[Beignet] [PATCH] Fix a bug in stack calculation.

[Zhigang Gong]

On Mon, Aug 05, 2013 at 03:14:39PM +0800, Ruiling Song wrote:
> 1. the thread_id is located in r0.5[0-8], so we need to get the correct bits.
For IVB, only r0.5 7:0 is used as FFTID, bit 8 is reserved. You may need to
check the spec again.

> 2. also, we don't need so much stack size, max_compute_unit have already
>    been treated as: #EU * max_thread_per_eu.
Good catch. This part is ok for me.

> 
> Signed-off-by: Ruiling Song <ruiling.song at intel.com>
> ---
>  backend/src/backend/gen_context.cpp |    2 +-
>  src/cl_command_queue_gen7.c         |    1 -
>  2 files changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/backend/src/backend/gen_context.cpp b/backend/src/backend/gen_context.cpp
> index e33d8da..12cc104 100644
> --- a/backend/src/backend/gen_context.cpp
> +++ b/backend/src/backend/gen_context.cpp
> @@ -118,7 +118,7 @@ namespace gbe
>      p->push();
>        p->curr.execWidth = 1;
>        p->curr.predicate = GEN_PREDICATE_NONE;
> -      p->SHR(GenRegister::ud1grf(126,0), GenRegister::ud1grf(0,5), GenRegister::immud(10));
> +      p->AND(GenRegister::ud1grf(126,0), GenRegister::ud1grf(0,5), GenRegister::immud(0x1ff));
>        p->curr.execWidth = this->simdWidth;
>        p->SHL(stackptr, stackptr, GenRegister::immud(perLaneShift));
>        p->curr.execWidth = 1;
> diff --git a/src/cl_command_queue_gen7.c b/src/cl_command_queue_gen7.c
> index 048595c..8933213 100644
> --- a/src/cl_command_queue_gen7.c
> +++ b/src/cl_command_queue_gen7.c
> @@ -180,7 +180,6 @@ cl_bind_stack(cl_gpgpu gpgpu, cl_kernel ker)
>    assert(offset >= 0);
>    stack_sz *= gbe_kernel_get_simd_width(ker->opaque);
>    stack_sz *= device->max_compute_unit;
> -  stack_sz *= device->max_thread_per_unit;
>    cl_gpgpu_set_stack(gpgpu, offset, stack_sz, cc_llc_l3);
>  }
>  
> -- 
> 1.7.9.5
> 
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet