369891b674
In most cases, the type information attached to load and store instructions is meaningless and inconsistently applied. We can usually use ".b" loads and avoid the complexity of trying to assign the correct type. The one expectation is sign-extending load, which will continue to use ".s" to ensure the sign extension into a larger register is done correctly.
28 lines
1.0 KiB
LLVM
28 lines
1.0 KiB
LLVM
; RUN: llc < %s -mtriple=nvptx -mcpu=sm_20 | FileCheck %s
|
|
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_20 | FileCheck %s
|
|
; RUN: %if ptxas && !ptxas-12.0 %{ llc < %s -mtriple=nvptx -mcpu=sm_20 | %ptxas-verify %}
|
|
; RUN: %if ptxas %{llc < %s -mtriple=nvptx64 -mcpu=sm_20 | %ptxas-verify %}
|
|
|
|
; This test makes sure that vector selects are scalarized by the type legalizer.
|
|
; If not, type legalization will fail.
|
|
|
|
; CHECK-LABEL: .visible .func foo(
|
|
define void @foo(ptr addrspace(1) %def_a, ptr addrspace(1) %def_b, ptr addrspace(1) %def_c) {
|
|
entry:
|
|
; CHECK: ld.global.v2.b32
|
|
; CHECK: ld.global.v2.b32
|
|
; CHECK: ld.global.v2.b32
|
|
%tmp4 = load <2 x i32>, ptr addrspace(1) %def_a
|
|
%tmp6 = load <2 x i32>, ptr addrspace(1) %def_c
|
|
%tmp8 = load <2 x i32>, ptr addrspace(1) %def_b
|
|
; CHECK: setp.gt.s32
|
|
; CHECK: setp.gt.s32
|
|
%0 = icmp sge <2 x i32> %tmp4, zeroinitializer
|
|
; CHECK: selp.b32
|
|
; CHECK: selp.b32
|
|
%cond = select <2 x i1> %0, <2 x i32> %tmp6, <2 x i32> %tmp8
|
|
; CHECK: st.global.v2.b32
|
|
store <2 x i32> %cond, ptr addrspace(1) %def_c
|
|
ret void
|
|
}
|