Skip to content

Linux: gate cblas dgemv on USE_OPENBLAS, autodetect via pkg-config #4

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
Oskar Austegard (oaustegard) wants to merge 1 commit into
base: main
Choose a base branch
from
+13 −1
Open

Linux: gate cblas dgemv on USE_OPENBLAS, autodetect via pkg-config #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 3 additions & 1 deletion transformer_vm/model/transformer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,8 @@
#ifdef __APPLE__
#define ACCELERATE_NEW_LAPACK
#include <Accelerate/Accelerate.h>
#elif defined(USE_OPENBLAS)
#include <cblas.h>
#endif

#include "hull2d_cht.h"
Expand Down Expand Up @@ -158,7 +160,7 @@ static inline void add_position_encoding(double* x, int pos) {
static inline void matvec(const double* __restrict__ W,
const double* __restrict__ x,
double* __restrict__ y, int rows, int cols) {
#if defined(__APPLE__) && !defined(NO_BLAS)
#if (defined(__APPLE__) || defined(USE_OPENBLAS)) && !defined(NO_BLAS)
cblas_dgemv(CblasRowMajor, CblasNoTrans, rows, cols,
1.0, W, cols, x, 1, 0.0, y, 1);
#else
Expand Down
10 changes: 10 additions & 0 deletions transformer_vm/runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,6 +113,16 @@ def _build_cpp_engine():
]
else:
cmd = ["g++", "-std=c++17", "-O3", "-I", attn_dir, source, "-o", binary]
# Opt into OpenBLAS for the matvec path if cblas.h is discoverable.
# Falls back silently to the scalar nested loop when not installed.
try:
pkg = subprocess.run(
["pkg-config", "--cflags", "--libs", "openblas"],
capture_output=True, text=True, check=True,
)
cmd += ["-DUSE_OPENBLAS"] + pkg.stdout.split()
except (subprocess.CalledProcessError, FileNotFoundError):
pass
try:
subprocess.check_call(cmd)
logger.info("[engine] Built: %s", binary)
Expand Down