microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-05-11 17:00:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

debug

Browse Source
This commit is contained in:
Ubuntu
2026-03-06 18:25:03 +00:00
committed by Ubuntu
parent ad56728c6d
commit e487f831e6
4 changed files with 237 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
cmake/FindGDRCopy.cmake
View File

@@ -30,15 +30,19 @@ find_library(GDRCOPY_LIBRARIES
${GDRCOPY_ROOT_DIR}/lib
/usr/local/lib
/usr/lib
/usr/lib/x86_64-linux-gnu)
/usr/lib/x86_64-linux-gnu
/usr/lib/aarch64-linux-gnu)
if(GDRCOPY_INCLUDE_DIRS)
include(CheckSymbolExists)
include(CheckCXXSourceCompiles)
set(CMAKE_REQUIRED_INCLUDES ${GDRCOPY_INCLUDE_DIRS})
set(CMAKE_REQUIRED_LIBRARIES ${GDRCOPY_LIBRARIES})
check_symbol_exists(gdr_pin_buffer_v2 "gdrapi.h" GDRCOPY_HAS_PIN_BUFFER_V2)
unset(CMAKE_REQUIRED_LIBRARIES)
check_cxx_source_compiles("
#include <gdrapi.h>
int main() { gdr_pin_buffer_v2(0, 0, 0, 0, 0); return 0; }
" GDRCOPY_HAS_PIN_BUFFER_V2)
unset(CMAKE_REQUIRED_INCLUDES)
unset(CMAKE_REQUIRED_LIBRARIES)
if(NOT GDRCOPY_HAS_PIN_BUFFER_V2)
message(STATUS "GDRCopy found but too old (gdr_pin_buffer_v2 not available). Requires >= 2.5.")
set(GDRCOPY_INCLUDE_DIRS GDRCOPY_INCLUDE_DIRS-NOTFOUND)

2
include/mscclpp/core.hpp
View File

@@ -390,7 +390,7 @@ struct EndpointConfig {
};
static constexpr int DefaultPort = -1;
static constexpr int DefaultGidIndex = 0;
static constexpr int DefaultGidIndex = 3;
static constexpr int DefaultMaxCqSize = 1024;
static constexpr int DefaultMaxCqPollNum = 1;
static constexpr int DefaultMaxSendWr = 8192;

15
python/mscclpp/language/rank.py
View File

@@ -304,11 +304,16 @@ class BaseBuffer:
self.size = offset + size
def __getitem__(self, key):
if self.offset + key.stop > self.size:
raise RuntimeError(
f"Index range from {self.offset + key.start} - {self.offset + key.stop} is out of bounds for buffer {self.buffer_type}. Buffer size: {self.size}"
)
return Chunk(self.rank, self.buffer_type, self.offset + key.start, key.stop - key.start)
if isinstance(key, slice):
start = key.start if key.start is not None else 0
stop = key.stop if key.stop is not None else (self.size - self.offset)
if self.offset + stop > self.size:
raise RuntimeError(
f"Index range from {self.offset + start} - {self.offset + stop} is out of bounds for buffer {self.buffer_type}. Buffer size: {self.size}"
)
return Chunk(self.rank, self.buffer_type, self.offset + start, stop - start)
else:
raise TypeError(f"Buffer indices must be slices, not {type(key).__name__}")
class Buffer(BaseBuffer):

218
test.json Normal file
View File

@@ -0,0 +1,218 @@
{
"name": "send_recv_test",
"collective": "test",
"protocol": "Simple",
"inplace": false,
"reuse_resources": false,
"gpus": [
{
"id": 0,
"input_chunks": 1,
"output_chunks": 1,
"scratch_chunks": 0,
"threadblocks": [
{
"id": 0,
"ops": [
{
"name": "signal",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "wait",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "nop"
},
{
"name": "put",
"src_buff": [
{
"type": "i",
"index": 0,
"size": 1
}
],
"dst_buff": [
{
"buffer_id": 0,
"index": 0,
"size": 1
}
],
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "nop"
},
{
"name": "signal",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "wait",
"channel_ids": [
0
],
"channel_type": "port"
}
],
"channels": [
{
"channel_type": "port",
"channel_ids": [
0
]
}
],
"remote_buffer_refs": [
{
"access_channel_type": "port",
"remote_buffer_ids": [
0
]
}
]
}
],
"channels": [
{
"channel_type": "port",
"connected_to": [
1
]
}
],
"remote_buffers": [
{
"rank": 1,
"type": "o",
"access_channel_types": [
"port"
]
}
],
"semaphores": []
},
{
"id": 1,
"input_chunks": 1,
"output_chunks": 1,
"scratch_chunks": 0,
"threadblocks": [
{
"id": 0,
"ops": [
{
"name": "signal",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "wait",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "nop"
},
{
"name": "put",
"src_buff": [
{
"type": "i",
"index": 0,
"size": 1
}
],
"dst_buff": [
{
"buffer_id": 0,
"index": 0,
"size": 1
}
],
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "nop"
},
{
"name": "signal",
"channel_ids": [
0
],
"channel_type": "port"
},
{
"name": "wait",
"channel_ids": [
0
],
"channel_type": "port"
}
],
"channels": [
{
"channel_type": "port",
"channel_ids": [
0
]
}
],
"remote_buffer_refs": [
{
"access_channel_type": "port",
"remote_buffer_ids": [
0
]
}
]
}
],
"channels": [
{
"channel_type": "port",
"connected_to": [
0
]
}
],
"remote_buffers": [
{
"rank": 0,
"type": "o",
"access_channel_types": [
"port"
]
}
],
"semaphores": []
}
],
"num_threads_per_block": 1024,
"use_double_scratch_buffer": false,
"buffer_alignment": 16,
"min_message_size": 0,
"max_message_size": 18446744073709551615
}