llvm-mirror/movntdq-no-avx.ll at da1417956356eabd9630af80ca3f0880b67c0d39 - llvm-mirror - Git.je (Gitea)

RPCS3/llvm-mirror

mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 20:23:11 +01:00

Ahmed Bougacha 0c13130c71 [X86] Require 32-byte alignment for 32-byte VMOVNTs.

We used to accept (and even test, and generate) 16-byte alignment
for 32-byte nontemporal stores, but they require 32-byte alignment,
per SDM. Found by inspection.

Instead of hardcoding 16 in the patfrag, check for natural alignment.
Also fix the autoupgrade and the various tests.

Also, use explicit -mattr instead of -mcpu: I stared at the output
several minutes wondering why I get 2x movntps for the unaligned
case (which is the ideal output, but needs some work: see FIXME),
until I remembered corei7-avx implies +slow-unaligned-mem-32.

llvm-svn: 246733

2015-09-02 23:25:39 +00:00

13 lines

312 B

LLVM

Raw Blame History

 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
 ; Test that we produce a movntdq, not a vmovntdq
 ; CHECK-NOT: vmovntdq
 define void @test(<2 x i64>* nocapture %a, <2 x i64> %b) nounwind optsize {
 entry:
   store <2 x i64> %b, <2 x i64>* %a, align 32, !nontemporal !0
   ret void
 }
 !0 = !{i32 1}