drm/amdgpu: add register distance
Subsystems (2)
cover drm/amdgpu: add register distance
During IP block updates in some asic, the register distribution pattern of context domains and invalidation engines will be changed in gc_v***.h and mmhub_v***.h. However the register distances are always constant in one IP block beside gfxhub and mmhub. So we can add register distance member in vmhub structure to store the register distance value, that's able to avoid the issue of hardcode at current in future updates.
Thanks,
Ray
Huang Rui (12):
drm/amdgpu: add register distance members into vmhub structure drm/amdgpu: use register distance member instead of hardcode in GMC9 drm/amdgpu: use register distance member instead of hardcode in GMC10 drm/amdgpu: use register distance member instead of hardcode in
VCN1/JEPG1
drm/amdgpu: use register distance member instead of hardcode in
UVD7/VCE4
drm/amdgpu: use register distance member instead of hardcode in
VCN2/JEPG2
drm/amdgpu: use register distance member instead of hardcode in gfxhub
v1
drm/amdgpu: use register distance member instead of hardcode in gfxhub
v2
drm/amdgpu: use register distance member instead of hardcode in gfxhub
v2.1
drm/amdgpu: use register distance member instead of hardcode in mmhub
v1
drm/amdgpu: use register distance member instead of hardcode in mmhub
v2
drm/amdgpu: use register distance member instead of hardcode in mmhub
v9.4drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 9 +++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 48 ++++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 47 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 47 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 43 ++++++++++++--------- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 39 ++++++++++++------- drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 45 ++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 46 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 46 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 5 ++- drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 3 +- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 5 ++- drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 5 ++- 15 files changed, 260 insertions(+), 132 deletions(-)
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 1/12 drm/amdgpu: add register distance members into vmhub structure
This patch is to abstract register distances between two continuous context domains and invalidation engines. In different ip headers, these distances may be differences.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 9 +++++++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 7 +++++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 8 ++++++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 8 ++++++++ drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 7 +++++++ drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 9 +++++++++ 7 files changed, 54 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +9 -0
@@ -83,6 +83,15 @@ struct amdgpu_vmhub {uint32_t vm_context0_cntl;uint32_t vm_l2_pro_fault_status;uint32_t vm_l2_pro_fault_cntl; + +/* +* store the register distances between two continuous context domain +* and invalidation engine. +*/ +uint32_t ctx_distance; +uint32_t ctx_addr_distance;/* include LO32/HI32 */ +uint32_t eng_distance; +uint32_t eng_addr_distance;/* include LO32/HI32 */ };/*
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +7 -0
@@ -386,4 +386,11 @@ void gfxhub_v1_0_init(struct amdgpu_device *adev)SOC15_REG_OFFSET (GC ,0 , mmVM_L2_PROTECTION_FAULT_STATUS); hub->vm_l2_pro_fault_cntl =SOC15_REG_OFFSET (GC ,0 , mmVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = mmVM_CONTEXT1_CNTL - mmVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = mmVM_INVALIDATE_ENG1_REQ - mmVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = mmVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; }
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c +8 -0
@@ -373,4 +373,12 @@ void gfxhub_v2_0_init(struct amdgpu_device *adev)SOC15_REG_OFFSET (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS); hub->vm_l2_pro_fault_cntl =SOC15_REG_OFFSET (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = mmGCVM_CONTEXT1_CNTL - mmGCVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = mmGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = mmGCVM_INVALIDATE_ENG1_REQ - + mmGCVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = mmGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; }
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +8 -0
@@ -372,6 +372,14 @@ void gfxhub_v2_1_init(struct amdgpu_device *adev)SOC15_REG_OFFSET (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS); hub->vm_l2_pro_fault_cntl =SOC15_REG_OFFSET (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = mmGCVM_CONTEXT1_CNTL - mmGCVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = mmGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = mmGCVM_INVALIDATE_ENG1_REQ - + mmGCVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = mmGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; }int gfxhub_v2_1_get_xgmi_info (struct amdgpu_device *adev)
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +6 -0
@@ -429,6 +429,12 @@ void mmhub_v1_0_init(struct amdgpu_device *adev) hub->vm_l2_pro_fault_cntl =SOC15_REG_OFFSET (MMHUB ,0 , mmVM_L2_PROTECTION_FAULT_CNTL); + hub->ctx_distance = mmVM_CONTEXT1_CNTL - mmVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = mmVM_INVALIDATE_ENG1_REQ - mmVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = mmVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; } static void mmhub_v1_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +7 -0
@@ -365,6 +365,13 @@ void mmhub_v2_0_init(struct amdgpu_device *adev) hub->vm_l2_pro_fault_cntl =SOC15_REG_OFFSET (MMHUB ,0 , mmMMVM_L2_PROTECTION_FAULT_CNTL); + hub->ctx_distance = mmMMVM_CONTEXT1_CNTL - mmMMVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = mmMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = mmMMVM_INVALIDATE_ENG1_REQ - + mmMMVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = mmMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; } static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +9 -0
@@ -534,6 +534,15 @@ void mmhub_v9_4_init(struct amdgpu_device *adev)SOC15_REG_OFFSET (MMHUB ,0 , mmVML2PF0_VM_L2_PROTECTION_FAULT_CNTL) + i *MMHUB_INSTANCE_REGISTER_OFFSET ; + + hub[i]->ctx_distance = mmVML2VC0_VM_CONTEXT1_CNTL - + mmVML2VC0_VM_CONTEXT0_CNTL; + hub[i]->ctx_addr_distance = mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub[i]->eng_distance = mmVML2VC0_VM_INVALIDATE_ENG1_REQ - + mmVML2VC0_VM_INVALIDATE_ENG0_REQ; + hub[i]->eng_addr_distance = mmVML2VC0_VM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + mmVML2VC0_VM_INVALIDATE_ENG0_ADDR_RANGE_LO32; } }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/12 drm/amdgpu: use register distance member instead of hardcode in GMC9
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in GMC9.
--- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 39 +++++++++++++++++---------- 1 file changed, 25 insertions(+), 14 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +25 -14
@@ -502,11 +502,11 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,if (adev->gfx.kiq.ring.sched.ready && (amdgpu_sriov_runtime (adev) || !amdgpu_sriov_vf (adev)) && !adev->in_gpu_reset) { -uint32_t req = hub->vm_inv_eng0_req + eng; -uint32_t ack = hub->vm_inv_eng0_ack + eng; +uint32_t req = hub->vm_inv_eng0_req + hub->eng_distance * eng; +uint32_t ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;amdgpu_virt_kiq_reg_write_reg_wait (adev, req, ack, inv_req, -1 << vmid); +1 << vmid);return ; }
@@ -523,7 +523,8 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,if (use_semaphore) {for (j =0 ; j < adev->usec_timeout; j++) {/* a read return value of 1 means semaphore acuqire */ - tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_sem + eng); + tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_sem + + hub->eng_distance * eng);if (tmp &0x1 )break ;udelay (1 );
@@ -534,7 +535,8 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, } do { - WREG32_NO_KIQ(hub->vm_inv_eng0_req + eng, inv_req); + WREG32_NO_KIQ(hub->vm_inv_eng0_req + + hub->eng_distance * eng, inv_req); /* * Issue a dummy read to wait for the ACK register to
@@ -542,10 +544,12 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,* GRBM interface. */ if (vmhub ==AMDGPU_GFXHUB_0 ) -RREG32_NO_KIQ (hub->vm_inv_eng0_req + eng); +RREG32_NO_KIQ (hub->vm_inv_eng0_req + + hub->eng_distance * eng);for (j =0 ; j < adev->usec_timeout; j++) { - tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_ack + eng); + tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_ack + + hub->eng_distance * eng);if (tmp & (1 << vmid))break ;udelay (1 );
@@ -561,7 +565,8 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,* add semaphore release after invalidation, * write with 0 means semaphore release */ -WREG32_NO_KIQ (hub->vm_inv_eng0_sem + eng,0 ); +WREG32_NO_KIQ (hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0 );spin_unlock (&adev->gmc.invalidate_lock);
@@ -676,16 +681,21 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,if (use_semaphore)/* a read return value of 1 means semaphore acuqire */ amdgpu_ring_emit_reg_wait (ring, - hub->vm_inv_eng0_sem + eng,0x1 ,0x1 ); + hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0x1 ,0x1 ); -amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid), +amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_lo32 + + (hub->ctx_addr_distance * vmid),lower_32_bits (pd_addr)); -amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_hi32 + (2 * vmid), +amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_hi32 + + (hub->ctx_addr_distance * vmid),upper_32_bits (pd_addr)); -amdgpu_ring_emit_reg_write_reg_wait (ring, hub->vm_inv_eng0_req + eng, - hub->vm_inv_eng0_ack + eng, +amdgpu_ring_emit_reg_write_reg_wait (ring, hub->vm_inv_eng0_req + + hub->eng_distance * eng, + hub->vm_inv_eng0_ack + + hub->eng_distance * eng, req,1 << vmid);/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
@@ -694,7 +704,8 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,* add semaphore release after invalidation, * write with 0 means semaphore release */ -amdgpu_ring_emit_wreg (ring, hub->vm_inv_eng0_sem + eng,0 ); +amdgpu_ring_emit_wreg (ring, hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0 );return pd_addr; }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 3/12 drm/amdgpu: use register distance member instead of hardcode in GMC10
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in GMC10.
--- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 43 ++++++++++++++++---------- 1 file changed, 26 insertions(+), 17 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +26 -17
@@ -86,7 +86,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,/* MM HUB */ hub = &adev->vmhub[AMDGPU_MMHUB_0 ];for (i =0 ; i <16 ; i++) { - reg = hub->vm_context0_cntl + i; + reg = hub->vm_context0_cntl + hub->ctx_distance * i; tmp =RREG32 (reg); tmp &= ~bits[AMDGPU_MMHUB_0 ];WREG32 (reg, tmp);
@@ -95,7 +95,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,/* GFX HUB */ hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];for (i =0 ; i <16 ; i++) { - reg = hub->vm_context0_cntl + i; + reg = hub->vm_context0_cntl + hub->ctx_distance * i; tmp =RREG32 (reg); tmp &= ~bits[AMDGPU_GFXHUB_0 ];WREG32 (reg, tmp);
@@ -105,7 +105,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,/* MM HUB */ hub = &adev->vmhub[AMDGPU_MMHUB_0 ];for (i =0 ; i <16 ; i++) { - reg = hub->vm_context0_cntl + i; + reg = hub->vm_context0_cntl + hub->ctx_distance * i; tmp =RREG32 (reg); tmp |= bits[AMDGPU_MMHUB_0 ];WREG32 (reg, tmp);
@@ -114,7 +114,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,/* GFX HUB */ hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];for (i =0 ; i <16 ; i++) { - reg = hub->vm_context0_cntl + i; + reg = hub->vm_context0_cntl + hub->ctx_distance * i; tmp =RREG32 (reg); tmp |= bits[AMDGPU_GFXHUB_0 ];WREG32 (reg, tmp);
@@ -283,7 +283,8 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,if (use_semaphore) {for (i =0 ; i < adev->usec_timeout; i++) {/* a read return value of 1 means semaphore acuqire */ - tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_sem + eng); + tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_sem + + hub->eng_distance * eng);if (tmp &0x1 )break ;udelay (1 );
@@ -293,18 +294,19 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,DRM_ERROR ("Timeout waiting for sem acquire in VM flush!\n" ); } -WREG32_NO_KIQ (hub->vm_inv_eng0_req + eng, inv_req); +WREG32_NO_KIQ (hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);/* * Issue a dummy read to wait for the ACK register to be cleared * to avoid a false ACK due to the new fast GRBM interface. */ if (vmhub ==AMDGPU_GFXHUB_0 ) -RREG32_NO_KIQ (hub->vm_inv_eng0_req + eng); +RREG32_NO_KIQ (hub->vm_inv_eng0_req + hub->eng_distance * eng);/* Wait for ACK with a delay.*/ for (i =0 ; i < adev->usec_timeout; i++) { - tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_ack + eng); + tmp =RREG32_NO_KIQ (hub->vm_inv_eng0_ack + + hub->eng_distance * eng); tmp &=1 << vmid;if (tmp)break ;
@@ -318,7 +320,8 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,* add semaphore release after invalidation, * write with 0 means semaphore release */ -WREG32_NO_KIQ (hub->vm_inv_eng0_sem + eng,0 ); +WREG32_NO_KIQ (hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0 );spin_unlock (&adev->gmc.invalidate_lock);
@@ -358,8 +361,8 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];const unsigned eng =17 ;u32 inv_req =gmc_v10_0_get_invalidate_req (vmid, flush_type); -u32 req = hub->vm_inv_eng0_req + eng; -u32 ack = hub->vm_inv_eng0_ack + eng; +u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng; +u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;amdgpu_virt_kiq_reg_write_reg_wait (adev, req, ack, inv_req,1 << vmid);
@@ -502,16 +505,21 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,if (use_semaphore)/* a read return value of 1 means semaphore acuqire */ amdgpu_ring_emit_reg_wait (ring, - hub->vm_inv_eng0_sem + eng,0x1 ,0x1 ); + hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0x1 ,0x1 ); -amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid), +amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_lo32 + + (hub->ctx_addr_distance * vmid),lower_32_bits (pd_addr)); -amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_hi32 + (2 * vmid), +amdgpu_ring_emit_wreg (ring, hub->ctx0_ptb_addr_hi32 + + (hub->ctx_addr_distance * vmid),upper_32_bits (pd_addr)); -amdgpu_ring_emit_reg_write_reg_wait (ring, hub->vm_inv_eng0_req + eng, - hub->vm_inv_eng0_ack + eng, +amdgpu_ring_emit_reg_write_reg_wait (ring, hub->vm_inv_eng0_req + + hub->eng_distance * eng, + hub->vm_inv_eng0_ack + + hub->eng_distance * eng, req,1 << vmid);/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
@@ -520,7 +528,8 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,* add semaphore release after invalidation, * write with 0 means semaphore release */ -amdgpu_ring_emit_wreg (ring, hub->vm_inv_eng0_sem + eng,0 ); +amdgpu_ring_emit_wreg (ring, hub->vm_inv_eng0_sem + + hub->eng_distance * eng,0 );return pd_addr; }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/12 drm/amdgpu: use register distance member instead of hardcode in VCN1/JEPG1
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in VCN1/JEPG1. They are for the same ASIC.
--- drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 5 +++-- 2 files changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c +1 -1
@@ -377,7 +377,7 @@ static void jpeg_v1_0_decode_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for register write */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid *2 ; + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; data1 =lower_32_bits (pd_addr); mask =0xffffffff ;jpeg_v1_0_decode_ring_emit_reg_wait (ring, data0, data1, mask);
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +3 -2
@@ -1539,7 +1539,7 @@ static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for register write */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid *2 ; + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; data1 =lower_32_bits (pd_addr); mask =0xffffffff ;vcn_v1_0_dec_ring_emit_reg_wait (ring, data0, data1, mask);
@@ -1679,7 +1679,8 @@ static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for reg writes */ -vcn_v1_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + vmid *2 , +vcn_v1_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + + vmid * hub->ctx_addr_distance,lower_32_bits (pd_addr),0xffffffff ); }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 5/12 drm/amdgpu: use register distance member instead of hardcode in UVD7/VCE4
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in UVD7/VCE4. They are for the same ASIC.
--- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 5 +++-- drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 3 ++- 2 files changed, 5 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +3 -2
@@ -1375,7 +1375,7 @@ static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for reg writes */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid *2 ; + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; data1 =lower_32_bits (pd_addr); mask =0xffffffff ;uvd_v7_0_ring_emit_reg_wait (ring, data0, data1, mask);
@@ -1417,7 +1417,8 @@ static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for reg writes */ -uvd_v7_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + vmid *2 , +uvd_v7_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + + vmid * hub->ctx_addr_distance,lower_32_bits (pd_addr),0xffffffff ); }
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +2 -1
@@ -991,7 +991,8 @@ static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for reg writes */ -vce_v4_0_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + vmid *2 , +vce_v4_0_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + + vmid * hub->ctx_addr_distance,lower_32_bits (pd_addr),0xffffffff ); }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 6/12 drm/amdgpu: use register distance member instead of hardcode in VCN2/JEPG2
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in VCN2/JEPG2. They are for the same ASIC.
--- drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 5 +++-- 2 files changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c +1 -1
@@ -629,7 +629,7 @@ void jpeg_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for register write */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid *2 ; + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; data1 =lower_32_bits (pd_addr); mask =0xffffffff ;jpeg_v2_0_dec_ring_emit_reg_wait (ring, data0, data1, mask);
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +3 -2
@@ -1505,7 +1505,7 @@ void vcn_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for register write */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid *2 ; + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; data1 =lower_32_bits (pd_addr); mask =0xffffffff ;vcn_v2_0_dec_ring_emit_reg_wait (ring, data0, data1, mask);
@@ -1660,7 +1660,8 @@ void vcn_v2_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, pd_addr =amdgpu_gmc_emit_flush_gpu_tlb (ring, vmid, pd_addr);/* wait for reg writes */ -vcn_v2_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + vmid *2 , +vcn_v2_0_enc_ring_emit_reg_wait (ring, hub->ctx0_ptb_addr_lo32 + + vmid * hub->ctx_addr_distance,lower_32_bits (pd_addr),0xffffffff ); }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 7/12 drm/amdgpu: use register distance member instead of hardcode in gfxhub v1
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in gfxhub v1.
--- drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 41 +++++++++++++++--------- 1 file changed, 25 insertions(+), 16 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +25 -16
@@ -38,15 +38,15 @@ u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev)void gfxhub_v1_0_setup_vm_pt_regs (struct amdgpu_device *adev,uint32_t vmid,uint64_t page_table_base) { -/* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */ -int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - offset * vmid,lower_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +lower_32_bits (page_table_base));WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - offset * vmid,upper_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +upper_32_bits (page_table_base)); }static void gfxhub_v1_0_init_gart_aperture_regs (struct amdgpu_device *adev)
@@ -207,6 +207,7 @@ static void gfxhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)static void gfxhub_v1_0_setup_vmid_config (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];unsigned num_level, block_size;uint32_t tmp;int i;
@@ -245,25 +246,31 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp =REG_SET_FIELD (tmp,VM_CONTEXT1_CNTL ,RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry); -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_CNTL, i, tmp); -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, i*2 , -lower_32_bits (adev->vm_manager.max_pfn -1 )); -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, i*2 , -upper_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, +lower_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, +upper_32_bits (adev->vm_manager.max_pfn -1 )); } }static void gfxhub_v1_0_program_invalidation (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (GC ,0 , mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, -2 * i,0xffffffff ); + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (GC ,0 , mmVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, -2 * i,0x1f ); + i * hub->eng_addr_distance,0x1f ); } }
@@ -299,12 +306,14 @@ int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev)void gfxhub_v1_0_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];u32 tmp;u32 i;/* Disable all tables */ for (i =0 ; i <16 ; i++) -WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT0_CNTL, i,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmVM_CONTEXT0_CNTL, + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp =RREG32_SOC15 (GC ,0 , mmMC_VM_MX_L1_TLB_CNTL);
@@ -360,7 +369,7 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, CRASH_ON_NO_RETRY_FAULT, 1); tmp =REG_SET_FIELD (tmp,VM_L2_PROTECTION_FAULT_CNTL ,CRASH_ON_RETRY_FAULT ,1 ); - } + }WREG32_SOC15 (GC ,0 , mmVM_L2_PROTECTION_FAULT_CNTL, tmp); }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 8/12 drm/amdgpu: use register distance member instead of hardcode in gfxhub v2
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in gfxhub v2.
--- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 39 +++++++++++++++--------- 1 file changed, 24 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c +24 -15
@@ -49,15 +49,15 @@ u64 gfxhub_v2_0_get_mc_fb_offset(struct amdgpu_device *adev)void gfxhub_v2_0_setup_vm_pt_regs (struct amdgpu_device *adev,uint32_t vmid,uint64_t page_table_base) { -/* two registers distance between mmGCVM_CONTEXT0_* to mmGCVM_CONTEXT1_* */ -int offset = mmGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - offset * vmid,lower_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +lower_32_bits (page_table_base));WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - offset * vmid,upper_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +upper_32_bits (page_table_base)); }static void gfxhub_v2_0_init_gart_aperture_regs (struct amdgpu_device *adev)
@@ -218,6 +218,7 @@ static void gfxhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)static void gfxhub_v2_0_setup_vmid_config (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];int i;uint32_t tmp;
@@ -247,25 +248,31 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev) tmp =REG_SET_FIELD (tmp,GCVM_CONTEXT1_CNTL ,RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_CNTL, i, tmp); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, i*2 , -lower_32_bits (adev->vm_manager.max_pfn -1 )); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, i*2 , -upper_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, +lower_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, +upper_32_bits (adev->vm_manager.max_pfn -1 )); } }static void gfxhub_v2_0_program_invalidation (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, -2 * i,0xffffffff ); + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, -2 * i,0x1f ); + i * hub->eng_addr_distance,0x1f ); } }
@@ -287,12 +294,14 @@ int gfxhub_v2_0_gart_enable(struct amdgpu_device *adev)void gfxhub_v2_0_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];u32 tmp;u32 i;/* Disable all tables */ for (i =0 ; i <16 ; i++) -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_CNTL, i,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_CNTL, + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp =RREG32_SOC15 (GC ,0 , mmGCMC_VM_MX_L1_TLB_CNTL);
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 9/12 drm/amdgpu: use register distance member instead of hardcode in gfxhub v2.1
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in gfxhub v2.1.
--- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 39 +++++++++++++++--------- 1 file changed, 24 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +24 -15
@@ -49,15 +49,15 @@ u64 gfxhub_v2_1_get_mc_fb_offset(struct amdgpu_device *adev)void gfxhub_v2_1_setup_vm_pt_regs (struct amdgpu_device *adev,uint32_t vmid,uint64_t page_table_base) { -/* two registers distance between mmGCVM_CONTEXT0_* to mmGCVM_CONTEXT1_* */ -int offset = mmGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - offset * vmid,lower_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +lower_32_bits (page_table_base));WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - offset * vmid,upper_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +upper_32_bits (page_table_base)); }static void gfxhub_v2_1_init_gart_aperture_regs (struct amdgpu_device *adev)
@@ -207,6 +207,7 @@ static void gfxhub_v2_1_disable_identity_aperture(struct amdgpu_device *adev)static void gfxhub_v2_1_setup_vmid_config (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];int i;uint32_t tmp;
@@ -236,25 +237,31 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev) tmp =REG_SET_FIELD (tmp,GCVM_CONTEXT1_CNTL ,RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_CNTL, i, tmp); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2 ,0 ); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, i*2 , -lower_32_bits (adev->vm_manager.max_pfn -1 )); -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, i*2 , -upper_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, +lower_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, +upper_32_bits (adev->vm_manager.max_pfn -1 )); } }static void gfxhub_v2_1_program_invalidation (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, -2 * i,0xffffffff ); + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, -2 * i,0x1f ); + i * hub->eng_addr_distance,0x1f ); } }
@@ -288,12 +295,14 @@ int gfxhub_v2_1_gart_enable(struct amdgpu_device *adev)void gfxhub_v2_1_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0 ];u32 tmp;u32 i;/* Disable all tables */ for (i =0 ; i <16 ; i++) -WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_CNTL, i,0 ); +WREG32_SOC15_OFFSET (GC ,0 , mmGCVM_CONTEXT0_CNTL, + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp =RREG32_SOC15 (GC ,0 , mmGCMC_VM_MX_L1_TLB_CNTL);
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 10/12 drm/amdgpu: use register distance member instead of hardcode in mmhub v1
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in mmhub v1.
--- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 39 +++++++++++++++---------- 1 file changed, 24 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +24 -15
@@ -54,15 +54,15 @@ u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev)void mmhub_v1_0_setup_vm_pt_regs (struct amdgpu_device *adev,uint32_t vmid,uint64_t page_table_base) { -/* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */ -int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - offset * vmid,lower_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +lower_32_bits (page_table_base));WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - offset * vmid,upper_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +upper_32_bits (page_table_base)); }static void mmhub_v1_0_init_gart_aperture_regs (struct amdgpu_device *adev)
@@ -230,6 +230,7 @@ static void mmhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)static void mmhub_v1_0_setup_vmid_config (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];unsigned num_level, block_size;uint32_t tmp;int i;
@@ -268,25 +269,31 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp =REG_SET_FIELD (tmp,VM_CONTEXT1_CNTL ,RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_CNTL, i, tmp); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2 ,0 ); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2 ,0 ); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, i*2 , -lower_32_bits (adev->vm_manager.max_pfn -1 )); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, i*2 , -upper_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, +lower_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, +upper_32_bits (adev->vm_manager.max_pfn -1 )); } }static void mmhub_v1_0_program_invalidation (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, -2 * i,0xffffffff ); + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, -2 * i,0x1f ); + i * hub->eng_addr_distance,0x1f ); } }
@@ -333,12 +340,14 @@ int mmhub_v1_0_gart_enable(struct amdgpu_device *adev)void mmhub_v1_0_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];u32 tmp;u32 i;/* Disable all tables */ for (i =0 ; i <16 ; i++) -WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT0_CNTL, i,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmVM_CONTEXT0_CNTL, + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp =RREG32_SOC15 (MMHUB ,0 , mmMC_VM_MX_L1_TLB_CNTL);
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 11/12 drm/amdgpu: use register distance member instead of hardcode in mmhub v2
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in mmhub v2.
--- drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 39 +++++++++++++++---------- 1 file changed, 24 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +24 -15
@@ -39,15 +39,15 @@void mmhub_v2_0_setup_vm_pt_regs (struct amdgpu_device *adev,uint32_t vmid,uint64_t page_table_base) { -/* two registers distance between mmMMVM_CONTEXT0_* to mmMMVM_CONTEXT1_* */ -int offset = mmMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - offset * vmid,lower_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +lower_32_bits (page_table_base));WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - offset * vmid,upper_32_bits (page_table_base)); + hub->ctx_addr_distance * vmid, +upper_32_bits (page_table_base)); }static void mmhub_v2_0_init_gart_aperture_regs (struct amdgpu_device *adev)
@@ -209,6 +209,7 @@ static void mmhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)static void mmhub_v2_0_setup_vmid_config (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];int i;uint32_t tmp;
@@ -239,25 +240,31 @@ static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev) tmp =REG_SET_FIELD (tmp,MMVM_CONTEXT1_CNTL ,RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_CNTL, i, tmp); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2 ,0 ); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2 ,0 ); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, i*2 , -lower_32_bits (adev->vm_manager.max_pfn -1 )); -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, i*2 , -upper_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, +lower_32_bits (adev->vm_manager.max_pfn -1 )); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, +upper_32_bits (adev->vm_manager.max_pfn -1 )); } }static void mmhub_v2_0_program_invalidation (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, -2 * i,0xffffffff ); + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, -2 * i,0x1f ); + i * hub->eng_addr_distance,0x1f ); } }
@@ -279,12 +286,14 @@ int mmhub_v2_0_gart_enable(struct amdgpu_device *adev)void mmhub_v2_0_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];u32 tmp;u32 i;/* Disable all tables */ for (i =0 ; i <16 ; i++) -WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT0_CNTL, i,0 ); +WREG32_SOC15_OFFSET (MMHUB ,0 , mmMMVM_CONTEXT0_CNTL, + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp =RREG32_SOC15 (MMHUB ,0 , mmMMMC_VM_MX_L1_TLB_CNTL);
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 12/12 drm/amdgpu: use register distance member instead of hardcode in mmhub v9.4
In reply to Huang RuiThis patch updates to use register distance member instead of hardcode in mmhub v9.4.
--- drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 37 ++++++++++++++----------- 1 file changed, 21 insertions(+), 16 deletions(-)
drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +21 -16
@@ -57,20 +57,16 @@ u64 mmhub_v9_4_get_fb_location(struct amdgpu_device *adev)static void mmhub_v9_4_setup_hubid_vm_pt_regs (struct amdgpu_device *adev,int hubid,uint32_t vmid,uint64_t value) { -/* two registers distance between mmVML2VC0_VM_CONTEXT0_* to -* mmVML2VC0_VM_CONTEXT1_* -*/ -int dist = mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - - mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - dist * vmid + hubid *MMHUB_INSTANCE_REGISTER_OFFSET , + hub->ctx_addr_distance * vmid + hubid *MMHUB_INSTANCE_REGISTER_OFFSET ,lower_32_bits (value));WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - dist * vmid + hubid *MMHUB_INSTANCE_REGISTER_OFFSET , + hub->ctx_addr_distance * vmid + hubid *MMHUB_INSTANCE_REGISTER_OFFSET ,upper_32_bits (value)); }
@@ -301,6 +297,7 @@ static void mmhub_v9_4_disable_identity_aperture(struct amdgpu_device *adev,static void mmhub_v9_4_setup_vmid_config (struct amdgpu_device *adev,int hubid) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];uint32_t tmp;int i;
@@ -335,21 +332,25 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)RETRY_PERMISSION_OR_INVALID_PAGE_FAULT , !amdgpu_noretry);WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT1_CNTL, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET + i, - tmp); + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->ctx_distance, tmp);WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET + i*2 ,0 ); + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->ctx_addr_distance,0 );WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET + i*2 ,0 ); + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->ctx_addr_distance,0 );WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET + i*2 , + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->ctx_addr_distance,lower_32_bits (adev->vm_manager.max_pfn -1 ));WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET + i*2 , + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->ctx_addr_distance,upper_32_bits (adev->vm_manager.max_pfn -1 )); } }
@@ -357,16 +358,19 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)static void mmhub_v9_4_program_invalidation (struct amdgpu_device *adev,int hubid) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];unsigned i;for (i =0 ; i <18 ; ++i) {WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_INVALIDATE_ENG0_ADDR_RANGE_LO32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET +2 * i, + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->eng_addr_distance,0xffffffff );WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_INVALIDATE_ENG0_ADDR_RANGE_HI32, - hubid *MMHUB_INSTANCE_REGISTER_OFFSET +2 * i, + hubid *MMHUB_INSTANCE_REGISTER_OFFSET + + i * hub->eng_addr_distance,0x1f ); } }
@@ -395,6 +399,7 @@ int mmhub_v9_4_gart_enable(struct amdgpu_device *adev)void mmhub_v9_4_gart_disable (struct amdgpu_device *adev) { +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0 ];u32 tmp;u32 i, j;
@@ -404,7 +409,7 @@ void mmhub_v9_4_gart_disable(struct amdgpu_device *adev)WREG32_SOC15_OFFSET (MMHUB ,0 , mmVML2VC0_VM_CONTEXT0_CNTL, j *MMHUB_INSTANCE_REGISTER_OFFSET + - i,0 ); + i * hub->ctx_distance,0 );/* Setup TLB control */ tmp = RREG32_SOC15_OFFSET(MMHUB, 0,
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Am 06.07.20 um 07:04 schrieb Huang Rui:
Quoted 6 lines
During IP block updates in some asic, the register distribution pattern of context domains and invalidation engines will be changed in gc_v***.h and mmhub_v***.h. However the register distances are always constant in one IP block beside gfxhub and mmhub. So we can add register distance member in vmhub structure to store the register distance value, that's able to avoid the issue of hardcode at current in future updates.
Quoted 27 lines
Thanks,
RayHuang Rui (12):
drm/amdgpu: add register distance members into vmhub structure drm/amdgpu: use register distance member instead of hardcode in GMC9 drm/amdgpu: use register distance member instead of hardcode in GMC10 drm/amdgpu: use register distance member instead of hardcode inVCN1/JEPG1 drm/amdgpu: use register distance member instead of hardcode in UVD7/VCE4 drm/amdgpu: use register distance member instead of hardcode in VCN2/JEPG2 drm/amdgpu: use register distance member instead of hardcode in gfxhub v1 drm/amdgpu: use register distance member instead of hardcode in gfxhub v2 drm/amdgpu: use register distance member instead of hardcode in gfxhub v2.1 drm/amdgpu: use register distance member instead of hardcode in mmhub v1 drm/amdgpu: use register distance member instead of hardcode in mmhub v2 drm/amdgpu: use register distance member instead of hardcode in mmhub v9.4
Quoted 17 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 9 +++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 48 ++++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 47 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 47 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 43 ++++++++++++--------- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 39 ++++++++++++------- drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 45 ++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 46 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 46 +++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 5 ++- drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 3 +- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 5 ++- drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 5 ++- 15 files changed, 260 insertions(+), 132 deletions(-)
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx