drm/amdgpu: clean up coding style a bit
Subsystems (2)
PATCH 1/4 drm/amdgpu: clean up coding style a bit
No need to bitcast a boolean and even if we should use "!!" instead.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +10 -10
@@ -328,35 +328,35 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_filecase AMDGPU_HW_IP_GFX : type =AMD_IP_BLOCK_TYPE_GFX ;for (i =0 ; i < adev->gfx.num_gfx_rings; i++) - ring_mask |= ((adev->gfx.gfx_ring[i].ready ?1 :0 ) << i); + ring_mask |= adev->gfx.gfx_ring[i].ready << i; ib_start_alignment =32 ; ib_size_alignment =32 ;break ;case AMDGPU_HW_IP_COMPUTE : type =AMD_IP_BLOCK_TYPE_GFX ;for (i =0 ; i < adev->gfx.num_compute_rings; i++) - ring_mask |= ((adev->gfx.compute_ring[i].ready ?1 :0 ) << i); + ring_mask |= adev->gfx.compute_ring[i].ready << i; ib_start_alignment =32 ; ib_size_alignment =32 ;break ;case AMDGPU_HW_IP_DMA : type =AMD_IP_BLOCK_TYPE_SDMA ;for (i =0 ; i < adev->sdma.num_instances; i++) - ring_mask |= ((adev->sdma.instance[i].ring.ready ?1 :0 ) << i); + ring_mask |= adev->sdma.instance[i].ring.ready << i; ib_start_alignment =256 ; ib_size_alignment =4 ;break ;case AMDGPU_HW_IP_UVD : type =AMD_IP_BLOCK_TYPE_UVD ;for (i =0 ; i < adev->uvd.num_uvd_inst; i++) - ring_mask |= ((adev->uvd.inst[i].ring.ready ?1 :0 ) << i); + ring_mask |= adev->uvd.inst[i].ring.ready << i; ib_start_alignment =64 ; ib_size_alignment =64 ;break ;case AMDGPU_HW_IP_VCE : type =AMD_IP_BLOCK_TYPE_VCE ;for (i =0 ; i < adev->vce.num_rings; i++) - ring_mask |= ((adev->vce.ring[i].ready ?1 :0 ) << i); + ring_mask |= adev->vce.ring[i].ready << i; ib_start_alignment =4 ; ib_size_alignment =1 ;break ;
@@ -365,27 +365,27 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_filefor (i =0 ; i < adev->uvd.num_uvd_inst; i++)for (j =0 ; j < adev->uvd.num_enc_rings; j++) ring_mask |= - ((adev->uvd.inst[i].ring_enc[j].ready ?1 :0 ) << - (j + i * adev->uvd.num_enc_rings)); + adev->uvd.inst[i].ring_enc[j].ready << + (j + i * adev->uvd.num_enc_rings); ib_start_alignment =64 ; ib_size_alignment =64 ;break ;case AMDGPU_HW_IP_VCN_DEC : type =AMD_IP_BLOCK_TYPE_VCN ; - ring_mask = adev->vcn.ring_dec.ready ?1 :0 ; + ring_mask = adev->vcn.ring_dec.ready; ib_start_alignment =16 ; ib_size_alignment =16 ;break ;case AMDGPU_HW_IP_VCN_ENC : type =AMD_IP_BLOCK_TYPE_VCN ;for (i =0 ; i < adev->vcn.num_enc_rings; i++) - ring_mask |= ((adev->vcn.ring_enc[i].ready ?1 :0 ) << i); + ring_mask |= adev->vcn.ring_enc[i].ready << i; ib_start_alignment =64 ; ib_size_alignment =1 ;break ;case AMDGPU_HW_IP_VCN_JPEG : type =AMD_IP_BLOCK_TYPE_VCN ; - ring_mask = adev->vcn.ring_jpeg.ready ?1 :0 ; + ring_mask = adev->vcn.ring_jpeg.ready; ib_start_alignment =16 ; ib_size_alignment =16 ;break ;
Signature · 6 lines
2.14.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 20180723151441.46790-1-christian.koenig-5C7GfCeVMHo@public.gmane.org
PATCH 2/4 drm/amdgpu: expose only the first UVD instance for now
Going to completely rework the context to ring mapping with Nayan's GSoC work, but for now just stopping to expose the second UVD instance should do it.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 13 +++++-------- drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c | 9 ++------- 2 files changed, 7 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +5 -8
@@ -286,7 +286,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_filestruct drm_crtc *crtc;uint32_t ui32 =0 ;uint64_t ui64 =0 ; -int i, j, found; +int i, found;int ui32_size =sizeof (ui32);if (!info->return_size || !info->return_pointer)
@@ -348,8 +348,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_filebreak ;case AMDGPU_HW_IP_UVD : type =AMD_IP_BLOCK_TYPE_UVD ; -for (i =0 ; i < adev->uvd.num_uvd_inst; i++) - ring_mask |= adev->uvd.inst[i].ring.ready << i; + ring_mask |= adev->uvd.inst[0 ].ring.ready; ib_start_alignment =64 ; ib_size_alignment =64 ;break ;
@@ -362,11 +361,9 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_filebreak ;case AMDGPU_HW_IP_UVD_ENC : type =AMD_IP_BLOCK_TYPE_UVD ; -for (i =0 ; i < adev->uvd.num_uvd_inst; i++) -for (j =0 ; j < adev->uvd.num_enc_rings; j++) - ring_mask |= - adev->uvd.inst[i].ring_enc[j].ready << - (j + i * adev->uvd.num_enc_rings); +for (i =0 ; i < adev->uvd.num_enc_rings; i++) + ring_mask |= + adev->uvd.inst[0 ].ring_enc[i].ready << i; ib_start_alignment =64 ; ib_size_alignment =64 ;break ;
drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +2 -7
@@ -66,8 +66,6 @@ static int amdgpu_identity_map(struct amdgpu_device *adev, u32 ring, struct amdgpu_ring **out_ring) { - u32 instance; -switch (mapper->hw_ip) {case AMDGPU_HW_IP_GFX : *out_ring = &adev->gfx.gfx_ring[ring];
@@ -79,16 +77,13 @@ static int amdgpu_identity_map(struct amdgpu_device *adev, *out_ring = &adev->sdma.instance[ring].ring;break ;case AMDGPU_HW_IP_UVD : - instance = ring; - *out_ring = &adev->uvd.inst[instance].ring; + *out_ring = &adev->uvd.inst[0 ].ring;break ;case AMDGPU_HW_IP_VCE : *out_ring = &adev->vce.ring[ring];break ;case AMDGPU_HW_IP_UVD_ENC : - instance = ring / adev->uvd.num_enc_rings; - *out_ring = - &adev->uvd.inst[instance].ring_enc[ring%adev->uvd.num_enc_rings]; + *out_ring = &adev->uvd.inst[0 ].ring_enc[ring];break ;case AMDGPU_HW_IP_VCN_DEC : *out_ring = &adev->vcn.ring_dec;
Signature · 6 lines
2.14.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 3/4 drm/amdgpu: add support for inplace IB patching for MM engines
In reply to a message not in the archiveWe are going to need that for the second UVD instance on Vega20.
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 22 +++++++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + 3 files changed, 17 insertions(+), 7 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
@@ -1751,6 +1751,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)#define amdgpu_vm_write_pte (adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))#define amdgpu_vm_set_pte_pde (adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))#define amdgpu_ring_parse_cs (r, p, ib) ((r)->funcs->parse_cs((p), (ib))) +#define amdgpu_ring_patch_cs (r, p, ib) ((r)->funcs->patch_cs((p), (ib)))#define amdgpu_ring_test_ring (r) (r)->funcs->test_ring((r))#define amdgpu_ring_test_ib (r, t) (r)->funcs->test_ib((r), (t))#define amdgpu_ring_get_rptr (r) (r)->funcs->get_rptr((r))
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +15 -7
@@ -916,7 +916,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,int r;/* Only for UVD/VCE VM emulation */ -if (p->ring->funcs->parse_cs) { +if (p->ring->funcs->parse_cs || p->ring->funcs->patch_cs) {unsigned i, j;for (i =0 , j =0 ; i < p->nchunks && j < p->job->num_ibs; i++) {
@@ -957,12 +957,20 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, offset = m->start *AMDGPU_GPU_PAGE_SIZE ; kptr += va_start - offset; -memcpy (ib->ptr, kptr, chunk_ib->ib_bytes); -amdgpu_bo_kunmap (aobj); - - r =amdgpu_ring_parse_cs (ring, p, j); -if (r) -return r; +if (p->ring->funcs->parse_cs) { +memcpy (ib->ptr, kptr, chunk_ib->ib_bytes); +amdgpu_bo_kunmap (aobj); + + r =amdgpu_ring_parse_cs (ring, p, j); +if (r) +return r; + }else { + ib->ptr = (uint32_t *)kptr; + r =amdgpu_ring_patch_cs (ring, p, j); +amdgpu_bo_kunmap (aobj); +if (r) +return r; + } j++; }
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +1 -0
@@ -123,6 +123,7 @@ struct amdgpu_ring_funcs {void (*set_wptr)(struct amdgpu_ring *ring);/* validating and patching of IBs */ int (*parse_cs)(struct amdgpu_cs_parser *p,uint32_t ib_idx); +int (*patch_cs)(struct amdgpu_cs_parser *p,uint32_t ib_idx);/* constants to calculate how many DW are needed for an emit */ unsigned emit_frame_size;unsigned emit_ib_size;
Signature · 6 lines
2.14.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 20180723151441.46790-3-christian.koenig-5C7GfCeVMHo@public.gmane.org
Alex Deucher
Reviewed-byRe: [PATCH 3/4] drm/amdgpu: add support for inplace IB patching for MM engines
Quoted 3 lines · Christian König
We are going to need that for the second UVD instance on Vega20.
Signed-off-by Christian König <[email protected]>
A couple of nits below. Patches 1, 2, 4 are:
Quoted 39 lines
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 22 +++++++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + 3 files changed, 17 insertions(+), 7 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-0@@ -1751,6 +1751,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)#define amdgpu_vm_write_pte (adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))#define amdgpu_vm_set_pte_pde (adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))#define amdgpu_ring_parse_cs (r, p, ib) ((r)->funcs->parse_cs((p), (ib))) +#define amdgpu_ring_patch_cs (r, p, ib) ((r)->funcs->patch_cs((p), (ib)))#define amdgpu_ring_test_ring (r) (r)->funcs->test_ring((r))#define amdgpu_ring_test_ib (r, t) (r)->funcs->test_ib((r), (t))#define amdgpu_ring_get_rptr (r) (r)->funcs->get_rptr((r))drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +1
-5@@ -916,7 +916,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,int r;/* Only for UVD/VCE VM emulation */ -if (p->ring->funcs->parse_cs) { +if (p->ring->funcs->parse_cs || p->ring->funcs->patch_cs) {unsigned i, j;for (i =0 , j =0 ; i < p->nchunks && j < p->job->num_ibs; i++) {@@ -957,12 +957,20 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, offset = m->start *AMDGPU_GPU_PAGE_SIZE ; kptr += va_start - offset; -memcpy (ib->ptr, kptr, chunk_ib->ib_bytes); -amdgpu_bo_kunmap (aobj); - - r =amdgpu_ring_parse_cs (ring, p, j);
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, · line 967
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +8
-2-if (r) -return r; +if (p->ring->funcs->parse_cs) { +memcpy (ib->ptr, kptr, chunk_ib->ib_bytes); +amdgpu_bo_kunmap (aobj); + + r =amdgpu_ring_parse_cs (ring, p, j); +if (r) +return r; + }else {
Maybe make this if (p->ring->funcs->patch_cs) {? Not a big deal either way. I doubt we'd every have both callbacks defined.
Quoted 14 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +6
-0+ ib->ptr = (uint32_t *)kptr; + r =amdgpu_ring_patch_cs (ring, p, j); +amdgpu_bo_kunmap (aobj); +if (r) +return r; + } j++; }drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h struct amdgpu_ring_funcs { · line 126
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +1
-0@@ -123,6 +123,7 @@ struct amdgpu_ring_funcs {void (*set_wptr)(struct amdgpu_ring *ring);/* validating and patching of IBs */ int (*parse_cs)(struct amdgpu_cs_parser *p,uint32_t ib_idx); +int (*patch_cs)(struct amdgpu_cs_parser *p,uint32_t ib_idx);
Can we come up with a better name? parse_vm_cs() or patch_in_place_cs() or something like that for consistency?
Alex
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +0
-1/* constants to calculate how many DW are needed for an emit */ unsigned emit_frame_size;unsigned emit_ib_size; --2.14.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/4 drm/amdgpu: patch the IBs for the second UVD instance v2
In reply to a message not in the archivePatch the IBs for the second UVD instance so that userspace don't need to care about the instance they submit to.
v2: use direct IB patching
--- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +27 -0
@@ -1205,6 +1205,32 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring)return r; } +/** +* uvd_v7_0_enc_ring_test_ring - test if UVD ENC ring is working +* +* @ring: the engine to test on +* +*/ +static int uvd_v7_0_ring_patch_cs (struct amdgpu_cs_parser *p,uint32_t ib_idx) +{ +struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; +unsigned i; + +/* No patching necessary for the first instance */ +if (!p->ring->me) +return 0 ; + +for (i =0 ; i < ib->length_dw; i +=2 ) { +uint32_t reg =amdgpu_get_ib_value (p, ib_idx, i); + + reg -= p->adev->reg_offset[UVD_HWIP ][0 ][1 ]; + reg += p->adev->reg_offset[UVD_HWIP ][1 ][1 ]; + +amdgpu_set_ib_value (p, ib_idx, i, reg); + } +return 0 ; +} +/** * uvd_v7_0_ring_emit_ib - execute indirect buffer *
@@ -1697,6 +1723,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .get_rptr = uvd_v7_0_ring_get_rptr, .get_wptr = uvd_v7_0_ring_get_wptr, .set_wptr = uvd_v7_0_ring_set_wptr, + .patch_cs = uvd_v7_0_ring_patch_cs, .emit_frame_size =6 +/* hdp invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
Signature · 6 lines
2.14.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx