[Mesa-dev] [PATCH 2/3] clover: Enable cl_khr_fp64 for devices that support doubles v2

[Matt Arsenault]

> On Feb 26, 2015, at 5:06 PM, Tom Stellard <thomas.stellard at amd.com> wrote:
> 
> v2:
>  - Report correct values for CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE and
>    CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE.
>  - Only define cl_khr_fp64 if the extension is supported.
>  - Remove trailing space from extension string.
>  - Rename device query function from cl_khr_fp64() to has_doubles().
> ---
> src/gallium/state_trackers/clover/api/device.cpp  | 6 +++---
> src/gallium/state_trackers/clover/core/device.cpp | 6 ++++++
> src/gallium/state_trackers/clover/core/device.hpp | 1 +
> 3 files changed, 10 insertions(+), 3 deletions(-)

There are two more changes that I don’t see here that should accompany this one. The correct CL_DEVICE_DOUBLE_FP_CONFIG values should be reported (CL_DEVICE_SINGLE_FP_CONFIG is also currently wrong as well). 
Plus, the FP config register needs to be set from the program binary so that double denormals are not flushed.

> 
> diff --git a/src/gallium/state_trackers/clover/api/device.cpp b/src/gallium/state_trackers/clover/api/device.cpp
> index e825468..217d2c3 100644
> --- a/src/gallium/state_trackers/clover/api/device.cpp
> +++ b/src/gallium/state_trackers/clover/api/device.cpp
> @@ -145,7 +145,7 @@ clGetDeviceInfo(cl_device_id d_dev, cl_device_info param,
>       break;
> 
>    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE:
> -      buf.as_scalar<cl_uint>() = 2;
> +      buf.as_scalar<cl_uint>() = dev.has_doubles() ? 2 : 0;
>       break;
> 
>    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF:
> @@ -283,7 +283,7 @@ clGetDeviceInfo(cl_device_id d_dev, cl_device_info param,
>       break;
> 
>    case CL_DEVICE_EXTENSIONS:
> -      buf.as_string() = "";
> +      buf.as_string() = dev.has_doubles() ? "cl_khr_fp64" : "";
>       break;
> 
>    case CL_DEVICE_PLATFORM:
> @@ -315,7 +315,7 @@ clGetDeviceInfo(cl_device_id d_dev, cl_device_info param,
>       break;
> 
>    case CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE:
> -      buf.as_scalar<cl_uint>() = 2;
> +      buf.as_scalar<cl_uint>() = dev.has_doubles() ? 2 : 0;
>       break;
> 
>    case CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF:
> diff --git a/src/gallium/state_trackers/clover/core/device.cpp b/src/gallium/state_trackers/clover/core/device.cpp
> index 688a7dd..c3f3b4e 100644
> --- a/src/gallium/state_trackers/clover/core/device.cpp
> +++ b/src/gallium/state_trackers/clover/core/device.cpp
> @@ -173,6 +173,12 @@ device::image_support() const {
>                                       PIPE_COMPUTE_CAP_IMAGES_SUPPORTED)[0];
> }
> 
> +bool
> +device::has_doubles() const {
> +   return pipe->get_shader_param(pipe, PIPE_SHADER_COMPUTE,
> +                                 PIPE_SHADER_CAP_DOUBLES);
> +}
> +
> std::vector<size_t>
> device::max_block_size() const {
>    auto v = get_compute_param<uint64_t>(pipe, PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE);
> diff --git a/src/gallium/state_trackers/clover/core/device.hpp b/src/gallium/state_trackers/clover/core/device.hpp
> index 2201700..de5fc6b 100644
> --- a/src/gallium/state_trackers/clover/core/device.hpp
> +++ b/src/gallium/state_trackers/clover/core/device.hpp
> @@ -64,6 +64,7 @@ namespace clover {
>       cl_uint max_clock_frequency() const;
>       cl_uint max_compute_units() const;
>       bool image_support() const;
> +      bool has_doubles() const;
> 
>       std::vector<size_t> max_block_size() const;
>       std::string device_name() const;
> -- 
> 2.0.4
> 
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev