_external.DpmActivityMonitorCoeffInt);
int workload_type, ret = 0;
+ u32 workload_mask;
smu->power_profile_mode = input[size];
@@ -2536,9 +2537,23 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu,
if (workload_type < 0)
return -EINVAL;
+ workload_mask = 1 << workload_type;
+
+ /* Add optimizations for SMU13.0.0. Reuse the power saving profile */
+ if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_COMPUTE &&
+ (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 0)) &&
+ ((smu->adev->pm.fw_version == 0x004e6601) ||
+ (smu->adev->pm.fw_version >= 0x004e7300))) {
+ workload_type = smu_cmn_to_asic_specific_index(smu,
+ CMN2ASIC_MAPPING_WORKLOAD,
+ PP_SMC_POWER_PROFILE_POWERSAVING);
+ if (workload_type >= 0)
+ workload_mask |= 1 << workload_type;
+ }
+
return smu_cmn_send_smc_msg_with_param(smu,
SMU_MSG_SetWorkloadMask,
- 1 << workload_type,
+ workload_mask,
NULL); }
--
2.41.0
More information about the amd-gfx
mailing list