1 //===-- AMDGPUFeatures.td - AMDGPU Feature Definitions -----*- tablegen -*-===//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 def FeatureFP64 : SubtargetFeature<"fp64",
12 "Enable double precision operations"
15 def FeatureFMA : SubtargetFeature<"fmaf",
18 "Enable single precision FMA (not as fast as mul+add, but fused)"
21 // Some instructions do not support denormals despite this flag. Using
22 // fp32 denormals also causes instructions to run at the double
23 // precision rate for the device.
24 def FeatureFP32Denormals : SubtargetFeature<"fp32-denormals",
27 "Enable single precision denormal handling"
30 class SubtargetFeatureLocalMemorySize <int Value> : SubtargetFeature<
31 "localmemorysize"#Value,
34 "The size of local memory in bytes"
37 def FeatureLocalMemorySize0 : SubtargetFeatureLocalMemorySize<0>;
38 def FeatureLocalMemorySize32768 : SubtargetFeatureLocalMemorySize<32768>;
39 def FeatureLocalMemorySize65536 : SubtargetFeatureLocalMemorySize<65536>;
41 class SubtargetFeatureWavefrontSize <int Value> : SubtargetFeature<
42 "wavefrontsize"#Value,
45 "The number of threads per wavefront"
48 def FeatureWavefrontSize16 : SubtargetFeatureWavefrontSize<16>;
49 def FeatureWavefrontSize32 : SubtargetFeatureWavefrontSize<32>;
50 def FeatureWavefrontSize64 : SubtargetFeatureWavefrontSize<64>;
52 class SubtargetFeatureGeneration <string Value, string FeatureName,
54 list<SubtargetFeature> Implies> :
55 SubtargetFeature <FeatureName, "Gen", Subtarget#"::"#Value,
56 Value#" GPU generation", Implies>;
58 def FeaturePromoteAlloca : SubtargetFeature <"promote-alloca",
59 "EnablePromoteAlloca",
61 "Enable promote alloca pass"