Skip to content

vulkan: workaround llvmpipe atan2 zero case #6729

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
nihui merged 1 commit into from
May 19, 2026
+76 −0
Merged

vulkan: workaround llvmpipe atan2 zero case #6729

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
14 changes: 14 additions & 0 deletions src/layer/vulkan/shader/binaryop.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,9 +97,23 @@ void main()
#if NCNN_moltenvk
if (op_type == 10) res = afp(atan(float(v1), float(v2)));
if (op_type == 11) res = afp(atan(float(v2), float(v1)));
#else
#if ncnn_driverID == 13
// llvmpipe driver (VK_DRIVER_ID_MESA_LLVMPIPE = 13) workaround for atan(0, 0), match atan2f(0, 0) == 0
if (op_type == 10)
{
afp v2_safe = (v1 == afp(0.f) && v2 == afp(0.f)) ? afp(1.f) : v2;
res = atan(v1, v2_safe);
}
if (op_type == 11)
{
afp v1_safe = (v1 == afp(0.f) && v2 == afp(0.f)) ? afp(1.f) : v1;
res = atan(v2, v1_safe);
}
#else
if (op_type == 10) res = atan(v1, v2);
if (op_type == 11) res = atan(v2, v1);
#endif
#endif
if (op_type == 12) res = v1 - trunc(v1 / v2) * v2;
if (op_type == 13) res = v2 - trunc(v2 / v1) * v1;
Expand Down
14 changes: 14 additions & 0 deletions src/layer/vulkan/shader/binaryop_broadcast.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,9 +166,23 @@ void main()
#if NCNN_moltenvk
if (op_type == 10) res = afp(atan(float(v1), float(v2)));
if (op_type == 11) res = afp(atan(float(v2), float(v1)));
#else
#if ncnn_driverID == 13
// llvmpipe driver (VK_DRIVER_ID_MESA_LLVMPIPE = 13) workaround for atan(0, 0), match atan2f(0, 0) == 0
if (op_type == 10)
{
afp v2_safe = (v1 == afp(0.f) && v2 == afp(0.f)) ? afp(1.f) : v2;
res = atan(v1, v2_safe);
}
if (op_type == 11)
{
afp v1_safe = (v1 == afp(0.f) && v2 == afp(0.f)) ? afp(1.f) : v1;
res = atan(v2, v1_safe);
}
#else
if (op_type == 10) res = atan(v1, v2);
if (op_type == 11) res = atan(v2, v1);
#endif
#endif
if (op_type == 12) res = v1 - trunc(v1 / v2) * v2;
if (op_type == 13) res = v2 - trunc(v2 / v1) * v1;
Expand Down
16 changes: 16 additions & 0 deletions src/layer/vulkan/shader/binaryop_broadcast_pack1to4.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,9 +97,25 @@ void main()
#if NCNN_moltenvk
if (op_type == 10) res = afpvec4(atan(vec4(v1), vec4(v2)));
if (op_type == 11) res = afpvec4(atan(vec4(v2), vec4(v1)));
#else
#if ncnn_driverID == 13
// llvmpipe driver (VK_DRIVER_ID_MESA_LLVMPIPE = 13) workaround for atan(0, 0), match atan2f(0, 0) == 0
if (op_type == 10)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v2_safe = mix(v2, afpvec4(1.f), zero_mask);
res = atan(v1, v2_safe);
}
if (op_type == 11)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v1_safe = mix(v1, afpvec4(1.f), zero_mask);
res = atan(v2, v1_safe);
}
#else
if (op_type == 10) res = atan(v1, v2);
if (op_type == 11) res = atan(v2, v1);
#endif
#endif
if (op_type == 12) res = v1 - trunc(v1 / v2) * v2;
if (op_type == 13) res = v2 - trunc(v2 / v1) * v1;
Expand Down
16 changes: 16 additions & 0 deletions src/layer/vulkan/shader/binaryop_broadcast_pack4.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,9 +166,25 @@ void main()
#if NCNN_moltenvk
if (op_type == 10) res = afpvec4(atan(vec4(v1), vec4(v2)));
if (op_type == 11) res = afpvec4(atan(vec4(v2), vec4(v1)));
#else
#if ncnn_driverID == 13
// llvmpipe driver (VK_DRIVER_ID_MESA_LLVMPIPE = 13) workaround for atan(0, 0), match atan2f(0, 0) == 0
if (op_type == 10)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v2_safe = mix(v2, afpvec4(1.f), zero_mask);
res = atan(v1, v2_safe);
}
if (op_type == 11)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v1_safe = mix(v1, afpvec4(1.f), zero_mask);
res = atan(v2, v1_safe);
}
#else
if (op_type == 10) res = atan(v1, v2);
if (op_type == 11) res = atan(v2, v1);
#endif
#endif
if (op_type == 12) res = v1 - trunc(v1 / v2) * v2;
if (op_type == 13) res = v2 - trunc(v2 / v1) * v1;
Expand Down
16 changes: 16 additions & 0 deletions src/layer/vulkan/shader/binaryop_pack4.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,9 +92,25 @@ void main()
#if NCNN_moltenvk
if (op_type == 10) res = afpvec4(atan(vec4(v1), vec4(v2)));
if (op_type == 11) res = afpvec4(atan(vec4(v2), vec4(v1)));
#else
#if ncnn_driverID == 13
// llvmpipe driver (VK_DRIVER_ID_MESA_LLVMPIPE = 13) workaround for atan(0, 0), match atan2f(0, 0) == 0
if (op_type == 10)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v2_safe = mix(v2, afpvec4(1.f), zero_mask);
res = atan(v1, v2_safe);
}
if (op_type == 11)
{
bvec4 zero_mask = bvec4(uvec4(equal(v1, afpvec4(0.f))) & uvec4(equal(v2, afpvec4(0.f))));
afpvec4 v1_safe = mix(v1, afpvec4(1.f), zero_mask);
res = atan(v2, v1_safe);
}
#else
if (op_type == 10) res = atan(v1, v2);
if (op_type == 11) res = atan(v2, v1);
#endif
#endif
if (op_type == 12) res = v1 - trunc(v1 / v2) * v2;
if (op_type == 13) res = v2 - trunc(v2 / v1) * v1;
Expand Down
Loading