1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-21 12:02:58 +02:00
llvm-mirror/test/CodeGen/PowerPC/vec-abi-align.ll
Hal Finkel 49a12f79c1 [PowerPC] Make LDtocL and friends invariant loads
LDtocL, and other loads that roughly correspond to the TOC_ENTRY SDAG node,
represent loads from the TOC, which is invariant. As a result, these loads can
be hoisted out of loops, etc. In order to do this, we need to generate
GOT-style MMOs for TOC_ENTRY, which requires treating it as a legitimate memory
intrinsic node type. Once this is done, the MMO transfer is automatically
handled for TableGen-driven instruction selection, and for nodes generated
directly in PPCISelDAGToDAG, we need to transfer the MMOs manually.

Also, we were not transferring MMOs associated with pre-increment loads, so do
that too.

Lastly, this fixes an exposed bug where R30 was not added as a defined operand of
UpdateGBR.

This problem was highlighted by an example (used to generate the test case)
posted to llvmdev by Francois Pichet.

llvm-svn: 230553
2015-02-25 21:36:59 +00:00

80 lines
2.6 KiB
LLVM

; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 -mattr=-vsx < %s | FileCheck %s
; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 -mattr=+vsx < %s | FileCheck -check-prefix=CHECK-VSX %s
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-f128:128:128-v128:128:128-n32:64"
target triple = "powerpc64-unknown-linux-gnu"
%struct.s2 = type { i64, <4 x float> }
@ve = external global <4 x float>
@n = external global i64
; Function Attrs: nounwind
define void @test1(i64 %d1, i64 %d2, i64 %d3, i64 %d4, i64 %d5, i64 %d6, i64 %d7, i64 %d8, i64 %d9, <4 x float> inreg %vs.coerce) #0 {
entry:
store <4 x float> %vs.coerce, <4 x float>* @ve, align 16
ret void
; CHECK-LABEL: @test1
; CHECK: stvx 2,
; CHECK: blr
; CHECK-VSX-LABEL: @test1
; CHECK-VSX: stxvw4x 34,
; CHECK-VSX: blr
}
; Function Attrs: nounwind
define void @test2(i64 %d1, i64 %d2, i64 %d3, i64 %d4, i64 %d5, i64 %d6, i64 %d7, i64 %d8, %struct.s2* byval nocapture readonly %vs) #0 {
entry:
%m = getelementptr inbounds %struct.s2* %vs, i64 0, i32 0
%0 = load i64* %m, align 8
store i64 %0, i64* @n, align 8
%v = getelementptr inbounds %struct.s2* %vs, i64 0, i32 1
%1 = load <4 x float>* %v, align 16
store <4 x float> %1, <4 x float>* @ve, align 16
ret void
; CHECK-LABEL: @test2
; CHECK-DAG: ld {{[0-9]+}}, 112(1)
; CHECK-DAG: li [[REG16:[0-9]+]], 16
; CHECK-DAG: addi [[REGB:[0-9]+]], 1, 112
; CHECK-DAG: lvx 2, [[REGB]], [[REG16]]
; CHECK: blr
; CHECK-VSX-LABEL: @test2
; CHECK-VSX-DAG: ld {{[0-9]+}}, 112(1)
; CHECK-VSX-DAG: li [[REG16:[0-9]+]], 16
; CHECK-VSX-DAG: addi [[REGB:[0-9]+]], 1, 112
; CHECK-VSX-DAG: lxvw4x {{[0-9]+}}, [[REGB]], [[REG16]]
; CHECK-VSX: blr
}
; Function Attrs: nounwind
define void @test3(i64 %d1, i64 %d2, i64 %d3, i64 %d4, i64 %d5, i64 %d6, i64 %d7, i64 %d8, i64 %d9, %struct.s2* byval nocapture readonly %vs) #0 {
entry:
%m = getelementptr inbounds %struct.s2* %vs, i64 0, i32 0
%0 = load i64* %m, align 8
store i64 %0, i64* @n, align 8
%v = getelementptr inbounds %struct.s2* %vs, i64 0, i32 1
%1 = load <4 x float>* %v, align 16
store <4 x float> %1, <4 x float>* @ve, align 16
ret void
; CHECK-LABEL: @test3
; CHECK-DAG: ld {{[0-9]+}}, 128(1)
; CHECK-DAG: li [[REG16:[0-9]+]], 16
; CHECK-DAG: addi [[REGB:[0-9]+]], 1, 128
; CHECK-DAG: lvx 2, [[REGB]], [[REG16]]
; CHECK: blr
; CHECK-VSX-LABEL: @test3
; CHECK-VSX-DAG: ld {{[0-9]+}}, 128(1)
; CHECK-VSX-DAG: li [[REG16:[0-9]+]], 16
; CHECK-VSX-DAG: addi [[REGB:[0-9]+]], 1, 128
; CHECK-VSX-DAG: lxvw4x {{[0-9]+}}, [[REGB]], [[REG16]]
; CHECK-VSX: blr
}
attributes #0 = { nounwind }