This is an archive of the discontinued LLVM Phabricator instance.

Differential D157348

Fix UB in benchmakrs
ClosedPublic

Authored by vitalybuka on Aug 7 2023, 4:26 PM.

Details

Reviewers
homerdin
rengolin
MaskRay
Commits
rTaf2e07172ef6: Fix UB in benchmakrs
Summary

This is not ideal fixes, but it does not change the
point of the benchmark, so should be good enough here.

  • ix[k] can be 0 causing negative index on array access.
  • p[ip][0] can be outside of Index_type.

Diff Detail

Repository
rT test-suite

Event Timeline

vitalybuka created this revision.Aug 7 2023, 4:26 PM
Herald added a project: Restricted Project. · View Herald TranscriptAug 7 2023, 4:26 PM
vitalybuka requested review of this revision.Aug 7 2023, 4:26 PM
Harbormaster completed remote builds in B250948: Diff 547992.Aug 7 2023, 4:26 PM
vitalybuka edited the summary of this revision. (Show Details)Aug 7 2023, 4:28 PM
vitalybuka added reviewers: homerdin, rengolin, MaskRay.
MaskRay accepted this revision.Aug 7 2023, 6:05 PM
MaskRay added inline comments.
MicroBenchmarks/LCALS/SubsetCRawLoops/RawSubsetCbenchmarks.cxx
552–553

this may express the intention better

This revision is now accepted and ready to land.Aug 7 2023, 6:05 PM
vitalybuka updated this revision to Diff 548019.Aug 7 2023, 6:37 PM

(Index_type)

Harbormaster completed remote builds in B250963: Diff 548019.Aug 7 2023, 6:37 PM
vitalybuka marked an inline comment as done.Aug 7 2023, 6:37 PM
MaskRay accepted this revision.Aug 7 2023, 6:42 PM
Closed by commit rTaf2e07172ef6: Fix UB in benchmakrs (authored by vitalybuka). · Explain WhyAug 7 2023, 6:43 PM
This revision was automatically updated to reflect the committed changes.
vitalybuka added a commit: rTaf2e07172ef6: Fix UB in benchmakrs.

Revision Contents

PathSize
MicroBenchmarks/
LCALS/
SubsetCLambdaLoops/
12 lines
SubsetCRawLoops/
12 lines

Diff 548023

MicroBenchmarks/LCALS/SubsetCLambdaLoops/LambdaSubsetCbenchmarks.cxx

Show First 20 Lines • Show All 373 Lines • ▼ Show 20 Linesstatic void BM_PIC_2D_LAMBDA(benchmark::State& state) {
Real_ptr* h = loop_data.array_2D_64x64_Real[0]; Real_ptr* h = loop_data.array_2D_64x64_Real[0];
for (auto _ : state) { for (auto _ : state) {
forall<exec_policy>(0, state.range(0), forall<exec_policy>(0, state.range(0),
[&] (Index_type ip) { [&] (Index_type ip) {
Index_type i1, j1, i2, j2; Index_type i1, j1, i2, j2;
i1 = (Index_type) p[ip][0]; i1 = (Index_type) (long long) p[ip][0];
j1 = (Index_type) p[ip][1]; j1 = (Index_type) (long long) p[ip][1];
i1 &= 64-1; i1 &= 64-1;
j1 &= 64-1; j1 &= 64-1;
p[ip][2] += b[j1][i1]; p[ip][2] += b[j1][i1];
p[ip][3] += c[j1][i1]; p[ip][3] += c[j1][i1];
p[ip][0] += p[ip][2]; p[ip][0] += p[ip][2];
p[ip][1] += p[ip][3]; p[ip][1] += p[ip][3];
i2 = (Index_type) p[ip][0]; i2 = (Index_type) (long long) p[ip][0];
j2 = (Index_type) p[ip][1]; j2 = (Index_type) (long long) p[ip][1];
i2 = ( i2 & 64-1 ) ; i2 = ( i2 & 64-1 ) ;
j2 = ( j2 & 64-1 ) ; j2 = ( j2 & 64-1 ) ;
p[ip][0] += y[i2+32]; p[ip][0] += y[i2+32];
p[ip][1] += z[j2+32]; p[ip][1] += z[j2+32];
i2 += e[i2+32]; i2 += e[i2+32];
j2 += f[j2+32]; j2 += f[j2+32];
h[j2][i2] += 1.0; h[j2][i2] += 1.0;
} ); } );
Show All 30 Linesstatic void BM_PIC_1D_LAMBDA(benchmark::State& state) {
for (auto _ : state) { for (auto _ : state) {
forall<exec_policy>(0, state.range(0), forall<exec_policy>(0, state.range(0),
[&] (Index_type k) { [&] (Index_type k) {
vx[k] = 0.0; vx[k] = 0.0;
xx[k] = 0.0; xx[k] = 0.0;
ix[k] = (Index_type) grd[k]; ix[k] = (Index_type) grd[k];
xi[k] = (Real_type) ix[k]; xi[k] = (Real_type) ix[k];
ex1[k] = ex[ ix[k] - 1 ]; ex1[k] = ix[k] ? ex[ ix[k] - 1 ] : 0;
dex1[k] = dex[ ix[k] - 1 ]; dex1[k] = ix[k] ? dex[ ix[k] - 1 ] : 0;
} ); } );
forall<exec_policy>(0, state.range(0), forall<exec_policy>(0, state.range(0),
[&] (Index_type k) { [&] (Index_type k) {
vx[k] = vx[k] + ex1[k] + ( xx[k] - xi[k] )*dex1[k]; vx[k] = vx[k] + ex1[k] + ( xx[k] - xi[k] )*dex1[k];
xx[k] = xx[k] + vx[k] + flx; xx[k] = xx[k] + vx[k] + flx;
ir[k] = (Index_type) xx[k]; ir[k] = (Index_type) xx[k];
rx[k] = xx[k] - ir[k]; rx[k] = xx[k] - ir[k];
▲ Show 20 Lines • Show All 271 Lines • Show Last 20 Lines

MicroBenchmarks/LCALS/SubsetCRawLoops/RawSubsetCbenchmarks.cxx

Show First 20 Lines • Show All 543 Lines • ▼ Show 20 Linesstatic void BM_PIC_2D_RAW(benchmark::State& state) {
Index_type* f = loop_data.array_1D_Indx[1]; Index_type* f = loop_data.array_1D_Indx[1];
Real_ptr* h = loop_data.array_2D_64x64_Real[0]; Real_ptr* h = loop_data.array_2D_64x64_Real[0];
for (auto _ : state) { for (auto _ : state) {
for (Index_type ip=0 ; ip< state.range(0) ; ip++ ) { for (Index_type ip=0 ; ip< state.range(0) ; ip++ ) {
Index_type i1, j1, i2, j2; Index_type i1, j1, i2, j2;
i1 = (Index_type) p[ip][0]; i1 = (Index_type) (long long) p[ip][0];
j1 = (Index_type) p[ip][1]; j1 = (Index_type) (long long) p[ip][1];
MaskRayUnsubmitted
Done
ReplyInline Actions
Index_type i1, j1, i2, j2;
- i1 = (long long) p[ip][0];
+ i1 = (Index_type)(long long) p[ip][0];
j1 = (long long) p[ip][1];

this may express the intention better

MaskRay: this may express the intention better
i1 &= 64-1; i1 &= 64-1;
j1 &= 64-1; j1 &= 64-1;
p[ip][2] += b[j1][i1]; p[ip][2] += b[j1][i1];
p[ip][3] += c[j1][i1]; p[ip][3] += c[j1][i1];
p[ip][0] += p[ip][2]; p[ip][0] += p[ip][2];
p[ip][1] += p[ip][3]; p[ip][1] += p[ip][3];
i2 = (Index_type) p[ip][0]; i2 = (Index_type) (long long) p[ip][0];
j2 = (Index_type) p[ip][1]; j2 = (Index_type) (long long) p[ip][1];
i2 = ( i2 & 64-1 ) ; i2 = ( i2 & 64-1 ) ;
j2 = ( j2 & 64-1 ) ; j2 = ( j2 & 64-1 ) ;
p[ip][0] += y[i2+32]; p[ip][0] += y[i2+32];
p[ip][1] += z[j2+32]; p[ip][1] += z[j2+32];
i2 += e[i2+32]; i2 += e[i2+32];
j2 += f[j2+32]; j2 += f[j2+32];
h[j2][i2] += 1.0; h[j2][i2] += 1.0;
} }
▲ Show 20 Lines • Show All 56 Lines • ▼ Show 20 Linesstatic void BM_PIC_1D_RAW(benchmark::State& state) {
for (auto _ : state) { for (auto _ : state) {
for (Index_type k=0 ; k< state.range(0) ; k++ ) { for (Index_type k=0 ; k< state.range(0) ; k++ ) {
vx[k] = 0.0; vx[k] = 0.0;
xx[k] = 0.0; xx[k] = 0.0;
ix[k] = (Index_type) grd[k]; ix[k] = (Index_type) grd[k];
xi[k] = (Real_type) ix[k]; xi[k] = (Real_type) ix[k];
ex1[k] = ex[ ix[k] - 1 ]; ex1[k] = ix[k] ? ex[ ix[k] - 1 ] : 0;
dex1[k] = dex[ ix[k] - 1 ]; dex1[k] = ix[k] ? dex[ ix[k] - 1 ] : 0;
} }
for (Index_type k=0 ; k< state.range(0) ; k++ ) { for (Index_type k=0 ; k< state.range(0) ; k++ ) {
vx[k] = vx[k] + ex1[k] + ( xx[k] - xi[k] )*dex1[k]; vx[k] = vx[k] + ex1[k] + ( xx[k] - xi[k] )*dex1[k];
xx[k] = xx[k] + vx[k] + flx; xx[k] = xx[k] + vx[k] + flx;
ir[k] = (Index_type) xx[k]; ir[k] = (Index_type) xx[k];
rx[k] = xx[k] - ir[k]; rx[k] = xx[k] - ir[k];
ir[k] = ( ir[k] & (2048-1) ) + 1; ir[k] = ( ir[k] & (2048-1) ) + 1;
▲ Show 20 Lines • Show All 374 Lines • Show Last 20 Lines