comparison clang/test/CodeGenCUDA/fp-contract.cu @ 236:c4bab56944e8 llvm-original

LLVM 16
author kono
date Wed, 09 Nov 2022 17:45:10 +0900
parents 79ff65ed7e25
children
comparison
equal deleted inserted replaced
232:70dce7da266c 236:c4bab56944e8
103 // RUN: -O3 -target-cpu gfx906 -o - -x ir %t.ll \ 103 // RUN: -O3 -target-cpu gfx906 -o - -x ir %t.ll \
104 // RUN: | FileCheck -check-prefixes=COMMON,AMD-OPT-FASTSTD %s 104 // RUN: | FileCheck -check-prefixes=COMMON,AMD-OPT-FASTSTD %s
105 105
106 // Explicit -ffp-contract=on -- fusing by front-end. 106 // Explicit -ffp-contract=on -- fusing by front-end.
107 // In IR, 107 // In IR,
108 // mult/add in the same statement - llvm.fmuladd instrinsic emitted 108 // mult/add in the same statement - llvm.fmuladd intrinsic emitted
109 // mult/add in different statement - fmul/fadd instructions without 109 // mult/add in different statement - fmul/fadd instructions without
110 // contract flag are emitted. 110 // contract flag are emitted.
111 // In backend 111 // In backend
112 // nvptx/amdgcn - assumes standard fp fuse option, which only 112 // nvptx/amdgcn - assumes standard fp fuse option, which only
113 // fuses mult/add insts with contract flag or 113 // fuses mult/add insts with contract flag or