[Beignet] [PATCH v2] Change the order of the code

[Zhigang Gong]

On Fri, Feb 21, 2014 at 04:25:20PM +0800, Xiuli Pan wrote:
> From: xiuli pan <xiuli.pan at intel.com>
> 
> Fix the 66K problem in the OpenCV testing.
> The bug was casued by the incorrect order
> of the code, it will result the beignet to
> calculate the whole localsize of the kernel
> file. Now the OpenCV test can pass.
> ---
>  backend/src/llvm/llvm_gen_backend.cpp |   21 +++++++++++----------
>  1 file changed, 11 insertions(+), 10 deletions(-)
> 
> diff --git a/backend/src/llvm/llvm_gen_backend.cpp b/backend/src/llvm/llvm_gen_backend.cpp
> index 2d4fb0a..e20f810 100644
> --- a/backend/src/llvm/llvm_gen_backend.cpp
> +++ b/backend/src/llvm/llvm_gen_backend.cpp
> @@ -1394,16 +1394,7 @@ namespace gbe
>  
>        ir::AddressSpace addrSpace = addressSpaceLLVMToGen(v.getType()->getAddressSpace());
>        if(addrSpace == ir::MEM_LOCAL) {
> -        ir::Function &f = ctx.getFunction();
> -        f.setUseSLM(true);
> -        const Constant *c = v.getInitializer();
> -        Type *ty = c->getType();
> -        uint32_t oldSlm = f.getSLMSize();
> -        uint32_t align = 8 * getAlignmentByte(unit, ty);
> -        uint32_t padding = getPadding(oldSlm*8, align);
> -
> -        f.setSLMSize(oldSlm + padding/8 + getTypeByteSize(unit, ty));
> -        const Value * val = cast<Value>(&v);
> +	const Value * val = cast<Value>(&v);
It seems that you forgot to fix this indent issue? Anyway, I will fix that and then push it latter.
Thanks for the patch.