LLVM 24.0.0git
AMDGPUCoExecSchedStrategy.cpp File Reference

Coexecution-focused scheduling strategy for AMDGPU. More...

Go to the source code of this file.

Macros

#define DEBUG_TYPE   "machine-scheduler"

Functions

static SUnit * pickOnlyChoice (SchedBoundary &Zone)
static void allowCoExec (llvm::AMDGPU::CoExecInfo &Info, llvm::AMDGPU::CoExecMaskT ExtraBits, unsigned StartIndex)
 Apply ExtraBits to every slot in Info starting with StartIndex Used by MFMA co-exec rules, because MFMA co-exec slots are incremental, i.e.

Variables

static constexpr unsigned DefaultCoExecVGPRThresholdPercent = 100

Detailed Description

Coexecution-focused scheduling strategy for AMDGPU.

Definition in file AMDGPUCoExecSchedStrategy.cpp.

Macro Definition Documentation

◆ DEBUG_TYPE

#define DEBUG_TYPE   "machine-scheduler"

Definition at line 23 of file AMDGPUCoExecSchedStrategy.cpp.

Function Documentation

◆ allowCoExec()

void allowCoExec ( llvm::AMDGPU::CoExecInfo & Info,
llvm::AMDGPU::CoExecMaskT ExtraBits,
unsigned StartIndex )
static

Apply ExtraBits to every slot in Info starting with StartIndex Used by MFMA co-exec rules, because MFMA co-exec slots are incremental, i.e.

for every slot N it supports all instructions which were supported by the previous slot N-1 and may support something extra.

Definition at line 55 of file AMDGPUCoExecSchedStrategy.cpp.

Referenced by llvm::AMDGPU::getMFMACoExecInfo().

◆ pickOnlyChoice()

Variable Documentation

◆ DefaultCoExecVGPRThresholdPercent

unsigned DefaultCoExecVGPRThresholdPercent = 100
staticconstexpr