mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-15 04:30:12 +00:00
6a7770b7ae
This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192750 91177308-0d34-0410-b5e6-96231b3b80d8
38 lines
1.5 KiB
LLVM
38 lines
1.5 KiB
LLVM
; RUN: llc < %s -mcpu=corei7-avx -mtriple=x86_64-linux | FileCheck %s
|
|
|
|
;CHECK-LABEL: cftx020:
|
|
;CHECK: vmovsd (%rdi), %xmm{{.*}}
|
|
;CHECK: vmovsd 16(%rdi), %xmm{{.*}}
|
|
;CHECK: vmovsd 24(%rdi), %xmm{{.*}}
|
|
;CHECK: vmovhpd 8(%rdi), %xmm{{.*}}
|
|
;CHECK: vmovupd %xmm{{.*}}, (%rdi)
|
|
;CHECK: vmovupd %xmm{{.*}}, 16(%rdi)
|
|
;CHECK: ret
|
|
|
|
; A test from pifft (after SLP-vectorization) that fails when we drop the chain on newly merged loads.
|
|
define void @cftx020(double* nocapture %a) {
|
|
entry:
|
|
%0 = load double* %a, align 8
|
|
%arrayidx1 = getelementptr inbounds double* %a, i64 2
|
|
%1 = load double* %arrayidx1, align 8
|
|
%arrayidx2 = getelementptr inbounds double* %a, i64 1
|
|
%2 = load double* %arrayidx2, align 8
|
|
%arrayidx3 = getelementptr inbounds double* %a, i64 3
|
|
%3 = load double* %arrayidx3, align 8
|
|
%4 = insertelement <2 x double> undef, double %0, i32 0
|
|
%5 = insertelement <2 x double> %4, double %3, i32 1
|
|
%6 = insertelement <2 x double> undef, double %1, i32 0
|
|
%7 = insertelement <2 x double> %6, double %2, i32 1
|
|
%8 = fadd <2 x double> %5, %7
|
|
%9 = bitcast double* %a to <2 x double>*
|
|
store <2 x double> %8, <2 x double>* %9, align 8
|
|
%10 = insertelement <2 x double> undef, double %0, i32 0
|
|
%11 = insertelement <2 x double> %10, double %2, i32 1
|
|
%12 = insertelement <2 x double> undef, double %1, i32 0
|
|
%13 = insertelement <2 x double> %12, double %3, i32 1
|
|
%14 = fsub <2 x double> %11, %13
|
|
%15 = bitcast double* %arrayidx1 to <2 x double>*
|
|
store <2 x double> %14, <2 x double>* %15, align 8
|
|
ret void
|
|
}
|