[Mesa-dev] [PATCH 2/9] glsl: Use the ir_intrinsic_* enums instead of the __intrinsic_* name strings

[Ilia Mirkin]

On Fri, Sep 2, 2016 at 10:37 AM, Ian Romanick <idr at freedesktop.org> wrote:
> On 09/01/2016 11:15 AM, Ilia Mirkin wrote:
>> On Thu, Sep 1, 2016 at 2:10 PM, Ian Romanick <idr at freedesktop.org> wrote:
>>> +      case ir_intrinsic_atomic_counter_comp_swap: {
>>>           opcode = TGSI_OPCODE_ATOMCAS;
>>>           param = param->get_next();
>>>           val = ((ir_instruction *)param)->as_rvalue();
>>>           val->accept(this);
>>>           data2 = this->result;
>>> -      } else if (!strcmp("__intrinsic_atomic_sub", callee)) {
>>> +         break;
>>> +      }
>>> +      case ir_intrinsic_atomic_sub: {
>>
>> I think you meant ir_intrinsic_atomic_counter_sub, and a similar issue
>> in the commit where you remove it.
>
> D'oh.  I only build-tested the final product, so I missed that one.  I
> fix that before I push.

Would you mind fixing it up and pushing it to a branch somewhere? Just
want to make sure you fix it in a way that doesn't end up breaking
something else :) [It's not that I don't trust you, but ... this stuff
is tedious and repetitive and it's easy to miss a spot. And you'll
have to do it anyways.]

>
>>>           opcode = TGSI_OPCODE_ATOMUADD;
>>>           st_src_reg res = get_temp(glsl_type::uvec4_type);
>>>           st_dst_reg dstres = st_dst_reg(res);
>>>           dstres.writemask = dst.writemask;
>>>           emit_asm(ir, TGSI_OPCODE_INEG, dstres, data);
>>>           data = res;
>>> -      } else {
>>> +         break;
>>> +      }
>