Skip to content

UBSAN: Avoid bitshifts in UNROLL #6

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mgondan wants to merge 2 commits into
base: master
Choose a base branch
from
Open

UBSAN: Avoid bitshifts in UNROLL #6

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5553c8a
UBSAN: Avoid bitshifts in UNROLL
mgondan Dec 12, 2025
bf93232
Update README.md
mgondan Dec 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,6 +77,11 @@ changes to
With `-ftree-vectorize`, it seems that this trick does wonders and speeds up
the code even more.

_Update:_ gcc-UBSAN complains about undefined behavior when the above bitshifts
are used, so the line has now been changed to:

uint32_t foo = y & 1 ? 0x12345678 : 0

Finally, note that people have done SIMD and CUDA implementations. If
you are looking for even more speed, I suggest you check them out.

Expand Down
5 changes: 2 additions & 3 deletions mersenne-twister.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ static MTState state;

#define UNROLL(expr) \
y = M32(state.MT[i]) | L31(state.MT[i+1]); \
state.MT[i] = state.MT[expr] ^ (y >> 1) ^ (((int32_t(y) << 31) >> 31) & MAGIC); \
state.MT[i] = state.MT[expr] ^ (y >> 1) ^ (y & 1 ? MAGIC : 0); \
++i;

static void generate_numbers()
Expand Down Expand Up @@ -99,8 +99,7 @@ static void generate_numbers()
{
// i = 623, last step rolls over
y = M32(state.MT[SIZE-1]) | L31(state.MT[0]);
state.MT[SIZE-1] = state.MT[PERIOD-1] ^ (y >> 1) ^ (((int32_t(y) << 31) >>
31) & MAGIC);
state.MT[SIZE-1] = state.MT[PERIOD-1] ^ (y >> 1) ^ (y & 1 ? MAGIC : 0);
}

// Temper all numbers in a batch
Expand Down