[PATCH 1/2] drm/amdkfd: Delete useless SDMA register setting on non HWS path
Zhao, Yong
Yong.Zhao at amd.com
Mon Sep 23 03:56:47 UTC 2019
It turns out when loading hqd for SDMA queues, RESUME_CTX of
SDMA*_GFX_CONTEXT_CNTL is already 0, so there is no need to set it
to 0 again. In addition, it does not seem right that we should touch
SDMA GFX queues when manipulating RLC queues.
Change-Id: I2c142d024e94f92194b1cb9feb7f44396b8f3ecc
Signed-off-by: Yong Zhao <Yong.Zhao at amd.com>
---
.../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 34 +------------------
.../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 9 +----
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 9 +----
3 files changed, 3 insertions(+), 49 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 4d9101834ba7..c9ce1516956e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -102,38 +102,12 @@ static uint32_t get_sdma_base_addr(struct amdgpu_device *adev,
return retval;
}
-static u32 sdma_v4_0_get_reg_offset(struct amdgpu_device *adev,
- u32 instance, u32 offset)
-{
- switch (instance) {
- case 0:
- return (adev->reg_offset[SDMA0_HWIP][0][0] + offset);
- case 1:
- return (adev->reg_offset[SDMA1_HWIP][0][1] + offset);
- case 2:
- return (adev->reg_offset[SDMA2_HWIP][0][1] + offset);
- case 3:
- return (adev->reg_offset[SDMA3_HWIP][0][1] + offset);
- case 4:
- return (adev->reg_offset[SDMA4_HWIP][0][1] + offset);
- case 5:
- return (adev->reg_offset[SDMA5_HWIP][0][1] + offset);
- case 6:
- return (adev->reg_offset[SDMA6_HWIP][0][1] + offset);
- case 7:
- return (adev->reg_offset[SDMA7_HWIP][0][1] + offset);
- default:
- break;
- }
- return 0;
-}
-
static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
uint32_t __user *wptr, struct mm_struct *mm)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
struct v9_sdma_mqd *m;
- uint32_t sdma_base_addr, sdmax_gfx_context_cntl;
+ uint32_t sdma_base_addr;
unsigned long end_jiffies;
uint32_t data;
uint64_t data64;
@@ -142,8 +116,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
m = get_sdma_mqd(mqd);
sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id,
m->sdma_queue_id);
- sdmax_gfx_context_cntl = sdma_v4_0_get_reg_offset(adev,
- m->sdma_engine_id, mmSDMA0_GFX_CONTEXT_CNTL);
WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL,
m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK));
@@ -157,10 +129,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
return -ETIME;
usleep_range(500, 1000);
}
- data = RREG32(sdmax_gfx_context_cntl);
- data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL,
- RESUME_CTX, 0);
- WREG32(sdmax_gfx_context_cntl, data);
WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL_OFFSET,
m->sdmax_rlcx_doorbell_offset);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 3aff2b5758e0..a4325db8d093 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -489,7 +489,7 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
struct v10_sdma_mqd *m;
- uint32_t sdma_base_addr, sdmax_gfx_context_cntl;
+ uint32_t sdma_base_addr;
unsigned long end_jiffies;
uint32_t data;
uint64_t data64;
@@ -499,9 +499,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id,
m->sdma_queue_id);
pr_debug("sdma load base addr %x for engine %d, queue %d\n", sdma_base_addr, m->sdma_engine_id, m->sdma_queue_id);
- sdmax_gfx_context_cntl = m->sdma_engine_id ?
- SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_GFX_CONTEXT_CNTL) :
- SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_CONTEXT_CNTL);
WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL,
m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK));
@@ -515,10 +512,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
return -ETIME;
usleep_range(500, 1000);
}
- data = RREG32(sdmax_gfx_context_cntl);
- data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL,
- RESUME_CTX, 0);
- WREG32(sdmax_gfx_context_cntl, data);
WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL_OFFSET,
m->sdmax_rlcx_doorbell_offset);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 741a3c54a21c..55437f160a72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -388,7 +388,7 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
struct v9_sdma_mqd *m;
- uint32_t sdma_base_addr, sdmax_gfx_context_cntl;
+ uint32_t sdma_base_addr;
unsigned long end_jiffies;
uint32_t data;
uint64_t data64;
@@ -397,9 +397,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
m = get_sdma_mqd(mqd);
sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id,
m->sdma_queue_id);
- sdmax_gfx_context_cntl = m->sdma_engine_id ?
- SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_GFX_CONTEXT_CNTL) :
- SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_CONTEXT_CNTL);
WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL,
m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK));
@@ -413,10 +410,6 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
return -ETIME;
usleep_range(500, 1000);
}
- data = RREG32(sdmax_gfx_context_cntl);
- data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL,
- RESUME_CTX, 0);
- WREG32(sdmax_gfx_context_cntl, data);
WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL_OFFSET,
m->sdmax_rlcx_doorbell_offset);
--
2.17.1
More information about the amd-gfx
mailing list