Add ck tile examples to package (#1880)

* add ck tile examples to package

* Update jenkinsfile

* fix for jenkinsfile

* fix for building ck tile code on non gfx9

* compile ck tile examples only for gfx94

* include ck tile examples in all target

* fix for basic gemm UseStructuredSparsity

* Update CMakeLists.txt

* Update gemm_pipeline_problem.hpp

* add targets to rocm install

---------

Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>
This commit is contained in:
jakpiase
2025-04-28 18:53:19 +02:00
committed by GitHub
parent 83394e40d2
commit 434d19f696
25 changed files with 83 additions and 48 deletions

View File

@@ -3,7 +3,8 @@ set(TILE_ADD_RMSNORM2D_RDQUANT_FWD "tile_add_rmsnorm2d_rdquant_fwd")
# to be included in "make all/install/check"
message("adding ${TILE_ADD_RMSNORM2D_RDQUANT_FWD}")
file(GLOB INSTANCE_SRCS instances/*.cpp)
add_executable(${TILE_ADD_RMSNORM2D_RDQUANT_FWD} EXCLUDE_FROM_ALL add_rmsnorm2d_rdquant_fwd.cpp)
add_executable(${TILE_ADD_RMSNORM2D_RDQUANT_FWD} add_rmsnorm2d_rdquant_fwd.cpp)
rocm_install(TARGETS ${TILE_ADD_RMSNORM2D_RDQUANT_FWD} COMPONENT examples)
target_include_directories(${TILE_ADD_RMSNORM2D_RDQUANT_FWD} PRIVATE ${CMAKE_CURRENT_LIST_DIR})
target_sources(${TILE_ADD_RMSNORM2D_RDQUANT_FWD} PRIVATE ${INSTANCE_SRCS})
@@ -15,7 +16,8 @@ list(APPEND TILE_ADD_RMSNORM2D_RDQUANT_FWD_COMPILE_OPTIONS -Wno-undefined-func-t
target_compile_options(${TILE_ADD_RMSNORM2D_RDQUANT_FWD} PRIVATE ${TILE_ADD_RMSNORM2D_RDQUANT_FWD_COMPILE_OPTIONS})
set(EXAMPLE_ADD_RMSNORM2D_RDQUANT_FWD "tile_example_add_rmsnorm2d_rdquant_fwd")
add_executable(${EXAMPLE_ADD_RMSNORM2D_RDQUANT_FWD} EXCLUDE_FROM_ALL example_add_rmsnorm2d_rdquant_fwd.cpp)
add_executable(${EXAMPLE_ADD_RMSNORM2D_RDQUANT_FWD} example_add_rmsnorm2d_rdquant_fwd.cpp)
rocm_install(TARGETS ${EXAMPLE_ADD_RMSNORM2D_RDQUANT_FWD} COMPONENT examples)
target_compile_options(${EXAMPLE_ADD_RMSNORM2D_RDQUANT_FWD} PRIVATE ${TILE_ADD_RMSNORM2D_RDQUANT_FWD_COMPILE_OPTIONS})
# TODO: we have to turn off this global prop, otherwise the progress bar generated

View File

@@ -67,13 +67,14 @@ bool run(const ck_tile::ArgParser& arg_parser)
using TypeConfig = AddRmsnormRdquantTypeConfig<InputDataType, QuantizedDataType>;
using ADataType = typename TypeConfig::ADataType;
using BDataType = typename TypeConfig::BDataType;
using GammaDataType = typename TypeConfig::GammaDataType;
using XDataType = typename TypeConfig::XDataType;
using YScaleDataType = typename TypeConfig::YScaleDataType;
using QYDataType = typename TypeConfig::QYDataType;
using ComputeDataType = float;
using ADataType = typename TypeConfig::ADataType;
using BDataType = typename TypeConfig::BDataType;
using GammaDataType = typename TypeConfig::GammaDataType;
using XDataType = typename TypeConfig::XDataType;
using UnquantYDataType = ck_tile::null_type;
using YScaleDataType = typename TypeConfig::YScaleDataType;
using QYDataType = typename TypeConfig::QYDataType;
using ComputeDataType = float;
// host verify
ck_tile::HostTensor<ADataType> a_host({m, n}, {stride, 1});
@@ -88,6 +89,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
ck_tile::HostTensor<QYDataType> qy_host_ref({m, n}, {stride, 1});
ck_tile::HostTensor<QYDataType> qy_host_dev({m, n}, {stride, 1});
ck_tile::HostTensor<UnquantYDataType> unquant_y_host_ref({m, n}, {stride, 1});
ck_tile::FillUniformDistribution<ADataType>{-.5f, .5f}(a_host);
ck_tile::FillUniformDistribution<BDataType>{-.5f, .5f}(b_host);
@@ -191,8 +193,9 @@ bool run(const ck_tile::ArgParser& arg_parser)
GammaDataType,
ComputeDataType,
YDataType,
InvRmsDataType>(
x_host_ref, gamma_host, y_host, invRms_host_ref, epsilon);
InvRmsDataType,
UnquantYDataType>(
x_host_ref, gamma_host, y_host, invRms_host_ref, unquant_y_host_ref, epsilon);
}
// yscale

View File

@@ -62,13 +62,14 @@ bool run(const ck_tile::ArgParser& arg_parser)
assert(stride >= n);
using ADataType = DataType;
using BDataType = DataType;
using GammaDataType = DataType;
using XDataType = DataType;
using YScaleDataType = float;
using QYDataType = ck_tile::int8_t;
using ComputeDataType = float;
using ADataType = DataType;
using BDataType = DataType;
using GammaDataType = DataType;
using XDataType = DataType;
using UnquantYDataType = ck_tile::null_type;
using YScaleDataType = float;
using QYDataType = ck_tile::int8_t;
using ComputeDataType = float;
// host verify
ck_tile::HostTensor<ADataType> a_host({m, n}, {stride, 1});
@@ -81,6 +82,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
ck_tile::HostTensor<YScaleDataType> yscale_host_dev({m}, {1});
ck_tile::HostTensor<QYDataType> qy_host_ref({m, n}, {stride, 1});
ck_tile::HostTensor<QYDataType> qy_host_dev({m, n}, {stride, 1});
ck_tile::HostTensor<UnquantYDataType> unquant_y_host_ref({m, n}, {stride, 1});
ck_tile::FillUniformDistribution<ADataType>{-.5f, .5f}(a_host);
ck_tile::FillUniformDistribution<BDataType>{-.5f, .5f}(b_host);
@@ -193,8 +195,9 @@ bool run(const ck_tile::ArgParser& arg_parser)
GammaDataType,
ComputeDataType,
YDataType,
InvRmsDataType>(
x_host_ref, gamma_host, y_host, invRms_host_ref, epsilon);
InvRmsDataType,
UnquantYDataType>(
x_host_ref, gamma_host, y_host, invRms_host_ref, unquant_y_host_ref, epsilon);
}
// yscale