24069718049

Committed 07 Apr 2026 07:27AM UTC coverage: 64.886%. First build

Build # 24069718049

Build Type

Pull #632

github

Committed by

web-flow

Commit Message

Merge 80d1d9fb5 into be4a18f39

Pull Request Pull Request #632: Separate can_be_applied and apply for GPUTilling during Loop Scheduling

Coverage Stats

272 of 338 new or added lines in 15 files covered. (80.47%)

28868 of 44490 relevant lines covered (64.89%)

606.24 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

0.0

/opt/include/sdfg/passes/offloading/gpu_nested_parallelization_pass.h

#pragma once

#include <vector>
#include "sdfg/passes/pass.h"
#include "sdfg/structured_control_flow/map.h"

namespace sdfg {
namespace passes {

enum class GPUTarget {
    CUDA,
    ROCM,
};

/**
 * @brief Phased GPU nested map parallelization pass.
 *
 * Given a set of outer maps, finds all nested maps within them and applies
 * nested parallelization (CUDA or ROCM) in two phases:
 * 1. can_be_applied phase: collects all nested maps where parallelization is applicable
 * 2. apply phase: applies parallelization to all collected maps
 */
class GPUNestedParallelizationPass : public Pass {
private:
    const std::vector<structured_control_flow::Map*>& maps_;
    GPUTarget target_;
    size_t block_size_;

public:
    GPUNestedParallelizationPass(
        const std::vector<structured_control_flow::Map*>& maps, GPUTarget target, size_t block_size
    );
    ~GPUNestedParallelizationPass() override = default;

    bool run_pass(builder::StructuredSDFGBuilder& builder, analysis::AnalysisManager& analysis_manager) override;

    std::string name() override { return "GPUNestedParallelizationPass"; }
};

} // namespace passes
} // namespace sdfg

1	#pragma once
2
3	#include <vector>
4	#include "sdfg/passes/pass.h"
5	#include "sdfg/structured_control_flow/map.h"
6
7	namespace sdfg {
8	namespace passes {
9
10	enum class GPUTarget {
11	CUDA,
12	ROCM,
13	};
14
15	/**
16	* @brief Phased GPU nested map parallelization pass.
17	*
18	* Given a set of outer maps, finds all nested maps within them and applies
19	* nested parallelization (CUDA or ROCM) in two phases:
20	* 1. can_be_applied phase: collects all nested maps where parallelization is applicable
21	* 2. apply phase: applies parallelization to all collected maps
22	*/
23	class GPUNestedParallelizationPass : public Pass {
24	private:
25	const std::vector<structured_control_flow::Map*>& maps_;
26	GPUTarget target_;
27	size_t block_size_;
28
29	public:
30	GPUNestedParallelizationPass(
31	const std::vector<structured_control_flow::Map*>& maps, GPUTarget target, size_t block_size
32	);
NEW 33	~GPUNestedParallelizationPass() override = default;	×
34
35	bool run_pass(builder::StructuredSDFGBuilder& builder, analysis::AnalysisManager& analysis_manager) override;
36
NEW 37	std::string name() override { return "GPUNestedParallelizationPass"; }	×
38	};
39
40	} // namespace passes
41	} // namespace sdfg

daisytuner / docc / 24069718049

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous