replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[GPGPU] Do not create copy statements when targetting managed memory 

Summary:
They are not used and consequently do not even need to be computed. This reduces
the overall compile time for our kernel from 1m33s to 17s.

Reviewers: Meinersbur, bollu, singam-sanjay

Reviewed By: bollu

Subscribers: nemanjai, pollydev, llvm-commits, kbarton

Tags: #polly

Differential Revision: https://reviews.llvm.org/D36868

llvm-svn: 311157
This commit is contained in:
Tobias Grosser 2017-08-18 13:11:05 +00:00
parent 7e680f15a5
commit 861a387fac
3 changed files with 13 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
polly/lib/CodeGen/PPCGCodeGeneration.cpp
View File

@ -3140,7 +3140,8 @@ public:
DEBUG(dbgs() << getUniqueScopName(S)
<< " does not have permutable bands. Bailing out\n";);
} else {
Schedule = map_to_device(PPCGGen, Schedule);
const bool CreateTransferToFromDevice = !PollyManagedMemory;
Schedule = map_to_device(PPCGGen, Schedule, CreateTransferToFromDevice);
PPCGGen->tree = generate_code(PPCGGen, isl_schedule_copy(Schedule));
}

10
polly/lib/External/ppcg/gpu.c vendored
View File

@ -5277,7 +5277,7 @@ static __isl_give isl_schedule_node *add_init_clear_device(
* around the entire schedule.
*/
__isl_give isl_schedule *map_to_device(struct gpu_gen *gen,
__isl_take isl_schedule *schedule)
__isl_take isl_schedule *schedule, int to_from_device)
{
isl_schedule_node *node;
isl_set *context;
@ -5309,7 +5309,12 @@ __isl_give isl_schedule *map_to_device(struct gpu_gen *gen,
prefix = isl_union_map_preimage_domain_union_pw_multi_aff(prefix,
contraction);
node = mark_kernels(gen, node);
if (to_from_device) {
node = add_to_from_device(node, domain, prefix, gen->prog);
} else {
isl_union_set_free(domain);
isl_union_map_free(prefix);
}
node = isl_schedule_node_root(node);
node = isl_schedule_node_child(node, 0);
node = isl_schedule_node_child(node, 0);
@ -5675,7 +5680,8 @@ static __isl_give isl_printer *generate(__isl_take isl_printer *p,
p = print_cpu(p, scop, options);
isl_schedule_free(schedule);
} else {
schedule = map_to_device(gen, schedule);
const int create_to_from_device = 1;
schedule = map_to_device(gen, schedule, create_to_from_device);
gen->tree = generate_code(gen, schedule);
p = ppcg_set_macro_names(p);
p = ppcg_print_exposed_declarations(p, prog->scop);

3
polly/lib/External/ppcg/gpu.h vendored
View File

@ -447,7 +447,8 @@ __isl_give isl_schedule_node *gpu_create_kernel(struct gpu_gen *gen,
__isl_give isl_schedule *get_schedule(struct gpu_gen *gen);
int has_any_permutable_node(__isl_keep isl_schedule *schedule);
__isl_give isl_schedule *map_to_device(struct gpu_gen *gen,
__isl_take isl_schedule *schedule);
__isl_take isl_schedule *schedule,
int to_from_device);
__isl_give isl_ast_node *generate_code(struct gpu_gen *gen,
__isl_take isl_schedule *schedule);