mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-09-05 17:30:02 +00:00
BBVectorize would, except for loads and stores, always fuse instructions so that the first instruction (in the current source order) would always represent the low part of the input vectors and the second instruction would always represent the high part. This lead to too many shuffles being produced because sometimes the opposite order produces fewer of them. With this change, BBVectorize tracks the kind of pair connections that form the DAG of candidate pairs, and uses that information to reorder the pairs to avoid excess shuffles. Using this information, a future commit will be able to add VTTI-based shuffle costs to the pair selection procedure. Importantly, the number of remaining shuffles can now be estimated during pair selection. There are some trivial instruction reorderings in the test cases, and one simple additional test where we certainly want to do a reordering to avoid an unnecessary shuffle. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167122 91177308-0d34-0410-b5e6-96231b3b80d8
47 lines
2.1 KiB
LLVM
47 lines
2.1 KiB
LLVM
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
|
|
; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -instcombine -gvn -S | FileCheck %s
|
|
; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -bb-vectorize-search-limit=4 -instcombine -gvn -S | FileCheck %s -check-prefix=CHECK-SL4
|
|
|
|
define double @test1(double %A1, double %A2, double %B1, double %B2) {
|
|
; CHECK: @test1
|
|
; CHECK-SL4: @test1
|
|
; CHECK-SL4-NOT: <2 x double>
|
|
; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0
|
|
; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1
|
|
; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0
|
|
; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1
|
|
%X1 = fsub double %A1, %B1
|
|
%X2 = fsub double %A2, %B2
|
|
; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2
|
|
%Y1 = fmul double %X1, %A1
|
|
%Y2 = fmul double %X2, %A2
|
|
; CHECK: %Y1 = fmul <2 x double> %X1, %X1.v.i0.2
|
|
%Z1 = fadd double %Y1, %B1
|
|
; Here we have a dependency chain: the short search limit will not
|
|
; see past this chain and so will not see the second part of the
|
|
; pair to vectorize.
|
|
%mul41 = fmul double %Z1, %Y2
|
|
%sub48 = fsub double %Z1, %mul41
|
|
%mul62 = fmul double %Z1, %sub48
|
|
%sub69 = fsub double %Z1, %mul62
|
|
%mul83 = fmul double %Z1, %sub69
|
|
%sub90 = fsub double %Z1, %mul83
|
|
%mul104 = fmul double %Z1, %sub90
|
|
%sub111 = fsub double %Z1, %mul104
|
|
%mul125 = fmul double %Z1, %sub111
|
|
%sub132 = fsub double %Z1, %mul125
|
|
%mul146 = fmul double %Z1, %sub132
|
|
%sub153 = fsub double %Z1, %mul146
|
|
; end of chain.
|
|
%Z2 = fadd double %Y2, %B2
|
|
; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2
|
|
%R1 = fdiv double %Z1, %Z2
|
|
%R = fmul double %R1, %sub153
|
|
; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0
|
|
; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1
|
|
; CHECK: %R1 = fdiv double %Z1.v.r1, %Z1.v.r2
|
|
ret double %R
|
|
; CHECK: ret double %R
|
|
}
|
|
|