This is an archive of the discontinued LLVM Phabricator instance.

Differential D110430

[OpenMP] Add new worksharing definitions into device RTL
ClosedPublic

Authored by jhuber6 on Sep 24 2021, 10:14 AM.

Details

Reviewers
jdoerfert
tianshilei1992
Commits
rG74d622dea450: [OpenMP] Add new worksharing definitions into device RTL
Summary

This path defines the newly added __kmpc_disitrute_static_init
functions in the device runtime library. These functions are currently
exact copies of the current worksharing method but can be tuned later.

Depends on D110429

Diff Detail

Event Timeline

jhuber6 created this revision.Sep 24 2021, 10:14 AM
Herald added subscribers: guansong, yaxunl. · View Herald TranscriptSep 24 2021, 10:14 AM
jhuber6 requested review of this revision.Sep 24 2021, 10:14 AM
Herald added a project: Restricted Project. · View Herald TranscriptSep 24 2021, 10:14 AM
Herald added subscribers: openmp-commits, sstefan1. · View Herald Transcript
tianshilei1992 accepted this revision.Sep 24 2021, 10:25 AM

LGTM. The dependent patch has a huge amount of changes. Hopefully the function signatures here are finalized. Otherwise...

This revision is now accepted and ready to land.Sep 24 2021, 10:25 AM
Harbormaster completed remote builds in B125603: Diff 374896.Sep 24 2021, 10:43 AM
jhuber6 added a comment.Sep 24 2021, 10:54 AM
In D110430#3021214, @tianshilei1992 wrote:

LGTM. The dependent patch has a huge amount of changes. Hopefully the function signatures here are finalized. Otherwise...

It's pretty much all fixing the tests, it's also to introduce stuff for the device library since we don't need to worry about backwards compatibility.

This revision was landed with ongoing or failed builds.Sep 27 2021, 8:37 AM
Closed by commit rG74d622dea450: [OpenMP] Add new worksharing definitions into device RTL (authored by jhuber6). · Explain Why
This revision was automatically updated to reflect the committed changes.
jhuber6 added a commit: rG74d622dea450: [OpenMP] Add new worksharing definitions into device RTL.
jhuber6 mentioned this in D110429: [OpenMP] Introduce a new worksharing RTL function for distribute.Sep 27 2021, 9:05 AM

Revision Contents

PathSize
openmp/
libomptarget/
DeviceRTL/
src/
42 lines
deviceRTLs/
common/
src/
50 lines
23 lines

Diff 375288

openmp/libomptarget/DeviceRTL/src/Workshare.cpp

Show First 20 Lines • Show All 586 Lines • ▼ Show 20 Linesvoid __kmpc_for_static_init_8u(IdentTy *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter, int32_t schedtype, int32_t *plastiter,
uint64_t *plower, uint64_t *pupper, uint64_t *plower, uint64_t *pupper,
int64_t *pstride, int64_t incr, int64_t chunk) { int64_t *pstride, int64_t incr, int64_t chunk) {
omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init( omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk, global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
mapping::isSPMDMode()); mapping::isSPMDMode());
} }
void __kmpc_distribute_static_init_4(IdentTy *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter,
int32_t *plower, int32_t *pupper,
int32_t *pstride, int32_t incr,
int32_t chunk) {
omptarget_nvptx_LoopSupport<int32_t, int32_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
mapping::isSPMDMode());
}
void __kmpc_distribute_static_init_4u(IdentTy *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter,
uint32_t *plower, uint32_t *pupper,
int32_t *pstride, int32_t incr,
int32_t chunk) {
omptarget_nvptx_LoopSupport<uint32_t, int32_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
mapping::isSPMDMode());
}
void __kmpc_distribute_static_init_8(IdentTy *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter,
int64_t *plower, int64_t *pupper,
int64_t *pstride, int64_t incr,
int64_t chunk) {
omptarget_nvptx_LoopSupport<int64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
mapping::isSPMDMode());
}
void __kmpc_distribute_static_init_8u(IdentTy *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter,
uint64_t *plower, uint64_t *pupper,
int64_t *pstride, int64_t incr,
int64_t chunk) {
omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
mapping::isSPMDMode());
}
void __kmpc_for_static_fini(IdentTy *loc, int32_t global_tid) {} void __kmpc_for_static_fini(IdentTy *loc, int32_t global_tid) {}
void __kmpc_distribute_static_fini(IdentTy *loc, int32_t global_tid) {}
} }
#pragma omp end declare target #pragma omp end declare target

openmp/libomptarget/deviceRTLs/common/src/loop.cu

Show First 20 Lines • Show All 654 Lines • ▼ Show 20 LinesEXTERN void __kmpc_for_static_init_8u(kmp_Ident *loc, int32_t global_tid,
int64_t *pstride, int64_t incr, int64_t *pstride, int64_t incr,
int64_t chunk) { int64_t chunk) {
PRINT0(LD_IO, "call kmpc_for_static_init_8u\n"); PRINT0(LD_IO, "call kmpc_for_static_init_8u\n");
omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init( omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk, global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
__kmpc_is_spmd_exec_mode()); __kmpc_is_spmd_exec_mode());
} }
EXTERN void __kmpc_distribute_static_init_4(kmp_Ident *loc, int32_t global_tid,
int32_t schedtype,
int32_t *plastiter, int32_t *plower,
int32_t *pupper, int32_t *pstride,
int32_t incr, int32_t chunk) {
PRINT0(LD_IO, "call kmpc_distribute_static_init_4\n");
omptarget_nvptx_LoopSupport<int32_t, int32_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
__kmpc_is_spmd_exec_mode());
}
EXTERN void __kmpc_distribute_static_init_4u(kmp_Ident *loc, int32_t global_tid,
int32_t schedtype,
int32_t *plastiter,
uint32_t *plower, uint32_t *pupper,
int32_t *pstride, int32_t incr,
int32_t chunk) {
PRINT0(LD_IO, "call kmpc_distribute_static_init_4u\n");
omptarget_nvptx_LoopSupport<uint32_t, int32_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
__kmpc_is_spmd_exec_mode());
}
EXTERN void __kmpc_distribute_static_init_8(kmp_Ident *loc, int32_t global_tid,
int32_t schedtype,
int32_t *plastiter, int64_t *plower,
int64_t *pupper, int64_t *pstride,
int64_t incr, int64_t chunk) {
PRINT0(LD_IO, "call kmpc_distribute_static_init_8\n");
omptarget_nvptx_LoopSupport<int64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
__kmpc_is_spmd_exec_mode());
}
EXTERN void __kmpc_distribute_static_init_8u(kmp_Ident *loc, int32_t global_tid,
int32_t schedtype,
int32_t *plastiter,
uint64_t *plower, uint64_t *pupper,
int64_t *pstride, int64_t incr,
int64_t chunk) {
PRINT0(LD_IO, "call kmpc_distribute_static_init_8u\n");
omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
__kmpc_is_spmd_exec_mode());
}
EXTERN EXTERN
void __kmpc_for_static_init_4_simple_spmd(kmp_Ident *loc, int32_t global_tid, void __kmpc_for_static_init_4_simple_spmd(kmp_Ident *loc, int32_t global_tid,
int32_t schedtype, int32_t *plastiter, int32_t schedtype, int32_t *plastiter,
int32_t *plower, int32_t *pupper, int32_t *plower, int32_t *pupper,
int32_t *pstride, int32_t incr, int32_t *pstride, int32_t incr,
int32_t chunk) { int32_t chunk) {
PRINT0(LD_IO, "call kmpc_for_static_init_4_simple_spmd\n"); PRINT0(LD_IO, "call kmpc_for_static_init_4_simple_spmd\n");
omptarget_nvptx_LoopSupport<int32_t, int32_t>::for_static_init( omptarget_nvptx_LoopSupport<int32_t, int32_t>::for_static_init(
▲ Show 20 Lines • Show All 80 Lines • ▼ Show 20 Linesvoid __kmpc_for_static_init_8u_simple_generic(
uint64_t *plower, uint64_t *pupper, int64_t *pstride, int64_t incr, uint64_t *plower, uint64_t *pupper, int64_t *pstride, int64_t incr,
int64_t chunk) { int64_t chunk) {
PRINT0(LD_IO, "call kmpc_for_static_init_8u_simple_generic\n"); PRINT0(LD_IO, "call kmpc_for_static_init_8u_simple_generic\n");
omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init( omptarget_nvptx_LoopSupport<uint64_t, int64_t>::for_static_init(
global_tid, schedtype, plastiter, plower, pupper, pstride, chunk, global_tid, schedtype, plastiter, plower, pupper, pstride, chunk,
/*IsSPMDExecutionMode=*/false); /*IsSPMDExecutionMode=*/false);
} }
EXTERN void __kmpc_distribute_static_fini(kmp_Ident *loc, int32_t global_tid) {
PRINT0(LD_IO, "call kmpc_distribute_static_fini\n");
}
EXTERN void __kmpc_for_static_fini(kmp_Ident *loc, int32_t global_tid) { EXTERN void __kmpc_for_static_fini(kmp_Ident *loc, int32_t global_tid) {
PRINT0(LD_IO, "call kmpc_for_static_fini\n"); PRINT0(LD_IO, "call kmpc_for_static_fini\n");
} }
#pragma omp end declare target #pragma omp end declare target

openmp/libomptarget/deviceRTLs/interface.h

Show First 20 Lines • Show All 249 Lines • ▼ Show 20 LinesEXTERN void __kmpc_for_static_init_8(kmp_Ident *loc, int32_t global_tid,
int64_t *plower, int64_t *pupper, int64_t *plower, int64_t *pupper,
int64_t *pstride, int64_t incr, int64_t *pstride, int64_t incr,
int64_t chunk); int64_t chunk);
EXTERN void __kmpc_for_static_init_8u(kmp_Ident *loc, int32_t global_tid, EXTERN void __kmpc_for_static_init_8u(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter1, int32_t sched, int32_t *plastiter1,
uint64_t *plower, uint64_t *pupper, uint64_t *plower, uint64_t *pupper,
int64_t *pstride, int64_t incr, int64_t *pstride, int64_t incr,
int64_t chunk); int64_t chunk);
// distribute static (no chunk or chunk)
EXTERN void __kmpc_distribute_static_init_4(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter,
int32_t *plower, int32_t *pupper,
int32_t *pstride, int32_t incr,
int32_t chunk);
EXTERN void __kmpc_distribute_static_init_4u(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter,
uint32_t *plower, uint32_t *pupper,
int32_t *pstride, int32_t incr,
int32_t chunk);
EXTERN void __kmpc_distribute_static_init_8(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter,
int64_t *plower, int64_t *pupper,
int64_t *pstride, int64_t incr,
int64_t chunk);
EXTERN void __kmpc_distribute_static_init_8u(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter1,
uint64_t *plower, uint64_t *pupper,
int64_t *pstride, int64_t incr,
int64_t chunk);
EXTERN EXTERN
void __kmpc_for_static_init_4_simple_spmd(kmp_Ident *loc, int32_t global_tid, void __kmpc_for_static_init_4_simple_spmd(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t *plastiter, int32_t sched, int32_t *plastiter,
int32_t *plower, int32_t *pupper, int32_t *plower, int32_t *pupper,
int32_t *pstride, int32_t incr, int32_t *pstride, int32_t incr,
int32_t chunk); int32_t chunk);
EXTERN EXTERN
void __kmpc_for_static_init_4u_simple_spmd(kmp_Ident *loc, int32_t global_tid, void __kmpc_for_static_init_4u_simple_spmd(kmp_Ident *loc, int32_t global_tid,
Show All 33 Lines
EXTERN EXTERN
void __kmpc_for_static_init_8u_simple_generic( void __kmpc_for_static_init_8u_simple_generic(
kmp_Ident *loc, int32_t global_tid, int32_t sched, int32_t *plastiter1, kmp_Ident *loc, int32_t global_tid, int32_t sched, int32_t *plastiter1,
uint64_t *plower, uint64_t *pupper, int64_t *pstride, int64_t incr, uint64_t *plower, uint64_t *pupper, int64_t *pstride, int64_t incr,
int64_t chunk); int64_t chunk);
EXTERN void __kmpc_for_static_fini(kmp_Ident *loc, int32_t global_tid); EXTERN void __kmpc_for_static_fini(kmp_Ident *loc, int32_t global_tid);
EXTERN void __kmpc_distribute_static_fini(kmp_Ident *loc, int32_t global_tid);
// for dynamic // for dynamic
EXTERN void __kmpc_dispatch_init_4(kmp_Ident *loc, int32_t global_tid, EXTERN void __kmpc_dispatch_init_4(kmp_Ident *loc, int32_t global_tid,
int32_t sched, int32_t lower, int32_t upper, int32_t sched, int32_t lower, int32_t upper,
int32_t incr, int32_t chunk); int32_t incr, int32_t chunk);
EXTERN void __kmpc_dispatch_init_4u(kmp_Ident *loc, int32_t global_tid, EXTERN void __kmpc_dispatch_init_4u(kmp_Ident *loc, int32_t global_tid,
int32_t sched, uint32_t lower, int32_t sched, uint32_t lower,
uint32_t upper, int32_t incr, uint32_t upper, int32_t incr,
int32_t chunk); int32_t chunk);
▲ Show 20 Lines • Show All 165 Lines • Show Last 20 Lines