Re: [PATCH 2/2] drm/amdgpu: cleanup VCN IB generation

[Leo Liu <leo.liu-5C7GfCeVMHo@public.gmane.org>]

On 02/09/2018 07:32 AM, Christian König wrote:
> Leo could you give this patch on Raven a try and review/ack it?

Sure.

Leo
>
> I've tested patch #1 on Vega10, but I don't have a stable Raven system 
> to test this one.
>
> Thanks,
> Christian.
>
> Am 07.02.2018 um 20:48 schrieb Christian König:
>> Start to use amdgpu_bo_create_reserved.
>>
>> Signed-off-by: Christian König <christian.koenig@amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 91 
>> ++++++++-------------------------
>>   1 file changed, 21 insertions(+), 70 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> index e86d0b2e9b7a..97bdde9a739e 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> @@ -270,43 +270,28 @@ int amdgpu_vcn_dec_ring_test_ring(struct 
>> amdgpu_ring *ring)
>>       return r;
>>   }
>>   -static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, 
>> struct amdgpu_bo *bo,
>> -                   bool direct, struct dma_fence **fence)
>> +static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
>> +                   struct amdgpu_bo *bo, bool direct,
>> +                   struct dma_fence **fence)
>>   {
>> -    struct ttm_operation_ctx ctx = { true, false };
>> -    struct ttm_validate_buffer tv;
>> -    struct ww_acquire_ctx ticket;
>> -    struct list_head head;
>> +    struct amdgpu_device *adev = ring->adev;
>> +    struct dma_fence *f = NULL;
>>       struct amdgpu_job *job;
>>       struct amdgpu_ib *ib;
>> -    struct dma_fence *f = NULL;
>> -    struct amdgpu_device *adev = ring->adev;
>>       uint64_t addr;
>>       int i, r;
>>   -    memset(&tv, 0, sizeof(tv));
>> -    tv.bo = &bo->tbo;
>> -
>> -    INIT_LIST_HEAD(&head);
>> -    list_add(&tv.head, &head);
>> -
>> -    r = ttm_eu_reserve_buffers(&ticket, &head, true, NULL);
>> -    if (r)
>> -        return r;
>> -
>> -    r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>> -    if (r)
>> -        goto err;
>> -
>>       r = amdgpu_job_alloc_with_ib(adev, 64, &job);
>>       if (r)
>>           goto err;
>>         ib = &job->ibs[0];
>>       addr = amdgpu_bo_gpu_offset(bo);
>> -    ib->ptr[0] = PACKET0(SOC15_REG_OFFSET(UVD, 0, 
>> mmUVD_GPCOM_VCPU_DATA0), 0);
>> +    ib->ptr[0] = PACKET0(SOC15_REG_OFFSET(UVD, 0, 
>> mmUVD_GPCOM_VCPU_DATA0),
>> +                 0);
>>       ib->ptr[1] = addr;
>> -    ib->ptr[2] = PACKET0(SOC15_REG_OFFSET(UVD, 0, 
>> mmUVD_GPCOM_VCPU_DATA1), 0);
>> +    ib->ptr[2] = PACKET0(SOC15_REG_OFFSET(UVD, 0, 
>> mmUVD_GPCOM_VCPU_DATA1),
>> +                 0);
>>       ib->ptr[3] = addr >> 32;
>>       ib->ptr[4] = PACKET0(SOC15_REG_OFFSET(UVD, 0, 
>> mmUVD_GPCOM_VCPU_CMD), 0);
>>       ib->ptr[5] = 0;
>> @@ -330,11 +315,12 @@ static int amdgpu_vcn_dec_send_msg(struct 
>> amdgpu_ring *ring, struct amdgpu_bo *b
>>               goto err_free;
>>       }
>>   -    ttm_eu_fence_buffer_objects(&ticket, &head, f);
>> +    amdgpu_bo_fence(bo, f, false);
>> +    amdgpu_bo_unreserve(bo);
>> +    amdgpu_bo_unref(&bo);
>>         if (fence)
>>           *fence = dma_fence_get(f);
>> -    amdgpu_bo_unref(&bo);
>>       dma_fence_put(f);
>>         return 0;
>> @@ -343,7 +329,8 @@ static int amdgpu_vcn_dec_send_msg(struct 
>> amdgpu_ring *ring, struct amdgpu_bo *b
>>       amdgpu_job_free(job);
>>     err:
>> -    ttm_eu_backoff_reservation(&ticket, &head);
>> +    amdgpu_bo_unreserve(bo);
>> +    amdgpu_bo_unref(&bo);
>>       return r;
>>   }
>>   @@ -351,31 +338,16 @@ static int 
>> amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
>>                     struct dma_fence **fence)
>>   {
>>       struct amdgpu_device *adev = ring->adev;
>> -    struct amdgpu_bo *bo;
>> +    struct amdgpu_bo *bo = NULL;
>>       uint32_t *msg;
>>       int r, i;
>>   -    r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
>> -                 AMDGPU_GEM_DOMAIN_VRAM,
>> -                 AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
>> -                 AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
>> -                 NULL, NULL, &bo);
>> +    r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE,
>> +                      AMDGPU_GEM_DOMAIN_VRAM,
>> +                      &bo, NULL, (void **)&msg);
>>       if (r)
>>           return r;
>>   -    r = amdgpu_bo_reserve(bo, false);
>> -    if (r) {
>> -        amdgpu_bo_unref(&bo);
>> -        return r;
>> -    }
>> -
>> -    r = amdgpu_bo_kmap(bo, (void **)&msg);
>> -    if (r) {
>> -        amdgpu_bo_unreserve(bo);
>> -        amdgpu_bo_unref(&bo);
>> -        return r;
>> -    }
>> -
>>       msg[0] = cpu_to_le32(0x00000028);
>>       msg[1] = cpu_to_le32(0x00000038);
>>       msg[2] = cpu_to_le32(0x00000001);
>> @@ -393,9 +365,6 @@ static int amdgpu_vcn_dec_get_create_msg(struct 
>> amdgpu_ring *ring, uint32_t hand
>>       for (i = 14; i < 1024; ++i)
>>           msg[i] = cpu_to_le32(0x0);
>>   -    amdgpu_bo_kunmap(bo);
>> -    amdgpu_bo_unreserve(bo);
>> -
>>       return amdgpu_vcn_dec_send_msg(ring, bo, true, fence);
>>   }
>>   @@ -407,27 +376,12 @@ static int 
>> amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
>>       uint32_t *msg;
>>       int r, i;
>>   -    r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
>> -                 AMDGPU_GEM_DOMAIN_VRAM,
>> -                 AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
>> -                 AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
>> -                 NULL, NULL, &bo);
>> +    r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE,
>> +                      AMDGPU_GEM_DOMAIN_VRAM,
>> +                      &bo, NULL, (void **)&msg);
>>       if (r)
>>           return r;
>>   -    r = amdgpu_bo_reserve(bo, false);
>> -    if (r) {
>> -        amdgpu_bo_unref(&bo);
>> -        return r;
>> -    }
>> -
>> -    r = amdgpu_bo_kmap(bo, (void **)&msg);
>> -    if (r) {
>> -        amdgpu_bo_unreserve(bo);
>> -        amdgpu_bo_unref(&bo);
>> -        return r;
>> -    }
>> -
>>       msg[0] = cpu_to_le32(0x00000028);
>>       msg[1] = cpu_to_le32(0x00000018);
>>       msg[2] = cpu_to_le32(0x00000000);
>> @@ -437,9 +391,6 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct 
>> amdgpu_ring *ring, uint32_t han
>>       for (i = 6; i < 1024; ++i)
>>           msg[i] = cpu_to_le32(0x0);
>>   -    amdgpu_bo_kunmap(bo);
>> -    amdgpu_bo_unreserve(bo);
>> -
>>       return amdgpu_vcn_dec_send_msg(ring, bo, direct, fence);
>>   }
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx