[PATCH] drm/amdkfd: bind cpu and hiveless gpu to a hive if xgmi connected

[PATCH] drm/amdkfd: bind cpu and hiveless gpu to a hive if xgmi connected

[Alex Deucher]

From: Jonathan Kim <jonathan.kim@amd.com>

If a CPU and GPU are xGMI connected but the GPU is hiveless with
respect to other GPUs, create a new CPU-GPU hive using the GPU's PCI
device location ID as the new hive ID to maintain fine grain memory
access usage.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 5373a79ac6a1..c7072fff778e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1328,8 +1328,15 @@ static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev)
 
 		/* Include the CPU peer in GPU hive if connected over xGMI. */
 		if (!peer_dev->gpu &&
-		    link->iolink_type == CRAT_IOLINK_TYPE_XGMI)
+		    link->iolink_type == CRAT_IOLINK_TYPE_XGMI) {
+			/*
+			 * If the GPU is not part of a GPU hive, use its pci
+			 * device location as the hive ID to bind with the CPU.
+			 */
+			if (!dev->node_props.hive_id)
+				dev->node_props.hive_id = pci_dev_id(dev->gpu->adev->pdev);
 			peer_dev->node_props.hive_id = dev->node_props.hive_id;
+		}
 
 		list_for_each_entry(inbound_link, &peer_dev->io_link_props,
 									list) {
-- 
2.39.2

[PATCH] drm/amdgpu: add vcn_4_0_3 codec query

[Alex Deucher]

From: Sonny Jiang <sonny.jiang@amd.com>

Add support for vcn_4_0_3 video codec query

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/soc15.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index d767a8834ef1..4626cae7eb18 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -153,6 +153,24 @@ static const struct amdgpu_video_codecs rn_video_codecs_decode =
 	.codec_array = rn_video_codecs_decode_array,
 };
 
+static const struct amdgpu_video_codec_info vcn_4_0_3_video_codecs_decode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs vcn_4_0_3_video_codecs_decode = {
+	.codec_count = ARRAY_SIZE(vcn_4_0_3_video_codecs_decode_array),
+	.codec_array = vcn_4_0_3_video_codecs_decode_array,
+};
+
+static const struct amdgpu_video_codecs vcn_4_0_3_video_codecs_encode = {
+	.codec_count = 0,
+	.codec_array = NULL,
+};
+
 static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode,
 				    const struct amdgpu_video_codecs **codecs)
 {
@@ -185,6 +203,12 @@ static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode,
 			else
 				*codecs = &rn_video_codecs_decode;
 			return 0;
+		case IP_VERSION(4, 0, 3):
+			if (encode)
+				*codecs = &vcn_4_0_3_video_codecs_encode;
+			else
+				*codecs = &vcn_4_0_3_video_codecs_decode;
+			return 0;
 		default:
 			return -EINVAL;
 		}
-- 
2.39.2

[PATCH] drm/amdgpu: enable context empty interrupt on sdma v4.4.2

[Alex Deucher]

From: Le Ma <le.ma@amd.com>

With SDMA_CTNL.CTXEMPTY_INT_ENABLE set, the F32 clock can be gated when
SDMA finishes all job and goes to idle.

And no specific interrupt handling is required in driver.

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 729e26a4a2e7..7aa26e716a2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -890,6 +890,8 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
 		/* set utc l1 enable flag always to 1 */
 		temp = RREG32_SDMA(i, regSDMA_CNTL);
 		temp = REG_SET_FIELD(temp, SDMA_CNTL, UTC_L1_ENABLE, 1);
+		/* enable context empty interrupt during initialization */
+		temp = REG_SET_FIELD(temp, SDMA_CNTL, CTXEMPTY_INT_ENABLE, 1);
 		WREG32_SDMA(i, regSDMA_CNTL, temp);
 
 		if (!amdgpu_sriov_vf(adev)) {
-- 
2.39.2

[PATCH] drm/amdgpu: Enable MGCG on SDMAv4.4.2

[Alex Deucher]

From: Lijo Lazar <lijo.lazar@amd.com>

Enable clock gating on SDMAv4.4.2 versions. Leave memory light sleep to
default.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 22 +++++++++++++---------
 drivers/gpu/drm/amd/amdgpu/soc15.c       |  4 ++--
 2 files changed, 15 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 7aa26e716a2d..dca0beec0252 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1683,7 +1683,11 @@ static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
 	uint32_t data, def;
 	int i;
 
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS))
+		return;
+
+	if (enable) {
 		for_each_inst(i, inst_mask) {
 			/* 1-not override: enable sdma mem light sleep */
 			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
@@ -1708,12 +1712,14 @@ static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
 	uint32_t data, def;
 	int i;
 
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG))
+		return;
+
+	if (enable) {
 		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1725,9 +1731,7 @@ static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
 	} else {
 		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1773,7 +1777,7 @@ static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags)
 
 	/* AMD_CG_SUPPORT_SDMA_MGCG */
 	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_CLK_CTRL));
-	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK))
+	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK))
 		*flags |= AMD_CG_SUPPORT_SDMA_MGCG;
 
 	/* AMD_CG_SUPPORT_SDMA_LS */
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index 4626cae7eb18..8550a6b0b890 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -1154,8 +1154,8 @@ static int soc15_common_early_init(void *handle)
 		adev->asic_funcs = &aqua_vanjaram_asic_funcs;
 		adev->cg_flags =
 			AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG |
-			AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_VCN_MGCG |
-			AMD_CG_SUPPORT_JPEG_MGCG;
+			AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_SDMA_MGCG |
+			AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG;
 		adev->pg_flags =
 			AMD_PG_SUPPORT_VCN |
 			AMD_PG_SUPPORT_VCN_DPG |
-- 
2.39.2