[Mesa-dev] [PATCH] radv: fix adjusting vertex fetches since 16bit support

[Bas Nieuwenhuizen]

On Wed, Jul 25, 2018 at 3:28 PM, Samuel Pitoiset
<samuel.pitoiset at gmail.com> wrote:
>
>
> On 07/25/2018 03:26 PM, Bas Nieuwenhuizen wrote:
>>
>> On Wed, Jul 25, 2018 at 3:25 PM, Samuel Pitoiset
>> <samuel.pitoiset at gmail.com> wrote:
>>>
>>>
>>>
>>> On 07/25/2018 02:55 PM, Bas Nieuwenhuizen wrote:
>>>>
>>>>
>>>> hmm, not sure why that did not fail on my CTS runs ...
>>>
>>>
>>>
>>> Because 16bit is only enabled on Polaris?
>>
>> Wait what, it should be enabled on Vega too right?
>
>
> It should but it's broken currently. Daniel will fix it later.
hmm?

Also the integer conversion should still kick in on vega right? Your
failed test is not a 16-bit test.

>
>
>>
>>>
>>>
>>>>
>>>> Reviewed-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
>>>>
>>>> On Wed, Jul 25, 2018 at 2:55 PM, Samuel Pitoiset
>>>> <samuel.pitoiset at gmail.com> wrote:
>>>>>
>>>>>
>>>>> Move the integer conversion after the fixup.
>>>>>
>>>>> This fixes some regressions with
>>>>> dEQP-VK.pipeline.vertex_input.single_attribute.mat4.as_a2r10g10b10*
>>>>>
>>>>> Fixes: b722b29f10 ("radv: add support for 16bit input/output")
>>>>> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
>>>>> ---
>>>>>    src/amd/vulkan/radv_nir_to_llvm.c | 7 ++++---
>>>>>    1 file changed, 4 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/src/amd/vulkan/radv_nir_to_llvm.c
>>>>> b/src/amd/vulkan/radv_nir_to_llvm.c
>>>>> index 9f9dc0d4fe..64b6522cd9 100644
>>>>> --- a/src/amd/vulkan/radv_nir_to_llvm.c
>>>>> +++ b/src/amd/vulkan/radv_nir_to_llvm.c
>>>>> @@ -2030,15 +2030,16 @@ handle_vs_input_decl(struct radv_shader_context
>>>>> *ctx,
>>>>>                                   output[chan] =
>>>>> LLVMBuildBitCast(ctx->ac.builder, output[chan], ctx->ac.f32, "");
>>>>>                                   output[chan] =
>>>>> LLVMBuildFPTrunc(ctx->ac.builder, output[chan], ctx->ac.f16, "");
>>>>>                           }
>>>>> -                       output[chan] = ac_to_integer(&ctx->ac,
>>>>> output[chan]);
>>>>> -                       if (type == GLSL_TYPE_UINT16 || type ==
>>>>> GLSL_TYPE_INT16)
>>>>> -                               output[chan] =
>>>>> LLVMBuildTrunc(ctx->ac.builder, output[chan], ctx->ac.i16, "");
>>>>>                   }
>>>>>
>>>>>                   unsigned alpha_adjust =
>>>>> (ctx->options->key.vs.alpha_adjust >> (attrib_index * 2)) & 3;
>>>>>                   output[3] = adjust_vertex_fetch_alpha(ctx,
>>>>> alpha_adjust,
>>>>> output[3]);
>>>>>
>>>>>                   for (unsigned chan = 0; chan < 4; chan++) {
>>>>> +                       output[chan] = ac_to_integer(&ctx->ac,
>>>>> output[chan]);
>>>>> +                       if (type == GLSL_TYPE_UINT16 || type ==
>>>>> GLSL_TYPE_INT16)
>>>>> +                               output[chan] =
>>>>> LLVMBuildTrunc(ctx->ac.builder, output[chan], ctx->ac.i16, "");
>>>>> +
>>>>>
>>>>> ctx->inputs[ac_llvm_reg_index_soa(variable->data.location + i, chan)] =
>>>>> output[chan];
>>>>>                   }
>>>>>           }
>>>>> --
>>>>> 2.18.0
>>>>>
>>>>> _______________________________________________
>>>>> mesa-dev mailing list
>>>>> mesa-dev at lists.freedesktop.org
>>>>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev