[PATCH i-g-t 1/3] lib/intel_compute: Align objects on 64K for xehp

[Zbigniew Kempczyński]

On Mon, Feb 03, 2025 at 02:38:46PM +0100, Francois Dugast wrote:
> Some of the buffers are aligned on 4K. Ensure 64K alignment
> in preparation for a switch to using xe_bo_create() during
> buffers creation.
> 
> Signed-off-by: Francois Dugast <francois.dugast at intel.com>
> ---
>  lib/intel_compute.c | 24 ++++++++++++------------
>  1 file changed, 12 insertions(+), 12 deletions(-)
> 
> diff --git a/lib/intel_compute.c b/lib/intel_compute.c
> index f1520aad4..983b2e2c3 100644
> --- a/lib/intel_compute.c
> +++ b/lib/intel_compute.c
> @@ -24,9 +24,9 @@
>  #define MEDIA_STATE_FLUSH		0x0
>  #define MAX(X, Y)			(((X) > (Y)) ? (X) : (Y))
>  #define SIZE_DATA			64
> -#define SIZE_BATCH			0x1000
> -#define SIZE_BUFFER_INPUT		MAX(sizeof(float) * SIZE_DATA, 0x1000)
> -#define SIZE_BUFFER_OUTPUT		MAX(sizeof(float) * SIZE_DATA, 0x1000)
> +#define SIZE_BATCH			0x10000
> +#define SIZE_BUFFER_INPUT		MAX(sizeof(float) * SIZE_DATA, 0x10000)
> +#define SIZE_BUFFER_OUTPUT		MAX(sizeof(float) * SIZE_DATA, 0x10000)
>  #define ADDR_BATCH			0x100000ULL
>  #define ADDR_INPUT			0x200000ULL
>  #define ADDR_OUTPUT			0x300000ULL
> @@ -38,7 +38,7 @@
>  
>  #define ADDR_GENERAL_STATE_BASE		0x80000000ULL
>  #define ADDR_INSTRUCTION_STATE_BASE	0x90000000ULL
> -#define OFFSET_BINDING_TABLE		0x1000
> +#define OFFSET_BINDING_TABLE		0x10000
>  
>  #define XE2_ADDR_STATE_CONTEXT_DATA_BASE	0x900000ULL
>  #define OFFSET_STATE_SIP			0xFFFF0000
> @@ -746,8 +746,8 @@ static void xehp_create_surface_state(uint32_t *addr_bo_buffer_batch,
>  	addr_bo_buffer_batch[b++] = 0x00000000;
>  	addr_bo_buffer_batch[b++] = 0x00000000;
>  
> -	addr_bo_buffer_batch[b++] = 0x00001000;
> -	addr_bo_buffer_batch[b++] = 0x00001040;
> +	addr_bo_buffer_batch[b++] = 0x00010000;
> +	addr_bo_buffer_batch[b++] = 0x00010040;
>  	addr_bo_buffer_batch[b++] = 0x00000000;
>  	addr_bo_buffer_batch[b++] = 0x00000000;
>  	addr_bo_buffer_batch[b++] = 0x00000000;
> @@ -825,7 +825,7 @@ static void xehp_compute_exec_compute(uint32_t *addr_bo_buffer_batch,
>  	addr_bo_buffer_batch[b++] = GEN8_3DSTATE_BINDING_TABLE_POOL_ALLOC | 2;
>  	addr_bo_buffer_batch[b++] = (addr_surface_state_base & 0xffffffff) | 0x6;
>  	addr_bo_buffer_batch[b++] = addr_surface_state_base >> 32;
> -	addr_bo_buffer_batch[b++] = 0x00002000;
> +	addr_bo_buffer_batch[b++] = 0x00012000;

I've checked this and it seems this line should be removed
(0x001ff000 is enough and pool alloc has 4 and not 5 dwords).
I mean this dword stays here by accident during reversing.
Rest looks good to me. With this nit:

Reviewed-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>

--
Zbigniew

>  	addr_bo_buffer_batch[b++] = 0x001ff000;
>  
>  	addr_bo_buffer_batch[b++] = XEHP_COMPUTE_WALKER | 0x25;
> @@ -852,7 +852,7 @@ static void xehp_compute_exec_compute(uint32_t *addr_bo_buffer_batch,
>  	addr_bo_buffer_batch[b++] = 0x00000000;
>  	addr_bo_buffer_batch[b++] = 0x00180000;
>  	addr_bo_buffer_batch[b++] = 0x00000000;
> -	addr_bo_buffer_batch[b++] = 0x00001080;
> +	addr_bo_buffer_batch[b++] = 0x00010080;
>  	addr_bo_buffer_batch[b++] = 0x0c000002;
>  
>  	addr_bo_buffer_batch[b++] = 0x00000008;
> @@ -895,10 +895,10 @@ static void xehp_compute_exec(int fd, const unsigned char *kernel,
>  		  .size = 0x100000,
>  		  .name = "dynamic state base"},
>  		{ .addr = ADDR_SURFACE_STATE_BASE,
> -		  .size = 0x1000,
> +		  .size = 0x10000,
>  		  .name = "surface state base"},
>  		{ .addr = ADDR_GENERAL_STATE_BASE + OFFSET_INDIRECT_DATA_START,
> -		  .size =  0x1000,
> +		  .size =  0x10000,
>  		  .name = "indirect object base"},
>  		{ .addr = ADDR_INPUT, .size = SIZE_BUFFER_INPUT,
>  		  .name = "addr input"},
> @@ -907,7 +907,7 @@ static void xehp_compute_exec(int fd, const unsigned char *kernel,
>  		{ .addr = ADDR_GENERAL_STATE_BASE, .size = 0x100000,
>  		  .name = "general state base" },
>  		{ .addr = ADDR_SURFACE_STATE_BASE + OFFSET_BINDING_TABLE,
> -		  .size = 0x1000,
> +		  .size = 0x10000,
>  		  .name = "binding table" },
>  		{ .addr = ADDR_BATCH, .size = SIZE_BATCH,
>  		  .name = "batch" },
> @@ -918,7 +918,7 @@ static void xehp_compute_exec(int fd, const unsigned char *kernel,
>  	bo_execenv_create(fd, &execenv, eci);
>  
>  	/* Sets Kernel size */
> -	bo_dict[0].size = ALIGN(size, 0x1000);
> +	bo_dict[0].size = ALIGN(size, 0x10000);
>  
>  	bo_execenv_bind(&execenv, bo_dict, XEHP_BO_DICT_ENTRIES);
>  
> -- 
> 2.43.0
>