diff options
| author | Daniil Fukalov <daniil.fukalov@amd.com> | 2019-07-17 16:51:29 +0000 |
|---|---|---|
| committer | Daniil Fukalov <daniil.fukalov@amd.com> | 2019-07-17 16:51:29 +0000 |
| commit | d912a9ba9b1647984dc65d34ba4422874bc28855 (patch) | |
| tree | 84cf7c74620164f6f92fb647f3faf67acfccde39 /llvm/lib/Target | |
| parent | 3fce6b5da169c77c4047a4632f28b41d33d7c0a9 (diff) | |
| download | bcm5719-llvm-d912a9ba9b1647984dc65d34ba4422874bc28855.tar.gz bcm5719-llvm-d912a9ba9b1647984dc65d34ba4422874bc28855.zip | |
[AMDGPU] Tune inlining parameters for AMDGPU target
Summary:
Since the target has no significant advantage of vectorization,
vector instructions bous threshold bonus should be optional.
amdgpu-inline-arg-alloca-cost parameter default value and the target
InliningThresholdMultiplier value tuned then respectively.
Reviewers: arsenm, rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, eraman, hiraditya, haicheng, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64642
llvm-svn: 366348
Diffstat (limited to 'llvm/lib/Target')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUInline.cpp | 2 | ||||
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h | 4 |
2 files changed, 4 insertions, 2 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp index ec0dd6df44a..f4df20b8f03 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp @@ -39,7 +39,7 @@ using namespace llvm; #define DEBUG_TYPE "inline" static cl::opt<int> -ArgAllocaCost("amdgpu-inline-arg-alloca-cost", cl::Hidden, cl::init(2200), +ArgAllocaCost("amdgpu-inline-arg-alloca-cost", cl::Hidden, cl::init(1500), cl::desc("Cost of alloca argument")); // If the amount of scratch memory to eliminate exceeds our ability to allocate diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h index 72882c83c01..6f1bf5a26f0 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h +++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h @@ -191,7 +191,9 @@ public: bool areInlineCompatible(const Function *Caller, const Function *Callee) const; - unsigned getInliningThresholdMultiplier() { return 9; } + unsigned getInliningThresholdMultiplier() { return 7; } + + int getInlinerVectorBonusPercent() { return 0; } int getArithmeticReductionCost(unsigned Opcode, Type *Ty, |
