Skip to content

[rocprofiler-sdk] HSA API changes for HSA_AMD_EXT_API_TABLE_STEP_VERSION 0x0A #3823

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
venkat1361 merged 10 commits into from
Mar 9, 2026
Merged

[rocprofiler-sdk] HSA API changes for HSA_AMD_EXT_API_TABLE_STEP_VERSION 0x0A #3823

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
a38620b
add new hsa APIs to rocprofiler-sdk table version 0x0A.
venky1306 Mar 6, 2026
4490e23
add fmt support for new hsa api.
venky1306 Mar 6, 2026
6476f40
Copilot suggestion.
venky1306 Mar 6, 2026
8124113
take _val by const reference.
venky1306 Mar 5, 2026
c5a4487
use internal fmt for ci builds.
venky1306 Mar 5, 2026
fbe4d34
address comments: remove default in hsa_amd_memory_copy_op_type_t and…
venky1306 Mar 6, 2026
57e0f76
add pip install for cmake 3.24. required by rocprofiler-register.
venky1306 Mar 6, 2026
b297d53
also install cmake 3.24 for sanitizers.
venky1306 Mar 6, 2026
0af73c9
bump patch version to 1.
venky1306 Mar 6, 2026
8c38286
add local path to GITHUB_ENV.
venky1306 Mar 6, 2026
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 projects/rocprofiler-sdk/source/include/rocprofiler-sdk/cxx/enum_string.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -401,6 +401,9 @@ ROCPROFILER_ENUM_LABEL(ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_ais_file_read);
ROCPROFILER_ENUM_LABEL(ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_counted_queue_acquire);
ROCPROFILER_ENUM_LABEL(ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_counted_queue_release);
# endif
# if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
ROCPROFILER_ENUM_LABEL(ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_memory_async_batch_copy);
# endif
#endif

#if HSA_AMD_EXT_API_TABLE_MAJOR_VERSION == 0x01
Expand All @@ -426,6 +429,8 @@ static_assert(ROCPROFILER_HSA_AMD_EXT_API_ID_LAST == 74);
static_assert(ROCPROFILER_HSA_AMD_EXT_API_ID_LAST == 76);
# elif HSA_AMD_EXT_API_TABLE_STEP_VERSION == 0x09
static_assert(ROCPROFILER_HSA_AMD_EXT_API_ID_LAST == 78);
# elif HSA_AMD_EXT_API_TABLE_STEP_VERSION == 0x0A
static_assert(ROCPROFILER_HSA_AMD_EXT_API_ID_LAST == 79);
# else
# if !defined(ROCPROFILER_UNSAFE_NO_VERSION_CHECK) && \
(defined(ROCPROFILER_CI) && ROCPROFILER_CI > 0)
Expand Down
3 changes: 3 additions & 0 deletions projects/rocprofiler-sdk/source/include/rocprofiler-sdk/hsa/amd_ext_api_id.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,6 +129,9 @@ typedef enum rocprofiler_hsa_amd_ext_api_id_t // NOLINT(performance-enum-size)
ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_counted_queue_acquire,
ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_counted_queue_release,
# endif
# if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_memory_async_batch_copy,
# endif
#endif

ROCPROFILER_HSA_AMD_EXT_API_ID_LAST,
Expand Down
9 changes: 9 additions & 0 deletions projects/rocprofiler-sdk/source/include/rocprofiler-sdk/hsa/api_args.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1480,6 +1480,15 @@ typedef union rocprofiler_hsa_api_args_t
hsa_queue_t* queue;
} hsa_amd_counted_queue_release;
# endif
# if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
struct
{
const hsa_amd_memory_copy_op_t* copy_ops;
uint32_t num_copy_ops;
uint32_t num_dep_signals;
const hsa_signal_t* dep_signals;
} hsa_amd_memory_async_batch_copy;
# endif
#endif
} rocprofiler_hsa_api_args_t;

Expand Down
5 changes: 5 additions & 0 deletions projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/hsa/abi.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,8 @@ ROCP_SDK_ENFORCE_ABI_VERSIONING(::AmdExtTable, 75);
ROCP_SDK_ENFORCE_ABI_VERSIONING(::AmdExtTable, 77);
#elif HSA_AMD_EXT_API_TABLE_STEP_VERSION == 0x09
ROCP_SDK_ENFORCE_ABI_VERSIONING(::AmdExtTable, 79);
#elif HSA_AMD_EXT_API_TABLE_STEP_VERSION == 0x0A
ROCP_SDK_ENFORCE_ABI_VERSIONING(::AmdExtTable, 80);
#else
INTERNAL_CI_ROCP_SDK_ENFORCE_ABI_VERSIONING(::AmdExtTable, 0);
#endif
Expand Down Expand Up @@ -305,6 +307,9 @@ ROCP_SDK_ENFORCE_ABI(::AmdExtTable, hsa_amd_ais_file_read_fn, 76);
ROCP_SDK_ENFORCE_ABI(::AmdExtTable, hsa_amd_counted_queue_acquire_fn, 77);
ROCP_SDK_ENFORCE_ABI(::AmdExtTable, hsa_amd_counted_queue_release_fn, 78);
#endif
#if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
ROCP_SDK_ENFORCE_ABI(::AmdExtTable, hsa_amd_memory_async_batch_copy_fn, 79);
#endif

ROCP_SDK_ENFORCE_ABI(::ImageExtTable, hsa_ext_image_get_capability_fn, 1);
ROCP_SDK_ENFORCE_ABI(::ImageExtTable, hsa_ext_image_data_get_info_fn, 2);
Expand Down
121 changes: 121 additions & 0 deletions projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/hsa/details/fmt.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,4 +188,125 @@ struct formatter<hsa_amd_ais_file_handle_t>
}
};
#endif
#if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
template <>
struct formatter<hsa_amd_memory_copy_op_type_t>
{
template <typename ParseContext>
constexpr auto parse(ParseContext& ctx)
{
return ctx.begin();
}

template <typename FormatContext>
auto format(hsa_amd_memory_copy_op_type_t const& op, FormatContext& ctx) const
{
switch(op)
{
case HSA_AMD_MEMORY_COPY_OP_LINEAR:
return fmt::format_to(ctx.out(), "HSA_AMD_MEMORY_COPY_OP_LINEAR");
case HSA_AMD_MEMORY_COPY_OP_LINEAR_BROADCAST:
return fmt::format_to(ctx.out(), "HSA_AMD_MEMORY_COPY_OP_LINEAR_BROADCAST");
case HSA_AMD_MEMORY_COPY_OP_LINEAR_SWAP:
return fmt::format_to(ctx.out(), "HSA_AMD_MEMORY_COPY_OP_LINEAR_SWAP");
case HSA_AMD_MEMORY_COPY_OP_LINEAR_INDIRECT:
return fmt::format_to(ctx.out(), "HSA_AMD_MEMORY_COPY_OP_LINEAR_INDIRECT");
default:
return fmt::format_to(
ctx.out(),
"hsa_amd_memory_copy_op_type_t({})",
static_cast<std::underlying_type_t<hsa_amd_memory_copy_op_type_t>>(op));
}
}
};

template <>
struct formatter<hsa_amd_memory_copy_op_t>
{
template <typename ParseContext>
constexpr auto parse(ParseContext& ctx)
{
return ctx.begin();
}

template <typename FormatContext>
auto format(hsa_amd_memory_copy_op_t const& op, FormatContext& ctx) const
{
auto out = fmt::format_to(ctx.out(),
"[MEMORY_COPY_OP type={}, version={}, num_dsts={}, "
"traffic_class={}, completion_signal={}",
op.type,
op.version,
op.num_dsts,
op.traffic_class,
op.completion_signal.handle);

switch(op.type)
{
case HSA_AMD_MEMORY_COPY_OP_LINEAR:
out = fmt::format_to(out,
", src={}, src_agent={}, dst={}, dst_agent={}, "
"size={}, unused_size={}",
fmt::ptr(op.src),
op.src_agent.handle,
fmt::ptr(op.dst),
op.dst_agent.handle,
op.size,
op.unused_size);
break;
case HSA_AMD_MEMORY_COPY_OP_LINEAR_BROADCAST:
out = fmt::format_to(
out,
", src={}, src_agent={}, dst_list={}, dst_agent_list={}, size={}",
fmt::ptr(op.src),
op.src_agent.handle,
fmt::ptr(op.dst_list),
fmt::ptr(op.dst_agent_list),
op.size);
break;
case HSA_AMD_MEMORY_COPY_OP_LINEAR_SWAP:
out = fmt::format_to(out,
", src={}, src_agent={}, dst={}, dst_agent={}, "
"src_size={}, dst_size={}",
fmt::ptr(op.src),
op.src_agent.handle,
fmt::ptr(op.dst),
op.dst_agent.handle,
op.src_size,
op.dst_size);
break;
case HSA_AMD_MEMORY_COPY_OP_LINEAR_INDIRECT:
out = fmt::format_to(out,
", src_ptr={}, src_agent={}, dst_ptr={}, dst_agent={}, "
"size={}, unused_size={}",
fmt::ptr(op.src),
op.src_agent.handle,
fmt::ptr(op.dst),
op.dst_agent.handle,
op.size,
op.unused_size);
break;
default:
out = fmt::format_to(out,
", src={}, src_agent={}, dst={}, dst_agent={}, "
"size={}, unused_size={}",
fmt::ptr(op.src),
op.src_agent.handle,
fmt::ptr(op.dst),
op.dst_agent.handle,
op.size,
op.unused_size);
break;
}

if(op.reserved[0] != 0 || op.reserved[1] != 0 || op.reserved[2] != 0)
{
out = fmt::format_to(
out, ", reserved=[{},{},{}]", op.reserved[0], op.reserved[1], op.reserved[2]);
}

return fmt::format_to(out, "]");
}
};
#endif
} // namespace fmt
10 changes: 10 additions & 0 deletions projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/hsa/hsa.def.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -536,6 +536,16 @@ HSA_API_INFO_DEFINITION_V(ROCPROFILER_HSA_TABLE_ID_AmdExt,
hsa_amd_counted_queue_release_fn,
queue)
# endif
# if HSA_AMD_EXT_API_TABLE_STEP_VERSION >= 0x0A
HSA_API_INFO_DEFINITION_V(ROCPROFILER_HSA_TABLE_ID_AmdExt,
ROCPROFILER_HSA_AMD_EXT_API_ID_hsa_amd_memory_async_batch_copy,
hsa_amd_memory_async_batch_copy,
hsa_amd_memory_async_batch_copy_fn,
copy_ops,
num_copy_ops,
num_dep_signals,
dep_signals)
# endif
# endif

#elif defined(ROCPROFILER_LIB_ROCPROFILER_HSA_ASYNC_COPY_CPP_IMPL) && \
Expand Down
Loading