`pyptx.debugkit`¶

Kernel debugging utilities: beacon waits and phase cycle counters.

Warp-specialized kernels fail in two characteristic ways that are painful to diagnose in environments without profiler access (containers commonly block ncu and cuda-gdb):

Deadlocks — some barrier wait never completes, the kernel hangs, and the host can never read anything back because the launch never finishes.
Latency mysteries — the pipeline runs but slower than it should, and you need per-phase cycle counts to see which wait is eating time.

DebugKit packages the two patterns that proved reliable while debugging the Blackwell flash-attention kernels:

Beacon waits: kit.wait(mbar, phase, site=...) compiles (when enabled) to a bounded mbarrier.try_wait loop. On timeout the thread records a first-blame (site, extra) code into a dedicated debug buffer and falls through pretending the wait succeeded, so the whole pipeline drains, the kernel exits, and the host can decode exactly which wait deadlocked first. When disabled it emits a plain blocking wait with zero overhead.
Cycle counters: t = kit.stamp(); ...; kit.accumulate("name", t) accumulates %clock deltas per thread; kit.flush() writes them to the debug buffer for host-side readout.

The kit needs a dedicated debug output tensor — never point it at a buffer the kernel also writes as real output; epilogue stores will clobber the diagnostics (ask us how we know). Declare an extra Tile(n_slots, SLOT_WORDS, b32) output when building in debug mode and attach it once at the top of the kernel body::

kit = DebugKit(enabled=debug)
...
def body(Q, K, V, O, DBG=None):
    ...
    slot = ...  # e.g. ctaid.x * block_threads + tid
    if kit.enabled:
        kit.attach(ptx.global_ptrs(DBG)[0], slot)

Host side, after the (now guaranteed to finish) run::

for rec in debugkit.decode_beacons(dbg_tensor):
    print(rec)  # {"slot": ..., "site": "p_full", "extra": 1}

`BEACON_MAGIC`¶

Kind: namespace
Type: int

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

Members¶

`conjugate`¶

Kind: attribute
Value: <built-in method conjugate of int object at 0x7f106ecf3490>

Returns self, the complex conjugate of any int.

`bit_length`¶

Kind: attribute
Value: <built-in method bit_length of int object at 0x7f106ecf3490>

Number of bits necessary to represent self in binary.

`bit_count`¶

Kind: attribute
Value: <built-in method bit_count of int object at 0x7f106ecf3490>

Number of ones in the binary representation of the absolute value of self.

Also known as the population count.

`to_bytes`¶

Kind: attribute
Value: <built-in method to_bytes of int object at 0x7f106ecf3490>

Return an array of bytes representing an integer.

length Length of bytes object to use. An OverflowError is raised if the integer is not representable with the given number of bytes. Default is length 1. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Determines whether two's complement is used to represent the integer. If signed is False and a negative integer is given, an OverflowError is raised.

`from_bytes`¶

Kind: attribute
Value: <built-in method from_bytes of type object at 0x7f106fb30840>

Return the integer represented by the given array of bytes.

bytes Holds the array of bytes to convert. The argument must either support the buffer protocol or be an iterable object producing bytes. Bytes and bytearray are examples of built-in objects that support the buffer protocol. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Indicates whether two's complement is used to represent the integer.

`as_integer_ratio`¶

Kind: attribute
Value: <built-in method as_integer_ratio of int object at 0x7f106ecf3490>

Return a pair of integers, whose ratio is equal to the original int.

The ratio is in lowest terms and has a positive denominator.

`is_integer`¶

Kind: attribute
Value: <built-in method is_integer of int object at 0x7f106ecf3490>

Returns True. Exists for duck type compatibility with float.is_integer.

`real`¶

Kind: attribute
Value: 48812

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`imag`¶

Kind: attribute
Value: 0

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`numerator`¶

Kind: attribute
Value: 48812

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`denominator`¶

Kind: attribute
Value: 1

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`SLOT_WORDS`¶

Kind: namespace
Type: int

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

Members¶

`conjugate`¶

Kind: attribute
Value: <built-in method conjugate of int object at 0x7f106fc1d3c8>

Returns self, the complex conjugate of any int.

`bit_length`¶

Kind: attribute
Value: <built-in method bit_length of int object at 0x7f106fc1d3c8>

Number of bits necessary to represent self in binary.

`bit_count`¶

Kind: attribute
Value: <built-in method bit_count of int object at 0x7f106fc1d3c8>

Number of ones in the binary representation of the absolute value of self.

Also known as the population count.

`to_bytes`¶

Kind: attribute
Value: <built-in method to_bytes of int object at 0x7f106fc1d3c8>

Return an array of bytes representing an integer.

length Length of bytes object to use. An OverflowError is raised if the integer is not representable with the given number of bytes. Default is length 1. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Determines whether two's complement is used to represent the integer. If signed is False and a negative integer is given, an OverflowError is raised.

`from_bytes`¶

Kind: attribute
Value: <built-in method from_bytes of type object at 0x7f106fb30840>

Return the integer represented by the given array of bytes.

bytes Holds the array of bytes to convert. The argument must either support the buffer protocol or be an iterable object producing bytes. Bytes and bytearray are examples of built-in objects that support the buffer protocol. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Indicates whether two's complement is used to represent the integer.

`as_integer_ratio`¶

Kind: attribute
Value: <built-in method as_integer_ratio of int object at 0x7f106fc1d3c8>

Return a pair of integers, whose ratio is equal to the original int.

The ratio is in lowest terms and has a positive denominator.

`is_integer`¶

Kind: attribute
Value: <built-in method is_integer of int object at 0x7f106fc1d3c8>

Returns True. Exists for duck type compatibility with float.is_integer.

`real`¶

Kind: attribute
Value: 8

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`imag`¶

Kind: attribute
Value: 0

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`numerator`¶

Kind: attribute
Value: 8

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`denominator`¶

Kind: attribute
Value: 1

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`DEFAULT_TIMEOUT_TRIES`¶

Kind: namespace
Type: int

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

Members¶

`conjugate`¶

Kind: attribute
Value: <built-in method conjugate of int object at 0x7f106ecf0c70>

Returns self, the complex conjugate of any int.

`bit_length`¶

Kind: attribute
Value: <built-in method bit_length of int object at 0x7f106ecf0c70>

Number of bits necessary to represent self in binary.

`bit_count`¶

Kind: attribute
Value: <built-in method bit_count of int object at 0x7f106ecf0c70>

Number of ones in the binary representation of the absolute value of self.

Also known as the population count.

`to_bytes`¶

Kind: attribute
Value: <built-in method to_bytes of int object at 0x7f106ecf0c70>

Return an array of bytes representing an integer.

length Length of bytes object to use. An OverflowError is raised if the integer is not representable with the given number of bytes. Default is length 1. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Determines whether two's complement is used to represent the integer. If signed is False and a negative integer is given, an OverflowError is raised.

`from_bytes`¶

Kind: attribute
Value: <built-in method from_bytes of type object at 0x7f106fb30840>

Return the integer represented by the given array of bytes.

bytes Holds the array of bytes to convert. The argument must either support the buffer protocol or be an iterable object producing bytes. Bytes and bytearray are examples of built-in objects that support the buffer protocol. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Indicates whether two's complement is used to represent the integer.

`as_integer_ratio`¶

Kind: attribute
Value: <built-in method as_integer_ratio of int object at 0x7f106ecf0c70>

Return a pair of integers, whose ratio is equal to the original int.

The ratio is in lowest terms and has a positive denominator.

`is_integer`¶

Kind: attribute
Value: <built-in method is_integer of int object at 0x7f106ecf0c70>

Returns True. Exists for duck type compatibility with float.is_integer.

`real`¶

Kind: attribute
Value: 3000000

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`imag`¶

Kind: attribute
Value: 0

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`numerator`¶

Kind: attribute
Value: 3000000

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`denominator`¶

Kind: attribute
Value: 1

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`DebugKit`¶

Kind: class

class DebugKit(enabled: 'bool' = False, timeout_tries: 'int' = 3000000, _sites: 'list[str]' = <factory>, _counters: 'list[str]' = <factory>, _counter_regs: 'dict' = <factory>, _ptr: 'Any' = None, _slot_off: 'Any' = None, _blamed: 'Any' = None, _label_n: 'int' = 0) -> None

Trace-time helper that instruments waits and phases in a kernel.

Members¶

`enabled`¶

Kind: attribute
Value: False

bool(x) -> bool

Returns True when the argument x is true, False otherwise. The builtins True and False are the only two instances of the class bool. The class bool is a subclass of the class int, and cannot be subclassed.

`timeout_tries`¶

Kind: attribute
Value: 3000000

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`attach(dbg_ptr: 'Any', slot_reg: 'Any') -> 'None'`¶

Kind: method

Bind the debug buffer pointer and this thread's slot index.

slot_reg is a u32 register unique per participating thread (e.g. ctaid.x * threads_per_cta + tid). Must be called before any wait/flush when the kit is enabled.

`wait(mbar: 'Any', phase: 'Any', *, site: 'str', extra: 'int' = 0) -> 'None'`¶

Kind: method

mbarrier parity wait; in debug mode, bounded with first-blame recording and fall-through on timeout.

`stamp() -> 'Any'`¶

Kind: method

Return a u32 register holding %clock (None when disabled).

`accumulate(name: 'str', t0: 'Any') -> 'None'`¶

Kind: method

counter[name] += clock() - t0.

`flush() -> 'None'`¶

Kind: method

Write all counters to this thread's slot (words 1..).

`decode(dbg_tensor) -> 'dict'`¶

Kind: method

Decode a debug tensor produced by this kit.

Returns {"beacons": [...], "counters": {slot: {name: cycles}}}. Beacons are aggregated by (site, extra) with slot lists.

`n_slot_words`¶

Kind: function

n_slot_words(n_slots: 'int') -> 'int'

Total b32 words to allocate for a debug buffer of n_slots.

`WAITMAP_MAX_SITES`¶

Kind: namespace
Type: int

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

Members¶

`conjugate`¶

Kind: attribute
Value: <built-in method conjugate of int object at 0x7f106fc1d4c8>

Returns self, the complex conjugate of any int.

`bit_length`¶

Kind: attribute
Value: <built-in method bit_length of int object at 0x7f106fc1d4c8>

Number of bits necessary to represent self in binary.

`bit_count`¶

Kind: attribute
Value: <built-in method bit_count of int object at 0x7f106fc1d4c8>

Number of ones in the binary representation of the absolute value of self.

Also known as the population count.

`to_bytes`¶

Kind: attribute
Value: <built-in method to_bytes of int object at 0x7f106fc1d4c8>

Return an array of bytes representing an integer.

length Length of bytes object to use. An OverflowError is raised if the integer is not representable with the given number of bytes. Default is length 1. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Determines whether two's complement is used to represent the integer. If signed is False and a negative integer is given, an OverflowError is raised.

`from_bytes`¶

Kind: attribute
Value: <built-in method from_bytes of type object at 0x7f106fb30840>

Return the integer represented by the given array of bytes.

bytes Holds the array of bytes to convert. The argument must either support the buffer protocol or be an iterable object producing bytes. Bytes and bytearray are examples of built-in objects that support the buffer protocol. byteorder The byte order used to represent the integer. If byteorder is 'big', the most significant byte is at the beginning of the byte array. If byteorder is 'little', the most significant byte is at the end of the byte array. To request the native byte order of the host system, use `sys.byteorder' as the byte order value. Default is to use 'big'. signed Indicates whether two's complement is used to represent the integer.

`as_integer_ratio`¶

Kind: attribute
Value: <built-in method as_integer_ratio of int object at 0x7f106fc1d4c8>

Return a pair of integers, whose ratio is equal to the original int.

The ratio is in lowest terms and has a positive denominator.

`is_integer`¶

Kind: attribute
Value: <built-in method is_integer of int object at 0x7f106fc1d4c8>

Returns True. Exists for duck type compatibility with float.is_integer.

`real`¶

Kind: attribute
Value: 16

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`imag`¶

Kind: attribute
Value: 0

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`numerator`¶

Kind: attribute
Value: 16

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`denominator`¶

Kind: attribute
Value: 1

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

`WaitMap`¶

Kind: class

class WaitMap(_sites: 'list' = <factory>, _accs: 'dict' = <factory>, _ptr: 'Any' = None, _row_off: 'Any' = None) -> None

Wraps every ptx.mbarrier.wait and ptx.bar.sync emitted while active, accumulating per-site %clock cycles into per-thread counters. Sites are auto-named from the caller's source line, so kernels need no source changes beyond attach/flush.

Usage in a builder::

wm = WaitMap()
...
def body(...):
    if wm_enabled:
        wm.attach_flush_late(...)  # see attach()/flush()
    with wm.active():
        ...   # trace the kernel body; waits get instrumented

Buffer layout: row per thread (tid), WAITMAP_MAX_SITES words per row.

Members¶

`site_names() -> 'list'`¶

Kind: method

No docstring yet.

`attach(dbg_ptr: 'Any', tid_reg: 'Any') -> 'None'`¶

Kind: method

No docstring yet.

`active()`¶

Kind: method

Context manager: instrument waits emitted inside.

`flush(pred: 'Any' = None) -> 'None'`¶

Kind: method

Write all site counters for this thread (guard with a pred that selects one CTA to keep the buffer small).

`decode(dbg_tensor, n_threads: 'int' = 512, roles: 'dict | None' = None) -> 'dict'`¶

Kind: method

Aggregate per-site cycles by role. roles: {name: (lo_tid, hi_tid)}.

pyptx.debugkit¶

Public API¶

BEACON_MAGIC¶

Members¶

conjugate¶

bit_length¶

bit_count¶

to_bytes¶

from_bytes¶

as_integer_ratio¶

is_integer¶

real¶

imag¶

numerator¶

denominator¶

SLOT_WORDS¶

Members¶

conjugate¶

bit_length¶

bit_count¶

to_bytes¶

from_bytes¶

as_integer_ratio¶

is_integer¶

real¶

imag¶

numerator¶

denominator¶

DEFAULT_TIMEOUT_TRIES¶

Members¶

conjugate¶

bit_length¶

bit_count¶

to_bytes¶

from_bytes¶

as_integer_ratio¶

is_integer¶

real¶

imag¶

numerator¶

denominator¶

DebugKit¶

Members¶

enabled¶

timeout_tries¶

attach(dbg_ptr: 'Any', slot_reg: 'Any') -> 'None'¶

wait(mbar: 'Any', phase: 'Any', *, site: 'str', extra: 'int' = 0) -> 'None'¶

stamp() -> 'Any'¶

accumulate(name: 'str', t0: 'Any') -> 'None'¶

flush() -> 'None'¶

decode(dbg_tensor) -> 'dict'¶

n_slot_words¶

WAITMAP_MAX_SITES¶

Members¶

conjugate¶

bit_length¶

bit_count¶

to_bytes¶

from_bytes¶

as_integer_ratio¶

is_integer¶

real¶

imag¶

numerator¶

denominator¶

WaitMap¶

Members¶

site_names() -> 'list'¶

attach(dbg_ptr: 'Any', tid_reg: 'Any') -> 'None'¶

active()¶

flush(pred: 'Any' = None) -> 'None'¶

decode(dbg_tensor, n_threads: 'int' = 512, roles: 'dict | None' = None) -> 'dict'¶

`pyptx.debugkit`¶

`BEACON_MAGIC`¶

`conjugate`¶

`bit_length`¶

`bit_count`¶

`to_bytes`¶

`from_bytes`¶

`as_integer_ratio`¶

`is_integer`¶

`real`¶

`imag`¶

`numerator`¶

`denominator`¶

`SLOT_WORDS`¶

`conjugate`¶

`bit_length`¶

`bit_count`¶

`to_bytes`¶

`from_bytes`¶

`as_integer_ratio`¶

`is_integer`¶

`real`¶

`imag`¶

`numerator`¶

`denominator`¶

`DEFAULT_TIMEOUT_TRIES`¶

`conjugate`¶

`bit_length`¶

`bit_count`¶

`to_bytes`¶

`from_bytes`¶

`as_integer_ratio`¶

`is_integer`¶

`real`¶

`imag`¶

`numerator`¶

`denominator`¶

`DebugKit`¶

`enabled`¶

`timeout_tries`¶

`attach(dbg_ptr: 'Any', slot_reg: 'Any') -> 'None'`¶

`wait(mbar: 'Any', phase: 'Any', *, site: 'str', extra: 'int' = 0) -> 'None'`¶

`stamp() -> 'Any'`¶

`accumulate(name: 'str', t0: 'Any') -> 'None'`¶

`flush() -> 'None'`¶

`decode(dbg_tensor) -> 'dict'`¶

`n_slot_words`¶

`WAITMAP_MAX_SITES`¶

`conjugate`¶

`bit_length`¶

`bit_count`¶

`to_bytes`¶

`from_bytes`¶

`as_integer_ratio`¶

`is_integer`¶

`real`¶

`imag`¶

`numerator`¶

`denominator`¶

`WaitMap`¶

`site_names() -> 'list'`¶

`attach(dbg_ptr: 'Any', tid_reg: 'Any') -> 'None'`¶

`active()`¶

`flush(pred: 'Any' = None) -> 'None'`¶

`decode(dbg_tensor, n_threads: 'int' = 512, roles: 'dict | None' = None) -> 'dict'`¶