TheRock/compiler/amd-llvm/llvm/test/CodeGen/RISCV/rvv/combine-splats.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv32 -mattr=+m,+v < %s | FileCheck %s
; RUN: llc -mtriple=riscv64 -mattr=+m,+v < %s | FileCheck %s

; fold (and (or x, C), D) -> D if (C & D) == D

define <vscale x 4 x i32> @and_or_nxv4i32(<vscale x 4 x i32> %A) {
; CHECK-LABEL: and_or_nxv4i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT:    vmv.v.i v8, 8
; CHECK-NEXT:    ret
  %v1 = or <vscale x 4 x i32> %A, splat (i32 255)
  %v2 = and <vscale x 4 x i32> %v1, splat (i32 8)
  ret <vscale x 4 x i32> %v2
}

; (or (and X, c1), c2) -> (and (or X, c2), c1|c2) iff (c1 & c2) != 0

define <vscale x 2 x i64> @or_and_nxv2i64(<vscale x 2 x i64> %a0) {
; CHECK-LABEL: or_and_nxv2i64:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
; CHECK-NEXT:    vor.vi v8, v8, 3
; CHECK-NEXT:    vand.vi v8, v8, 7
; CHECK-NEXT:    ret
  %v1 = and <vscale x 2 x i64> %a0, splat (i64 7)
  %v2 = or <vscale x 2 x i64> %v1, splat (i64 3)
  ret <vscale x 2 x i64> %v2
}

; If all masked bits are going to be set, that's a constant fold.

define <vscale x 2 x i64> @or_and_nxv2i64_fold(<vscale x 2 x i64> %a0) {
; CHECK-LABEL: or_and_nxv2i64_fold:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
; CHECK-NEXT:    vmv.v.i v8, 3
; CHECK-NEXT:    ret
  %v1 = and <vscale x 2 x i64> %a0, splat (i64 1)
  %v2 = or <vscale x 2 x i64> %v1, splat (i64 3)
  ret <vscale x 2 x i64> %v2
}

; fold (shl (shl x, c1), c2) -> (shl x, (add c1, c2))

define <vscale x 4 x i32> @combine_vec_shl_shl(<vscale x 4 x i32> %x) {
; CHECK-LABEL: combine_vec_shl_shl:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT:    vsll.vi v8, v8, 6
; CHECK-NEXT:    ret
  %v1 = shl <vscale x 4 x i32> %x, splat (i32 2)
  %v2 = shl <vscale x 4 x i32> %v1, splat (i32 4)
  ret <vscale x 4 x i32> %v2
}

; fold (sra (sra x, c1), c2) -> (sra x, (add c1, c2))

define <vscale x 2 x i32> @combine_vec_ashr_ashr(<vscale x 2 x i32> %x) {
; CHECK-LABEL: combine_vec_ashr_ashr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
; CHECK-NEXT:    vsra.vi v8, v8, 6
; CHECK-NEXT:    ret
  %v1 = ashr <vscale x 2 x i32> %x, splat (i32 2)
  %v2 = ashr <vscale x 2 x i32> %v1, splat (i32 4)
  ret <vscale x 2 x i32> %v2
}

; fold (srl (srl x, c1), c2) -> (srl x, (add c1, c2))

define <vscale x 8 x i16> @combine_vec_lshr_lshr(<vscale x 8 x i16> %x) {
; CHECK-LABEL: combine_vec_lshr_lshr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
; CHECK-NEXT:    vsrl.vi v8, v8, 8
; CHECK-NEXT:    ret
  %v1 = lshr <vscale x 8 x i16> %x, splat (i16 4)
  %v2 = lshr <vscale x 8 x i16> %v1, splat (i16 4)
  ret <vscale x 8 x i16> %v2
}

; fold (fmul x, 1.0) -> x
define <vscale x 2 x float> @combine_fmul_one(<vscale x 2 x float> %x) {
; CHECK-LABEL: combine_fmul_one:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret
  %v = fmul <vscale x 2 x float> %x, splat (float 1.0)
  ret <vscale x 2 x float> %v
}

; fold (fmul 1.0, x) -> x
define <vscale x 2 x float> @combine_fmul_one_commuted(<vscale x 2 x float> %x) {
; CHECK-LABEL: combine_fmul_one_commuted:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret
  %v = fmul <vscale x 2 x float> splat (float 1.0), %x
  ret <vscale x 2 x float> %v
}