RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-31 07:52:55 +01:00
Code Issues Projects Releases Wiki Activity
b98a4accfa
llvm-mirror/test/CodeGen/ARM/2010-05-20-NEONSpillCrash.ll

46 lines
3.3 KiB
LLVM
Raw Normal View History

RegAlloc superpass: includes phi elimination, coalescing, and scheduling. Creates a configurable regalloc pipeline. Ensure specific llc options do what they say and nothing more: -reglloc=... has no effect other than selecting the allocator pass itself. This patch introduces a new umbrella flag, "-optimize-regalloc", to enable/disable the optimizing regalloc "superpass". This allows for example testing coalscing and scheduling under -O0 or vice-versa. When a CodeGen pass requires the MachineFunction to have a particular property, we need to explicitly define that property so it can be directly queried rather than naming a specific Pass. For example, to check for SSA, use MRI->isSSA, not addRequired<PHIElimination>. CodeGen transformation passes are never "required" as an analysis ProcessImplicitDefs does not require LiveVariables. We have a plan to massively simplify some of the early passes within the regalloc superpass. llvm-svn: 150226
2012-02-10 05:10:36 +01:00
; RUN: llc < %s -march=arm -mattr=+neon -O0 -optimize-regalloc -regalloc=basic
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
; This test would crash the rewriter when trying to handle a spill after one of
; the @llvm.arm.neon.vld3.v8i8 defined three parts of a register.
%struct.__neon_int8x8x3_t = type { <8 x i8>, <8 x i8>, <8 x i8> }
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
declare %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8*, i32) nounwind readonly
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
declare void @llvm.arm.neon.vst3.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, i32) nounwind
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
define <8 x i8> @t3(i8* %A1, i8* %A2, i8* %A3, i8* %A4, i8* %A5, i8* %A6, i8* %A7, i8* %A8, i8* %B) nounwind {
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1b = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A2, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=2]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2b = extractvalue %struct.__neon_int8x8x3_t %tmp1b, 0 ; <<8 x i8>> [#uses=1]
%tmp4b = extractvalue %struct.__neon_int8x8x3_t %tmp1b, 1 ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1d = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A4, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=2]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2d = extractvalue %struct.__neon_int8x8x3_t %tmp1d, 0 ; <<8 x i8>> [#uses=1]
%tmp4d = extractvalue %struct.__neon_int8x8x3_t %tmp1d, 1 ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1e = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A5, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=1]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2e = extractvalue %struct.__neon_int8x8x3_t %tmp1e, 0 ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1f = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A6, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=1]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2f = extractvalue %struct.__neon_int8x8x3_t %tmp1f, 0 ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1g = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A7, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=2]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2g = extractvalue %struct.__neon_int8x8x3_t %tmp1g, 0 ; <<8 x i8>> [#uses=1]
%tmp4g = extractvalue %struct.__neon_int8x8x3_t %tmp1g, 1 ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
%tmp1h = call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3.v8i8(i8* %A8, i32 1) ; <%struct.__neon_int8x8x3_t> [#uses=2]
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2h = extractvalue %struct.__neon_int8x8x3_t %tmp1h, 0 ; <<8 x i8>> [#uses=1]
%tmp3h = extractvalue %struct.__neon_int8x8x3_t %tmp1h, 2 ; <<8 x i8>> [#uses=1]
%tmp2bd = add <8 x i8> %tmp2b, %tmp2d ; <<8 x i8>> [#uses=1]
%tmp4bd = add <8 x i8> %tmp4b, %tmp4d ; <<8 x i8>> [#uses=1]
%tmp2abcd = mul <8 x i8> undef, %tmp2bd ; <<8 x i8>> [#uses=1]
%tmp4abcd = mul <8 x i8> undef, %tmp4bd ; <<8 x i8>> [#uses=2]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
call void @llvm.arm.neon.vst3.v8i8(i8* %A1, <8 x i8> %tmp4abcd, <8 x i8> zeroinitializer, <8 x i8> %tmp2abcd, i32 1)
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp2ef = sub <8 x i8> %tmp2e, %tmp2f ; <<8 x i8>> [#uses=1]
%tmp2gh = sub <8 x i8> %tmp2g, %tmp2h ; <<8 x i8>> [#uses=1]
%tmp3gh = sub <8 x i8> zeroinitializer, %tmp3h ; <<8 x i8>> [#uses=1]
%tmp4ef = sub <8 x i8> zeroinitializer, %tmp4g ; <<8 x i8>> [#uses=1]
%tmp2efgh = mul <8 x i8> %tmp2ef, %tmp2gh ; <<8 x i8>> [#uses=1]
%tmp3efgh = mul <8 x i8> undef, %tmp3gh ; <<8 x i8>> [#uses=1]
%tmp4efgh = mul <8 x i8> %tmp4ef, undef ; <<8 x i8>> [#uses=2]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
call void @llvm.arm.neon.vst3.v8i8(i8* %A2, <8 x i8> %tmp4efgh, <8 x i8> %tmp3efgh, <8 x i8> %tmp2efgh, i32 1)
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
%tmp4 = sub <8 x i8> %tmp4efgh, %tmp4abcd ; <<8 x i8>> [#uses=1]
Add alignment arguments to all the NEON load/store intrinsics. Update all the tests using those intrinsics and add support for auto-upgrading bitcode files with the old versions of the intrinsics. llvm-svn: 112271
2010-08-27 19:13:24 +02:00
tail call void @llvm.arm.neon.vst3.v8i8(i8* %B, <8 x i8> zeroinitializer, <8 x i8> undef, <8 x i8> undef, i32 1)
Teach VirtRegRewriter to handle spilling in instructions that have multiple definitions of the virtual register. This happens when spilling the registers produced by REG_SEQUENCE: %reg1047:5<def>, %reg1047:6<def>, %reg1047:7<def> = VLD3d8 %reg1033, 0, pred:14, pred:%reg0 The rewriter would spill the register multiple times, dead store elimination tried to keep up, but ended up cutting the branch it was sitting on. llvm-svn: 104321
2010-05-21 18:36:13 +02:00
ret <8 x i8> %tmp4
}
Reference in New Issue Copy Permalink