Skip to content

Fix support of float atomics in threadgroup memory. #1

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
duburcqa wants to merge 1 commit into
base: main
Choose a base branch
from
+15 −1
Open

Fix support of float atomics in threadgroup memory. #1

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
16 changes: 15 additions & 1 deletion spirv_msl.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11098,6 +11098,20 @@ void CompilerMSL::emit_atomic_func_op(uint32_t result_type, uint32_t result_id,
else if (opcode == OpAtomicSMax || opcode == OpAtomicSMin)
expected_type = to_signed_basetype(type.width);

// Metal does not support float atomics in threadgroup memory. When the
// atomic result type is integer but the pointee type is float (CAS-based
// emulation of float atomics), override the pointer cast type to emit
// atomic_uint / atomic_ulong instead of atomic_float. We leave
// expected_type unchanged so the result bitcast logic is not affected.
auto atomic_cast_type = expected_type;
if (ptr_type.storage == spv::StorageClassWorkgroup &&
type_is_floating_point(type))
{
auto &res_type = get<SPIRType>(result_type);
if (type_is_integral(res_type))
atomic_cast_type = res_type.basetype;
}

bool use_native_image_atomic;
if (msl_options.supports_msl_version(3, 1))
use_native_image_atomic = check_atomic_image(obj);
Expand All @@ -11108,7 +11122,7 @@ void CompilerMSL::emit_atomic_func_op(uint32_t result_type, uint32_t result_id,
SPIRV_CROSS_THROW("MSL currently does not support 64-bit atomics.");

auto remapped_type = type;
remapped_type.basetype = expected_type;
remapped_type.basetype = atomic_cast_type;

auto *var = maybe_get_backing_variable(obj);
const auto *res_type = var ? &get<SPIRType>(var->basetype) : nullptr;
Expand Down