Skip to content

fp16 for intelblas implementation #20

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

Merged
ptillet merged 3 commits into from
Apr 5, 2017
Merged

fp16 for intelblas implementation #20

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
735dc3d
intelblas_gemm enabled
Feb 27, 2017
adcff23
Add image block fp16 version.
wujunkai166 Feb 22, 2017
efdf9e1
Added new tuned skl json
Mar 15, 2017
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
1 change: 0 additions & 1 deletion include/isaac/jit/generation/gemm.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,7 +79,6 @@ class intelblas_gemm_image : public external_base
bool init_;
};


class gemm : public parameterized_base
{
private:
Expand Down
735 changes: 440 additions & 295 deletions lib/jit/generation/gemm.cpp
View file
Open in desktop

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion lib/runtime/database/intel/json/skylake.json
View file
Open in desktop

Large diffs are not rendered by default.

163,933 changes: 75,004 additions & 88,929 deletions lib/runtime/database/intel/skylake.hpp
View file
Open in desktop

Large diffs are not rendered by default.

8 changes: 4 additions & 4 deletions tune/android/tune/tools.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -131,22 +131,22 @@ def external_profiles(template):
res = []
if template is sc.templates.gemm_nn:
res += [sc.templates.cublas_gemm('N','N')]
if 'float32' in dtype.__name__:
if 'float64' not in dtype.__name__:
res += [sc.templates.intelblas_gemm('N','N')]
res += [sc.templates.intelblas_gemm_image('N','N')]
elif template is sc.templates.gemm_tn:
res += [sc.templates.cublas_gemm('T','N')]
if 'float32' in dtype.__name__:
if 'float64' not in dtype.__name__:
res += [sc.templates.intelblas_gemm('T','N')]
res += [sc.templates.intelblas_gemm_image('T','N')]
elif template is sc.templates.gemm_nt:
res += [sc.templates.cublas_gemm('N','T')]
if 'float32' in dtype.__name__:
if 'float64' not in dtype.__name__:
res += [sc.templates.intelblas_gemm('N','T')]
res += [sc.templates.intelblas_gemm_image('N','T')]
elif template is sc.templates.gemm_tt:
res += [sc.templates.cublas_gemm('T','T')]
if 'float32' in dtype.__name__:
if 'float64' not in dtype.__name__:
res += [sc.templates.intelblas_gemm('T','T')]
res += [sc.templates.intelblas_gemm_image('T','T')]
return res
Expand Down
12 changes: 7 additions & 5 deletions tune/android/tune/tune.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,11 +184,13 @@ def save():
save()
#Adding external profiles
for prof in tools.external_profiles(operation):
profiles.append(prof.__class__.__name__)
for x, y in zip(X, Y):
tree, operands = tools.tree_of(operation, x, context)
perf = performance(x,tools.benchmark(prof, tree, operation))
y.append(perf)
profiles.append(prof.__class__.__name__)
for x, y in zip(X, Y):
tree, operands = tools.tree_of(operation, x, context)
perf = performance(x,tools.benchmark(prof, tree, operation))
if max(y) < perf:
print x, '\t', prof.__class__.__name__, '\toutperform: \t', int(perf), tools.metric_name_of(operation)
y.append(perf)
#Pruning of useless profiles
X = np.array(X)
Y = np.array(Y)
Expand Down