Files
clang-p2996/llvm/test/CodeGen/VE/Vector/vp_reduce_add.ll
Fangrui Song 5240e0b891 [VE,test] Change llc -march= to -mtriple=
Similar to 806761a762

-mtriple= specifies the full target triple while -march= merely sets the
architecture part of the default target triple, leaving a target triple which
may not make sense.

Therefore, -march= is error-prone and not recommended for tests without a target
triple. The issue has been benign as we recognize $ve-apple-darwin as ELF instead
of rejecting it outrightly.
2024-12-15 10:24:14 -08:00

42 lines
1.5 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -O0 --mtriple=ve -mattr=+vpu | FileCheck %s
declare i64 @llvm.vp.reduce.add.v256i64(i64, <256 x i64>, <256 x i1>, i32)
define fastcc i64 @vp_reduce_add_v256i64(i64 %s, <256 x i64> %v, <256 x i1> %m, i32 %n) {
; CHECK-LABEL: vp_reduce_add_v256i64:
; CHECK: # %bb.0:
; CHECK-NEXT: and %s1, %s1, (32)0
; CHECK-NEXT: # kill: def $sw1 killed $sw1 killed $sx1
; CHECK-NEXT: lvl %s1
; CHECK-NEXT: vsum.l %v0, %v0, %vm1
; CHECK-NEXT: lvs %s1, %v0(0)
; CHECK-NEXT: adds.l %s0, %s0, %s1
; CHECK-NEXT: b.l.t (, %s10)
%r = call i64 @llvm.vp.reduce.add.v256i64(i64 %s, <256 x i64> %v, <256 x i1> %m, i32 %n)
ret i64 %r
}
declare i32 @llvm.vp.reduce.add.v256i32(i32, <256 x i32>, <256 x i1>, i32)
define fastcc i32 @vp_reduce_add_v256i32(i32 %s, <256 x i32> %v, <256 x i1> %m, i32 %n) {
; CHECK-LABEL: vp_reduce_add_v256i32:
; CHECK: # %bb.0:
; CHECK-NEXT: and %s0, %s0, (32)0
; CHECK-NEXT: # kill: def $sw0 killed $sw0 killed $sx0
; CHECK-NEXT: and %s1, %s1, (32)0
; CHECK-NEXT: # kill: def $sw1 killed $sw1 killed $sx1
; CHECK-NEXT: lvl %s1
; CHECK-NEXT: vsum.w.sx %v0, %v0, %vm1
; CHECK-NEXT: lvs %s1, %v0(0)
; CHECK-NEXT: or %s1, 0, %s1
; CHECK-NEXT: adds.w.sx %s1, %s0, %s1
; CHECK-NEXT: # implicit-def: $sx0
; CHECK-NEXT: or %s0, 0, %s1
; CHECK-NEXT: b.l.t (, %s10)
%r = call i32 @llvm.vp.reduce.add.v256i32(i32 %s, <256 x i32> %v, <256 x i1> %m, i32 %n)
ret i32 %r
}