[PATCH] lib/intel_compute_square_kernels: Extend compute test for ptl

[Zbigniew Kempczyński]

On Fri, Feb 14, 2025 at 07:05:18AM +0000, sai.gowtham.ch at intel.com wrote:
> From: Sai Gowtham Ch <sai.gowtham.ch at intel.com>
> 
> Current compute square kernel tests dosen't support on PTL, this enables
> tests to work on PTL.
> 
> CC: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> Signed-off-by: Sai Gowtham Ch <sai.gowtham.ch at intel.com>
> ---
>  lib/intel_compute.c                | 10 +++++++
>  lib/intel_compute_square_kernels.c | 42 ++++++++++++++++++++++++++++++
>  2 files changed, 52 insertions(+)
> 
> diff --git a/lib/intel_compute.c b/lib/intel_compute.c
> index 233835c6c..5350389df 100644
> --- a/lib/intel_compute.c
> +++ b/lib/intel_compute.c
> @@ -1729,6 +1729,16 @@ static const struct {
>  		.compute_exec = xe2lpg_compute_exec,
>  		.compat = COMPAT_DRIVER_XE,
>  	},
> +	{
> +		.ip_ver = IP_VER(30, 00),
> +		.compute_exec = xe2lpg_compute_exec,
> +		.compat = COMPAT_DRIVER_XE,
> +	},
> +	{
> +		.ip_ver = IP_VER(30, 01),
> +		.compute_exec = xe2lpg_compute_exec,
> +		.compat = COMPAT_DRIVER_XE,

This entry seems to be unused at the moment.

> +	},
>  };
>  
>  static bool __run_intel_compute_kernel(int fd,
> diff --git a/lib/intel_compute_square_kernels.c b/lib/intel_compute_square_kernels.c
> index b0912e7e4..18f90c4fe 100644
> --- a/lib/intel_compute_square_kernels.c
> +++ b/lib/intel_compute_square_kernels.c
> @@ -3889,6 +3889,38 @@ unsigned char xelpg_kernel_square_bin[] = {
>  	0x00, 0x00, 0x00, 0x00
>  };
>  
> +static const unsigned char ptl_kernel_square_bin[] = {
> +	0x65, 0xa0, 0x00, 0x80, 0x20, 0x82, 0x05, 0x1f, 0x04, 0x00, 0x00,
> +	0x02, 0xc0, 0xff, 0xff, 0xff, 0x40, 0x19, 0x00, 0x80, 0x20, 0x82,
> +	0x05, 0x1f, 0x04, 0x1f, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x31,
> +	0x22, 0x03, 0x80, 0x00, 0x00, 0x0c, 0x05, 0x8f, 0x1f, 0x00, 0xfa,
> +	0x03, 0x00, 0x70, 0xf6, 0x61, 0x80, 0x24, 0x2c, 0x02, 0x00, 0x10,
> +	0x00, 0x66, 0x09, 0x00, 0x80, 0x20, 0x82, 0x01, 0x80, 0x00, 0x80,
> +	0x00, 0x02, 0xc0, 0x04, 0x00, 0x40, 0x41, 0x22, 0x03, 0x80, 0x60,
> +	0x06, 0x01, 0x20, 0x54, 0x05, 0x00, 0x01, 0x14, 0x02, 0x00, 0x00,
> +	0x53, 0x81, 0x00, 0x80, 0x60, 0x06, 0x05, 0x03, 0x54, 0x05, 0x00,
> +	0x06, 0x14, 0x02, 0x00, 0x00, 0x52, 0x19, 0x14, 0x00, 0x60, 0x06,
> +	0x04, 0x06, 0x04, 0x03, 0x0e, 0x01, 0x04, 0x01, 0x04, 0x04, 0x70,
> +	0x19, 0x14, 0x00, 0x20, 0x02, 0x01, 0x00, 0x04, 0x06, 0x10, 0x52,
> +	0x44, 0x05, 0x00, 0x00, 0x2e, 0x00, 0x14, 0x14, 0x00, 0xc0, 0x00,
> +	0x00, 0x88, 0x00, 0x00, 0x00, 0x88, 0x00, 0x00, 0x00, 0x61, 0x00,
> +	0x00, 0xb4, 0x15, 0x06, 0x10, 0x00, 0x61, 0x00, 0x08, 0xb4, 0x17,
> +	0x07, 0x10, 0x00, 0x69, 0x1a, 0x10, 0x00, 0x70, 0x86, 0x05, 0x19,
> +	0x04, 0x15, 0x20, 0x05, 0x02, 0x00, 0x02, 0x00, 0x69, 0x1a, 0x10,
> +	0x02, 0x70, 0x86, 0x05, 0x1b, 0x04, 0x17, 0x20, 0x05, 0x02, 0x00,
> +	0x02, 0x00, 0x40, 0x1a, 0x00, 0x38, 0x08, 0x19, 0x10, 0x05, 0x40,
> +	0x1a, 0x08, 0x38, 0x0a, 0x1b, 0x10, 0x05, 0x31, 0x23, 0x17, 0x00,
> +	0x00, 0x00, 0x14, 0x0c, 0x24, 0x08, 0x00, 0xfb, 0x00, 0x00, 0x00,
> +	0x00, 0x40, 0x00, 0x00, 0x38, 0x10, 0x19, 0x30, 0x05, 0x40, 0x00,
> +	0x08, 0x38, 0x12, 0x1b, 0x30, 0x05, 0x41, 0x83, 0x20, 0x20, 0x0e,
> +	0x0c, 0x00, 0x0c, 0x31, 0x24, 0x15, 0x00, 0x00, 0x00, 0x00, 0x00,
> +	0x24, 0x10, 0x08, 0xfb, 0x14, 0x0e, 0x00, 0x00, 0x2f, 0x00, 0x14,
> +	0x00, 0x00, 0x40, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x10, 0x00,
> +	0x00, 0x00, 0x61, 0x00, 0x10, 0x28, 0x14, 0x02, 0x10, 0x00, 0x31,
> +	0x20, 0x02, 0x80, 0x04, 0x00, 0x00, 0x00, 0x0c, 0x14, 0x20, 0x30,
> +	0x00, 0x00, 0x00, 0x00
> +};

Shader looks correct.

> +
>  const struct intel_compute_kernels intel_compute_square_kernels[] = {
>  	{
>  		.ip_ver = IP_VER(12, 0),
> @@ -3933,5 +3965,15 @@ const struct intel_compute_kernels intel_compute_square_kernels[] = {
>  		.sip_kernel = xe2lpg_kernel_sip_bin,
>  		.sip_kernel_size = sizeof(xe2lpg_kernel_sip_bin),
>  	},
> +	{
> +		.ip_ver = IP_VER(30, 00),
> +		.size = sizeof(ptl_kernel_square_bin),
> +		.kernel = ptl_kernel_square_bin,
> +	},
> +	{
> +		.ip_ver = IP_VER(30, 01),
> +		.size = sizeof(ptl_kernel_square_bin),
> +		.kernel = ptl_kernel_square_bin,

Same as first comment, there's no rel = 1 for PTL defined
(good argument to start rework for getting ver/rel from hw).

--
Zbigniew

> +	},
>  	{}
>  };
> -- 
> 2.43.0
>