maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/NVPTX/addrspacecast-gvar.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

14 lines
726 B
LLVM
Raw Normal View History

[NVPTX] Add support for addrspacecast in global variable initializers, including emitting generic() when casting to address space 0. llvm-svn: 205906
2014-04-09 23:39:11 +08:00
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
; CHECK: .visible .global .align 4 .u32 g = 42;
; CHECK: .visible .global .align 4 .u32 g2 = generic(g);
; CHECK: .visible .global .align 4 .u32 g3 = g;
[NVPTX] Emits "generic()" depending on the original address space Summary: Fixes a bug in the NVPTX codegen. The code used to miss necessary "generic()" on aggregates of addrspacecasts. Test Plan: addrspacecast-gvar.ll Reviewers: eliben, jholewinski Reviewed By: jholewinski Subscribers: jholewinski, llvm-commits Differential Revision: http://reviews.llvm.org/D9130 llvm-svn: 235689
2015-04-24 10:57:30 +08:00
; CHECK: .visible .global .align 8 .u32 g4[2] = {0, generic(g)};
[NVPTX] Handle addrspacecast constant expressions in aggregate initializers We need to track if an AddrSpaceCast expression was seen when generating an MCExpr for a ConstantExpr. This change introduces a custom lowerConstant method to the NVPTX asm printer that will create NVPTXGenericMCSymbolRefExpr nodes at the appropriate places to encode the information that a given symbol needs to be casted to a generic address. llvm-svn: 236000
2015-04-29 01:18:30 +08:00
; CHECK: .visible .global .align 8 .u32 g5[2] = {0, generic(g)+8};
[NVPTX] Add support for addrspacecast in global variable initializers, including emitting generic() when casting to address space 0. llvm-svn: 205906
2014-04-09 23:39:11 +08:00
@g = addrspace(1) global i32 42
@g2 = addrspace(1) global i32* addrspacecast (i32 addrspace(1)* @g to i32*)
@g3 = addrspace(1) global i32 addrspace(1)* @g
[NVPTX] Emits "generic()" depending on the original address space Summary: Fixes a bug in the NVPTX codegen. The code used to miss necessary "generic()" on aggregates of addrspacecasts. Test Plan: addrspacecast-gvar.ll Reviewers: eliben, jholewinski Reviewed By: jholewinski Subscribers: jholewinski, llvm-commits Differential Revision: http://reviews.llvm.org/D9130 llvm-svn: 235689
2015-04-24 10:57:30 +08:00
@g4 = constant {i32*, i32*} {i32* null, i32* addrspacecast (i32 addrspace(1)* @g to i32*)}
[NVPTX] Handle addrspacecast constant expressions in aggregate initializers We need to track if an AddrSpaceCast expression was seen when generating an MCExpr for a ConstantExpr. This change introduces a custom lowerConstant method to the NVPTX asm printer that will create NVPTXGenericMCSymbolRefExpr nodes at the appropriate places to encode the information that a given symbol needs to be casted to a generic address. llvm-svn: 236000
2015-04-29 01:18:30 +08:00
@g5 = constant {i32*, i32*} {i32* null, i32* addrspacecast (i32 addrspace(1)* getelementptr (i32, i32 addrspace(1)* @g, i32 2) to i32*)}