[PATCH] drm/amdgpu: set vm size and block size by individual gmc by default

Christian König deathsimple at vodafone.de
Wed Apr 5 09:01:04 UTC 2017


Am 05.04.2017 um 08:43 schrieb Junwei Zhang:
> By default, the value is set by individual gmc.
> if a specific value is input, it overrides the global value for all
>
> Signed-off-by: Junwei Zhang <Jerry.Zhang at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  2 ++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 44 ++++++++++++++++++------------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c    |  2 +-
>   drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c      | 10 +++++--
>   drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c      | 10 +++++--
>   drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c      | 10 +++++--
>   drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c      | 30 ++++++++++++++------
>   7 files changed, 74 insertions(+), 34 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index 65021df..d7f75ce 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -1647,6 +1647,8 @@ void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
>   bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type);
>   bool amdgpu_device_has_dc_support(struct amdgpu_device *adev);
>   
> +uint32_t amdgpu_get_block_size(int vm_size);
> +
>   /*
>    * Registers read & write functions.
>    */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 1d0c742..2f91c2f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -1039,35 +1039,44 @@ static bool amdgpu_check_pot_argument(int arg)
>   	return (arg & (arg - 1)) == 0;
>   }
>   
> -static void amdgpu_get_block_size(struct amdgpu_device *adev)
> +uint32_t amdgpu_get_block_size(int vm_size)
> +{
> +	/* Total bits covered by PD + PTs */
> +	unsigned bits = ilog2(vm_size) + 18;
> +
> +	/* Make sure the PD is 4K in size up to 8GB address space.
> +	   Above that split equal between PD and PTs */
> +	if (vm_size <= 8)
> +		return (bits - 9);
> +	else
> +		return ((bits + 3) / 2);
> +}

Please move that helper into amdgpu_vmc.c.

> +
> +static void amdgpu_check_block_size(struct amdgpu_device *adev)
>   {
>   	/* defines number of bits in page table versus page directory,
>   	 * a page is 4KB so we have 12 bits offset, minimum 9 bits in the
>   	 * page table and the remaining bits are in the page directory */
> -	if (amdgpu_vm_block_size == -1) {
> -
> -		/* Total bits covered by PD + PTs */
> -		unsigned bits = ilog2(amdgpu_vm_size) + 18;
> -
> -		/* Make sure the PD is 4K in size up to 8GB address space.
> -		   Above that split equal between PD and PTs */
> -		if (amdgpu_vm_size <= 8)
> -			amdgpu_vm_block_size = bits - 9;
> -		else
> -			amdgpu_vm_block_size = (bits + 3) / 2;
> +	if (amdgpu_vm_block_size == -1)
> +		return;
>   
> -	} else if (amdgpu_vm_block_size < 9) {
> +	if (amdgpu_vm_block_size < 9) {
>   		dev_warn(adev->dev, "VM page table size (%d) too small\n",
>   			 amdgpu_vm_block_size);
> -		amdgpu_vm_block_size = 9;
> +		goto def_value;
>   	}
>   
>   	if (amdgpu_vm_block_size > 24 ||
>   	    (amdgpu_vm_size * 1024) < (1ull << amdgpu_vm_block_size)) {
>   		dev_warn(adev->dev, "VM page table size (%d) too large\n",
>   			 amdgpu_vm_block_size);
> -		amdgpu_vm_block_size = 9;
> +		goto def_value;
>   	}
> +
> +	return;
> +
> +def_value:
> +	amdgpu_vm_block_size = -1;
>   }
>   
>   static void amdgpu_check_vm_size(struct amdgpu_device *adev)
> @@ -1096,8 +1105,7 @@ static void amdgpu_check_vm_size(struct amdgpu_device *adev)
>   	return;
>   
>   def_value:
> -	amdgpu_vm_size = 8;
> -	dev_info(adev->dev, "set default VM size %dGB\n", amdgpu_vm_size);
> +	amdgpu_vm_size = -1;;
>   }
>   
>   /**
> @@ -1131,7 +1139,7 @@ static void amdgpu_check_arguments(struct amdgpu_device *adev)
>   
>   	amdgpu_check_vm_size(adev);
>   
> -	amdgpu_get_block_size(adev);
> +	amdgpu_check_block_size(adev);
>   
>   	if (amdgpu_vram_page_split != -1 && (amdgpu_vram_page_split < 16 ||
>   	    !amdgpu_check_pot_argument(amdgpu_vram_page_split))) {
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> index bfd945b..6238e2e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> @@ -86,7 +86,7 @@
>   unsigned amdgpu_ip_block_mask = 0xffffffff;
>   int amdgpu_bapm = -1;
>   int amdgpu_deep_color = 0;
> -int amdgpu_vm_size = 64;
> +int amdgpu_vm_size = -1;
>   int amdgpu_vm_block_size = -1;
>   int amdgpu_vm_fault_stop = 0;
>   int amdgpu_vm_debug = 0;
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
> index 30d5c42..f46c52c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
> @@ -849,8 +849,14 @@ static int gmc_v6_0_sw_init(void *handle)
>   	if (r)
>   		return r;
>   
> -	adev->vm_manager.vm_size = amdgpu_vm_size;
> -	adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	if (amdgpu_vm_size == -1) {
> +		adev->vm_manager.vm_size = 64;
> +		adev->vm_manager.block_size =
> +			amdgpu_get_block_size(adev->vm_manager.vm_size);
> +	} else {
> +		adev->vm_manager.vm_size = amdgpu_vm_size;
> +		adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	}
>   	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
>   
>   	DRM_INFO("vm size is %d GB, block size is %d-bit\n",
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
> index 7113765..80b31f4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
> @@ -1003,8 +1003,14 @@ static int gmc_v7_0_sw_init(void *handle)
>   	 * Currently set to 4GB ((1 << 20) 4k pages).
>   	 * Max GPUVM size for cayman and SI is 40 bits.
>   	 */
> -	adev->vm_manager.vm_size = amdgpu_vm_size;
> -	adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	if (amdgpu_vm_size == -1) {
> +		adev->vm_manager.vm_size = 64;
> +		adev->vm_manager.block_size =
> +			amdgpu_get_block_size(adev->vm_manager.vm_size);
> +	} else {
> +		adev->vm_manager.vm_size = amdgpu_vm_size;
> +		adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	}

That's a bit to simple. The logic should probably rather be:

if (amdgpu_vm_size == -1)
     adev->vm_manager.vm_size = 64;
else
     adev->vm_manager.vm_size = amdgpu_vm_size;

if (amdgpu_vm_block_size == -1)
     adev->vm_manager.block_size = 
amdgpu_get_block_size(adev->vm_manager.vm_size);
else
     adev->vm_manager.block_size = amdgpu_vm_block_size;


Since the logic is the same for GFX6,7 and 8 we might put that into a 
helper as well.

>   	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
>   
>   	DRM_INFO("vm size is %d GB, block size is %d-bit\n",
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> index b3d1f1b..f1f135a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> @@ -1087,8 +1087,14 @@ static int gmc_v8_0_sw_init(void *handle)
>   	 * Currently set to 4GB ((1 << 20) 4k pages).
>   	 * Max GPUVM size for cayman and SI is 40 bits.
>   	 */
> -	adev->vm_manager.vm_size = amdgpu_vm_size;
> -	adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	if (amdgpu_vm_size == -1) {
> +		adev->vm_manager.vm_size = 64;
> +		adev->vm_manager.block_size =
> +			amdgpu_get_block_size(adev->vm_manager.vm_size);
> +	} else {
> +		adev->vm_manager.vm_size = amdgpu_vm_size;
> +		adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	}
>   	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
>   
>   	DRM_INFO("vm size is %d GB, block size is %d-bit\n",
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> index 0cae7f0..8e12ff3 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> @@ -532,18 +532,9 @@ static int gmc_v9_0_sw_init(void *handle)
>   
>   	if (adev->flags & AMD_IS_APU) {
>   		adev->mc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
> -		adev->vm_manager.vm_size = amdgpu_vm_size;
> -		adev->vm_manager.block_size = amdgpu_vm_block_size;
>   	} else {
>   		/* XXX Don't know how to get VRAM type yet. */
>   		adev->mc.vram_type = AMDGPU_VRAM_TYPE_HBM;
> -		/*
> -		 * To fulfill 4-level page support,
> -		 * vm size is 256TB (48bit), maximum size of Vega10,
> -		 * block size 512 (9bit)
> -		 */
> -		adev->vm_manager.vm_size = 1U << 18;
> -		adev->vm_manager.block_size = 9;
>   	}
>   
>   	DRM_INFO("vm size is %d GB, block size is %d-bit\n",
> @@ -558,6 +549,27 @@ static int gmc_v9_0_sw_init(void *handle)
>   	if (r)
>   		return r;
>   
> +	if (amdgpu_vm_size == -1) {
> +		if (adev->flags & AMD_IS_APU) {
> +			adev->vm_manager.vm_size = 64;
> +			adev->vm_manager.block_size =
> +				amdgpu_get_block_size(adev->vm_manager.vm_size);
> +		} else {
> +			/*
> +			 * To fulfill 4-level page support,
> +			 * vm size is 256TB (48bit), maximum size of Vega10,
> +			 * block size 512 (9bit)
> +			 */
> +			adev->vm_manager.vm_size = 1U << 18;
> +			adev->vm_manager.block_size = 9;
> +		}
> +	} else {
> +		adev->vm_manager.vm_size = amdgpu_vm_size;
> +		adev->vm_manager.block_size = amdgpu_vm_block_size;
> +	}
> +
> +	adev->vm_manager.vm_size = amdgpu_vm_size;
> +	adev->vm_manager.block_size = amdgpu_vm_block_size;

NAK. On GFX9 vm_size and block_size should be hard coded to 256TB and 9.

Christian.

>   	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
>   
>   	/* Set the internal MC address mask




More information about the amd-gfx mailing list