Untrack allocation

@brief Set the per-wavefront scratch base address. @param val Scratch base byte address (set at dispatch by CP).

@brief Unified VGPR index for this operand (0-511). @details Maps AMDGPU encoding ranges to a unified index space: VGPRs 0-255, AccVGPRs 256-511. Only

@returns This Isa's minor version.

Returns true iff the test part failed.

Compute a stable hash of a SemaBlock's semantic content. Returns a 32-byte SHA-256 digest. Raises: AssertionError: If ``block.enrich

Helper function to convert IB work completion opcode to string

Register an LDS write and validate against outstanding reads.

A kernel_dispatch buffer record mapping a dispatch to a kernel.

@brief Byte offset of the first instruction. @returns Start offset within the code object.

@returns The bumped total per-lane scratch bytes.

The code in this function illustrates how to load a kernel from pre-compiled code. The goal is to get a handle that can be later used in an AQL packet

Map canonical dtype string to SemaType. Unmapped narrow dtypes (i16, u16, i8, u8, i24, u24, b16) intentionally fall back to U32 — the wired c

@brief Encode an s_endpgm instruction for the given target ISA. @param arch Target ISA architecture. @returns The encoded 32-bit instruction word.

Access to raw binary data (mutable).

* Returns the number of threads in the caller's flattened thread block. */

Register a global load into VGPRs (tracked by vmcnt).

Interface for Info Log

Initializes a new NCCL communicator. Creates a communicator that connects multiple ranks. All ranks must call this method wi

Compute input element location for the GFX9 MFMA register layout. @param dim Outer dimension (M for A matrix, N for B matrix) @param K Reduction di

Section size in bytes.

@brief Return the raw status register value. @returns Raw status register value.

Get virtual device for the current command queue

If the test case one big test, you should use RunGenericTest() to run the test case. OTOH, if the test case consists of multiple functions to be run a

Attempt to find and set test->cpu_device and test->gpu_device1

Lower an expression node to a C++ expression string.

Return a C++ expression for a physical VGPR base. gfx1250 uses MODE-controlled high-bank bits for VGPR operands. The generated operan

Returns image's row pitch in bytes

================================================================================================

@brief return whether value is an object @sa https://json.nlohmann.me/api/basic_json/is_object/

Compute 32-bit output element location for the GFX9 MFMA register layout.

* @brief Query Counter info such as name or description. * * @param [in] counter_id counter to get info for * @param [in] version Version of struct

@brief Return the current execution state. @returns Current WfState.