1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 04:32:44 +01:00
llvm-mirror/test/CodeGen/Hexagon/v60Intrins.ll
serge-sans-paille 73bc91a5e6 Revert "[NFC] remove explicit default value for strboolattr attribute in tests"
This reverts commit bda6e5bee04c75b1f1332b4fd1ac4e8ef6c3c247.

See https://lab.llvm.org/buildbot/#/builders/109/builds/15424 for instance
2021-05-24 19:43:40 +02:00

2559 lines
193 KiB
LLVM

; RUN: llc -march=hexagon -mcpu=hexagonv60 -O2 -disable-post-ra < %s | FileCheck %s
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.eq(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.eq(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.eq(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.eq(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.eq(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.eq(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.eq(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.eq(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.eq(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.eq(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.eq(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.eq(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.uw,v{{[0-9]*}}.uw)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vcmp.gt(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.uw,v{{[0-9]*}}.uw)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} &= vcmp.gt(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.uw,v{{[0-9]*}}.uw)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} |= vcmp.gt(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.uw,v{{[0-9]*}}.uw)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} ^= vcmp.gt(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = xor{{[0-9]*}}(q{{[0-3]}},q{{[0-3]}})
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = v
; CHECK: v{{[0-9]*}} = valign(v{{[0-9]*}},v{{[0-9]*}},#1)
; CHECK: v{{[0-9]*}} = valign(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vand(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} |= vand(q{{[0-3]}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vdelta(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vlalign(v{{[0-9]*}},v{{[0-9]*}},#1)
; CHECK: v{{[0-9]*}} = vlalign(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vmux(q{{[0-3]}},v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vnot(v{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vor{{[0-9]*}}(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vr{{[0-9]*}}delta(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vr{{[0-9]*}}or{{[0-9]*}}(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}} = vxor{{[0-9]*}}(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}}.b = vadd(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vasr{{[0-9]*}}(v{{[0-9]*}}.h,v{{[0-9]*}}.h,r{{[0-9]*}}):{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.b = vdeal(v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vdeale(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vlut32(v{{[0-9]*}}.b,v{{[0-9]*}}.b,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.b |= vlut32(v{{[0-9]*}}.b,v{{[0-9]*}}.b,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.b = vnav{{[0-9]*}}g(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.b = vpack(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.b = vpacke(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.b = vpacko(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.b = vr{{[0-9]*}}ound(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.b = vshuff(v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vshuffe(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vshuffo(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.b = vsub(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.h = vabs(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vabs(v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vadd(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vadd(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vasl(v{{[0-9]*}}.h,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.h = vasl(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vasr{{[0-9]*}}(v{{[0-9]*}}.h,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.h = vasr{{[0-9]*}}(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vasr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.h = vasr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w,r{{[0-9]*}}):{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vasr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w,r{{[0-9]*}}):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vav{{[0-9]*}}g(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vav{{[0-9]*}}g(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}r{{[0-9]*}}nd
; CHECK: v{{[0-9]*}}.h = vdeal(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vdmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.h += vdmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.h = vlsr{{[0-9]*}}(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vmax(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vmin(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vmpyi(v{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.h = vmpyi(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h += vmpyi(v{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.h += vmpyi(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vnav{{[0-9]*}}g(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vnor{{[0-9]*}}mamt(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vpack(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vpacke(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.h = vpacko(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.h = vpopcount(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vr{{[0-9]*}}ound(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.h = vsat(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.h = vshuff(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vshuffe(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vshuffo(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vsub(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.h = vsub(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vabsdiff(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.ub = vadd(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vasr{{[0-9]*}}(v{{[0-9]*}}.h,v{{[0-9]*}}.h,r{{[0-9]*}}):{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vasr{{[0-9]*}}(v{{[0-9]*}}.h,v{{[0-9]*}}.h,r{{[0-9]*}}):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vav{{[0-9]*}}g(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.ub = vav{{[0-9]*}}g(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub):{{[0-9]*}}r{{[0-9]*}}nd
; CHECK: v{{[0-9]*}}.ub = vmax(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.ub = vmin(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.ub = vpack(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vr{{[0-9]*}}ound(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.ub = vsat(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.ub = vsub(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uh = vabsdiff(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.uh = vabsdiff(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.uh = vadd(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uh = vasr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w,r{{[0-9]*}}):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uh = vav{{[0-9]*}}g(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.uh = vav{{[0-9]*}}g(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh):{{[0-9]*}}r{{[0-9]*}}nd
; CHECK: v{{[0-9]*}}.uh = vcl0(v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.uh = vlsr{{[0-9]*}}(v{{[0-9]*}}.uh,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.uh = vmax(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.uh = vmin(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.uh = vpack(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uh = vr{{[0-9]*}}ound(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uh = vsub(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.uw = vabsdiff(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.uw = vcl0(v{{[0-9]*}}.uw)
; CHECK: v{{[0-9]*}}.uw = vlsr{{[0-9]*}}(v{{[0-9]*}}.uw,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.uw = vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.uw = vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.uw += vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.uw += vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}.w = vabs(v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vabs(v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vadd(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vadd(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vasl(v{{[0-9]*}}.w,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vasl(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w += vasl(v{{[0-9]*}}.w,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vasr{{[0-9]*}}(v{{[0-9]*}}.w,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vasr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w += vasr{{[0-9]*}}(v{{[0-9]*}}.w,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vav{{[0-9]*}}g(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vav{{[0-9]*}}g(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}r{{[0-9]*}}nd
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.uh,#1):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.uh,#1):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vinser{{[0-9]*}}t(r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vinser{{[0-9]*}}t(r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vinser{{[0-9]*}}t(r{{[0-9]*}})
; CHECK: v{{[0-9]*}}.w = vlsr{{[0-9]*}}(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vmax(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vmin(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vmpye(v{{[0-9]*}}.w,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.w = vmpyi(v{{[0-9]*}}.w,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w = vmpyi(v{{[0-9]*}}.w,r{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.w += vmpyi(v{{[0-9]*}}.w,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w += vmpyi(v{{[0-9]*}}.w,r{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.w = vmpyie(v{{[0-9]*}}.w,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.w += vmpyie(v{{[0-9]*}}.w,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.w += vmpyie(v{{[0-9]*}}.w,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}.w = vmpyieo(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.w = vmpyio(v{{[0-9]*}}.w,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}.w = vmpyo(v{{[0-9]*}}.w,v{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w = vmpyo(v{{[0-9]*}}.w,v{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}.w += vmpyo(v{{[0-9]*}}.w,v{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}r{{[0-9]*}}nd:{{[0-9]*}}sat:{{[0-9]*}}shift
; CHECK: v{{[0-9]*}}.w += vmpyo(v{{[0-9]*}}.w,v{{[0-9]*}}.h):{{[0-9]*}}<<1:{{[0-9]*}}sat:{{[0-9]*}}shift
; CHECK: v{{[0-9]*}}.w = vnav{{[0-9]*}}g(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vnor{{[0-9]*}}mamt(v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vr{{[0-9]*}}mpy(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w = vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w = vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w += vr{{[0-9]*}}mpy(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w += vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w += vr{{[0-9]*}}mpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}.w = vsub(v{{[0-9]*}}.w,v{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}.w = vsub(v{{[0-9]*}}.w,v{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vcombine(v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vdeal(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vshuff(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vshuff(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vshuff(v{{[0-9]*}},v{{[0-9]*}},r{{[0-9]*}})
; CHECK: q{{[0-3]}} = vand(v{{[0-9]*}},r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}} = vswap(q{{[0-3]}},v{{[0-9]*}},v{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.b = vadd(v{{[0-9]*}}:{{[0-9]*}}.b,v{{[0-9]*}}:{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.b = vshuffoe(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.b = vsub(v{{[0-9]*}}:{{[0-9]*}}.b,v{{[0-9]*}}:{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vadd(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vadd(v{{[0-9]*}}:{{[0-9]*}}.h,v{{[0-9]*}}:{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vadd(v{{[0-9]*}}:{{[0-9]*}}.h,v{{[0-9]*}}:{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vdmpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vdmpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vlut16(v{{[0-9]*}}.b,v{{[0-9]*}}.h,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h |= vlut16(v{{[0-9]*}}.b,v{{[0-9]*}}.h,r{{[0-9]*}})
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpa(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpa(v{{[0-9]*}}:{{[0-9]*}}.ub,v{{[0-9]*}}:{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpa(v{{[0-9]*}}:{{[0-9]*}}.ub,v{{[0-9]*}}:{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vmpa(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vmpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vmpy(v{{[0-9]*}}.b,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vmpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vshuffoe(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vsub(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vsub(v{{[0-9]*}}:{{[0-9]*}}.h,v{{[0-9]*}}:{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vsub(v{{[0-9]*}}:{{[0-9]*}}.h,v{{[0-9]*}}:{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vsxt(v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vtmpy(v{{[0-9]*}}:{{[0-9]*}}.b,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vtmpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vtmpy(v{{[0-9]*}}:{{[0-9]*}}.b,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h += vtmpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h = vunpack(v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.h |= vunpacko(v{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.ub = vadd(v{{[0-9]*}}:{{[0-9]*}}.ub,v{{[0-9]*}}:{{[0-9]*}}.ub):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.ub = vsub(v{{[0-9]*}}:{{[0-9]*}}.ub,v{{[0-9]*}}:{{[0-9]*}}.ub):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vadd(v{{[0-9]*}}:{{[0-9]*}}.uh,v{{[0-9]*}}:{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vmpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh += vmpy(v{{[0-9]*}}.ub,r{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh += vmpy(v{{[0-9]*}}.ub,v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vsub(v{{[0-9]*}}:{{[0-9]*}}.uh,v{{[0-9]*}}:{{[0-9]*}}.uh):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vunpack(v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uh = vzxt(v{{[0-9]*}}.ub)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vdsad(v{{[0-9]*}}:{{[0-9]*}}.uh,r{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw += vdsad(v{{[0-9]*}}:{{[0-9]*}}.uh,r{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vmpy(v{{[0-9]*}}.uh,r{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vmpy(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw += vmpy(v{{[0-9]*}}.uh,r{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw += vmpy(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vr{{[0-9]*}}mpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.ub,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw += vr{{[0-9]*}}mpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.ub,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vr{{[0-9]*}}sad(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.ub,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw += vr{{[0-9]*}}sad(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.ub,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vunpack(v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.uw = vzxt(v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vadd(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vadd(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vadd(v{{[0-9]*}}:{{[0-9]*}}.w,v{{[0-9]*}}:{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vadd(v{{[0-9]*}}:{{[0-9]*}}.w,v{{[0-9]*}}:{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vdmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vmpa(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vmpa(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vmpy(v{{[0-9]*}}.h,r{{[0-9]*}}.h):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vmpy(v{{[0-9]*}}.h,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vr{{[0-9]*}}mpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vr{{[0-9]*}}mpy(v{{[0-9]*}}:{{[0-9]*}}.ub,r{{[0-9]*}}.b,#0)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vsub(v{{[0-9]*}}.h,v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vsub(v{{[0-9]*}}.uh,v{{[0-9]*}}.uh)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vsub(v{{[0-9]*}}:{{[0-9]*}}.w,v{{[0-9]*}}:{{[0-9]*}}.w)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vsub(v{{[0-9]*}}:{{[0-9]*}}.w,v{{[0-9]*}}:{{[0-9]*}}.w):{{[0-9]*}}sat
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vsxt(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vtmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w += vtmpy(v{{[0-9]*}}:{{[0-9]*}}.h,r{{[0-9]*}}.b)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w = vunpack(v{{[0-9]*}}.h)
; CHECK: v{{[0-9]*}}:{{[0-9]*}}.w |= vunpacko(v{{[0-9]*}}.h)
target datalayout = "e-m:e-p:32:32:32-i64:64:64-i32:32:32-i16:16:16-i1:8:8-f64:64:64-f32:32:32-v64:64:64-v32:32:32-a:0-n16:32"
target triple = "hexagon"
@K = global i64 0, align 8
@src = global i8 -1, align 1
@vecpreds = common global [15 x <16 x i32>] zeroinitializer, align 64
@Q6VecPredResult = common global <16 x i32> zeroinitializer, align 64
@vectors = common global [15 x <16 x i32>] zeroinitializer, align 64
@VectorResult = common global <16 x i32> zeroinitializer, align 64
@vector_pairs = common global [15 x <32 x i32>] zeroinitializer, align 128
@VectorPairResult = common global <32 x i32> zeroinitializer, align 128
@dst_addresses = common global [15 x i8] zeroinitializer, align 8
@ptr_addresses = common global [15 x i8*] zeroinitializer, align 8
@src_addresses = common global [15 x i8*] zeroinitializer, align 8
@dst = common global i8 0, align 1
@ptr = common global [32768 x i8] zeroinitializer, align 8
; Function Attrs: nounwind
define i32 @main() #0 {
entry:
%retval = alloca i32, align 4
store i32 0, i32* %retval, align 4
%0 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%1 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %0, i32 -1)
%2 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 1), align 64
%3 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %2, i32 -1)
%4 = call <64 x i1> @llvm.hexagon.V6.pred.and(<64 x i1> %1, <64 x i1> %3)
%5 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %4, i32 -1)
store volatile <16 x i32> %5, <16 x i32>* @Q6VecPredResult, align 64
%6 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%7 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %6, i32 -1)
%8 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 1), align 64
%9 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %8, i32 -1)
%10 = call <64 x i1> @llvm.hexagon.V6.pred.and.n(<64 x i1> %7, <64 x i1> %9)
%11 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %10, i32 -1)
store volatile <16 x i32> %11, <16 x i32>* @Q6VecPredResult, align 64
%12 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%13 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %12, i32 -1)
%14 = call <64 x i1> @llvm.hexagon.V6.pred.not(<64 x i1> %13)
%15 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %14, i32 -1)
store volatile <16 x i32> %15, <16 x i32>* @Q6VecPredResult, align 64
%16 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%17 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %16, i32 -1)
%18 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 1), align 64
%19 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %18, i32 -1)
%20 = call <64 x i1> @llvm.hexagon.V6.pred.or(<64 x i1> %17, <64 x i1> %19)
%21 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %20, i32 -1)
store volatile <16 x i32> %21, <16 x i32>* @Q6VecPredResult, align 64
%22 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%23 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %22, i32 -1)
%24 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 1), align 64
%25 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %24, i32 -1)
%26 = call <64 x i1> @llvm.hexagon.V6.pred.or.n(<64 x i1> %23, <64 x i1> %25)
%27 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %26, i32 -1)
store volatile <16 x i32> %27, <16 x i32>* @Q6VecPredResult, align 64
%28 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%29 = call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %28, i32 -1)
%30 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %29, i32 -1)
store volatile <16 x i32> %30, <16 x i32>* @Q6VecPredResult, align 64
%31 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%32 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %31, i32 -1)
%33 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%34 = call <64 x i1> @llvm.hexagon.V6.vandvrt.acc(<64 x i1> %32, <16 x i32> %33, i32 -1)
%35 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %34, i32 -1)
store volatile <16 x i32> %35, <16 x i32>* @Q6VecPredResult, align 64
%36 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%37 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%38 = call <64 x i1> @llvm.hexagon.V6.veqb(<16 x i32> %36, <16 x i32> %37)
%39 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %38, i32 -1)
store volatile <16 x i32> %39, <16 x i32>* @Q6VecPredResult, align 64
%40 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%41 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%42 = call <64 x i1> @llvm.hexagon.V6.veqh(<16 x i32> %40, <16 x i32> %41)
%43 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %42, i32 -1)
store volatile <16 x i32> %43, <16 x i32>* @Q6VecPredResult, align 64
%44 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%45 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%46 = call <64 x i1> @llvm.hexagon.V6.veqw(<16 x i32> %44, <16 x i32> %45)
%47 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %46, i32 -1)
store volatile <16 x i32> %47, <16 x i32>* @Q6VecPredResult, align 64
%48 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%49 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %48, i32 -1)
%50 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%51 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%52 = call <64 x i1> @llvm.hexagon.V6.veqb.and(<64 x i1> %49, <16 x i32> %50, <16 x i32> %51)
%53 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %52, i32 -1)
store volatile <16 x i32> %53, <16 x i32>* @Q6VecPredResult, align 64
%54 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%55 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %54, i32 -1)
%56 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%57 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%58 = call <64 x i1> @llvm.hexagon.V6.veqh.and(<64 x i1> %55, <16 x i32> %56, <16 x i32> %57)
%59 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %58, i32 -1)
store volatile <16 x i32> %59, <16 x i32>* @Q6VecPredResult, align 64
%60 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%61 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %60, i32 -1)
%62 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%63 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%64 = call <64 x i1> @llvm.hexagon.V6.veqw.and(<64 x i1> %61, <16 x i32> %62, <16 x i32> %63)
%65 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %64, i32 -1)
store volatile <16 x i32> %65, <16 x i32>* @Q6VecPredResult, align 64
%66 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%67 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %66, i32 -1)
%68 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%69 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%70 = call <64 x i1> @llvm.hexagon.V6.veqb.or(<64 x i1> %67, <16 x i32> %68, <16 x i32> %69)
%71 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %70, i32 -1)
store volatile <16 x i32> %71, <16 x i32>* @Q6VecPredResult, align 64
%72 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%73 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %72, i32 -1)
%74 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%75 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%76 = call <64 x i1> @llvm.hexagon.V6.veqh.or(<64 x i1> %73, <16 x i32> %74, <16 x i32> %75)
%77 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %76, i32 -1)
store volatile <16 x i32> %77, <16 x i32>* @Q6VecPredResult, align 64
%78 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%79 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %78, i32 -1)
%80 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%81 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%82 = call <64 x i1> @llvm.hexagon.V6.veqw.or(<64 x i1> %79, <16 x i32> %80, <16 x i32> %81)
%83 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %82, i32 -1)
store volatile <16 x i32> %83, <16 x i32>* @Q6VecPredResult, align 64
%84 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%85 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %84, i32 -1)
%86 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%87 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%88 = call <64 x i1> @llvm.hexagon.V6.veqb.xor(<64 x i1> %85, <16 x i32> %86, <16 x i32> %87)
%89 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %88, i32 -1)
store volatile <16 x i32> %89, <16 x i32>* @Q6VecPredResult, align 64
%90 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%91 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %90, i32 -1)
%92 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%93 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%94 = call <64 x i1> @llvm.hexagon.V6.veqh.xor(<64 x i1> %91, <16 x i32> %92, <16 x i32> %93)
%95 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %94, i32 -1)
store volatile <16 x i32> %95, <16 x i32>* @Q6VecPredResult, align 64
%96 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%97 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %96, i32 -1)
%98 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%99 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%100 = call <64 x i1> @llvm.hexagon.V6.veqw.xor(<64 x i1> %97, <16 x i32> %98, <16 x i32> %99)
%101 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %100, i32 -1)
store volatile <16 x i32> %101, <16 x i32>* @Q6VecPredResult, align 64
%102 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%103 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%104 = call <64 x i1> @llvm.hexagon.V6.vgtb(<16 x i32> %102, <16 x i32> %103)
%105 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %104, i32 -1)
store volatile <16 x i32> %105, <16 x i32>* @Q6VecPredResult, align 64
%106 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%107 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%108 = call <64 x i1> @llvm.hexagon.V6.vgth(<16 x i32> %106, <16 x i32> %107)
%109 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %108, i32 -1)
store volatile <16 x i32> %109, <16 x i32>* @Q6VecPredResult, align 64
%110 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%111 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%112 = call <64 x i1> @llvm.hexagon.V6.vgtub(<16 x i32> %110, <16 x i32> %111)
%113 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %112, i32 -1)
store volatile <16 x i32> %113, <16 x i32>* @Q6VecPredResult, align 64
%114 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%115 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%116 = call <64 x i1> @llvm.hexagon.V6.vgtuh(<16 x i32> %114, <16 x i32> %115)
%117 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %116, i32 -1)
store volatile <16 x i32> %117, <16 x i32>* @Q6VecPredResult, align 64
%118 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%119 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%120 = call <64 x i1> @llvm.hexagon.V6.vgtuw(<16 x i32> %118, <16 x i32> %119)
%121 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %120, i32 -1)
store volatile <16 x i32> %121, <16 x i32>* @Q6VecPredResult, align 64
%122 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%123 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%124 = call <64 x i1> @llvm.hexagon.V6.vgtw(<16 x i32> %122, <16 x i32> %123)
%125 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %124, i32 -1)
store volatile <16 x i32> %125, <16 x i32>* @Q6VecPredResult, align 64
%126 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%127 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %126, i32 -1)
%128 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%129 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%130 = call <64 x i1> @llvm.hexagon.V6.vgtb.and(<64 x i1> %127, <16 x i32> %128, <16 x i32> %129)
%131 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %130, i32 -1)
store volatile <16 x i32> %131, <16 x i32>* @Q6VecPredResult, align 64
%132 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%133 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %132, i32 -1)
%134 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%135 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%136 = call <64 x i1> @llvm.hexagon.V6.vgth.and(<64 x i1> %133, <16 x i32> %134, <16 x i32> %135)
%137 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %136, i32 -1)
store volatile <16 x i32> %137, <16 x i32>* @Q6VecPredResult, align 64
%138 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%139 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %138, i32 -1)
%140 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%141 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%142 = call <64 x i1> @llvm.hexagon.V6.vgtub.and(<64 x i1> %139, <16 x i32> %140, <16 x i32> %141)
%143 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %142, i32 -1)
store volatile <16 x i32> %143, <16 x i32>* @Q6VecPredResult, align 64
%144 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%145 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %144, i32 -1)
%146 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%147 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%148 = call <64 x i1> @llvm.hexagon.V6.vgtuh.and(<64 x i1> %145, <16 x i32> %146, <16 x i32> %147)
%149 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %148, i32 -1)
store volatile <16 x i32> %149, <16 x i32>* @Q6VecPredResult, align 64
%150 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%151 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %150, i32 -1)
%152 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%153 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%154 = call <64 x i1> @llvm.hexagon.V6.vgtuw.and(<64 x i1> %151, <16 x i32> %152, <16 x i32> %153)
%155 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %154, i32 -1)
store volatile <16 x i32> %155, <16 x i32>* @Q6VecPredResult, align 64
%156 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%157 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %156, i32 -1)
%158 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%159 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%160 = call <64 x i1> @llvm.hexagon.V6.vgtw.and(<64 x i1> %157, <16 x i32> %158, <16 x i32> %159)
%161 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %160, i32 -1)
store volatile <16 x i32> %161, <16 x i32>* @Q6VecPredResult, align 64
%162 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%163 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %162, i32 -1)
%164 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%165 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%166 = call <64 x i1> @llvm.hexagon.V6.vgtb.or(<64 x i1> %163, <16 x i32> %164, <16 x i32> %165)
%167 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %166, i32 -1)
store volatile <16 x i32> %167, <16 x i32>* @Q6VecPredResult, align 64
%168 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%169 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %168, i32 -1)
%170 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%171 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%172 = call <64 x i1> @llvm.hexagon.V6.vgth.or(<64 x i1> %169, <16 x i32> %170, <16 x i32> %171)
%173 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %172, i32 -1)
store volatile <16 x i32> %173, <16 x i32>* @Q6VecPredResult, align 64
%174 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%175 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %174, i32 -1)
%176 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%177 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%178 = call <64 x i1> @llvm.hexagon.V6.vgtub.or(<64 x i1> %175, <16 x i32> %176, <16 x i32> %177)
%179 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %178, i32 -1)
store volatile <16 x i32> %179, <16 x i32>* @Q6VecPredResult, align 64
%180 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%181 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %180, i32 -1)
%182 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%183 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%184 = call <64 x i1> @llvm.hexagon.V6.vgtuh.or(<64 x i1> %181, <16 x i32> %182, <16 x i32> %183)
%185 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %184, i32 -1)
store volatile <16 x i32> %185, <16 x i32>* @Q6VecPredResult, align 64
%186 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%187 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %186, i32 -1)
%188 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%189 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%190 = call <64 x i1> @llvm.hexagon.V6.vgtuw.or(<64 x i1> %187, <16 x i32> %188, <16 x i32> %189)
%191 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %190, i32 -1)
store volatile <16 x i32> %191, <16 x i32>* @Q6VecPredResult, align 64
%192 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%193 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %192, i32 -1)
%194 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%195 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%196 = call <64 x i1> @llvm.hexagon.V6.vgtw.or(<64 x i1> %193, <16 x i32> %194, <16 x i32> %195)
%197 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %196, i32 -1)
store volatile <16 x i32> %197, <16 x i32>* @Q6VecPredResult, align 64
%198 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%199 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %198, i32 -1)
%200 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%201 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%202 = call <64 x i1> @llvm.hexagon.V6.vgtb.xor(<64 x i1> %199, <16 x i32> %200, <16 x i32> %201)
%203 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %202, i32 -1)
store volatile <16 x i32> %203, <16 x i32>* @Q6VecPredResult, align 64
%204 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%205 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %204, i32 -1)
%206 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%207 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%208 = call <64 x i1> @llvm.hexagon.V6.vgth.xor(<64 x i1> %205, <16 x i32> %206, <16 x i32> %207)
%209 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %208, i32 -1)
store volatile <16 x i32> %209, <16 x i32>* @Q6VecPredResult, align 64
%210 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%211 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %210, i32 -1)
%212 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%213 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%214 = call <64 x i1> @llvm.hexagon.V6.vgtub.xor(<64 x i1> %211, <16 x i32> %212, <16 x i32> %213)
%215 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %214, i32 -1)
store volatile <16 x i32> %215, <16 x i32>* @Q6VecPredResult, align 64
%216 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%217 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %216, i32 -1)
%218 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%219 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%220 = call <64 x i1> @llvm.hexagon.V6.vgtuh.xor(<64 x i1> %217, <16 x i32> %218, <16 x i32> %219)
%221 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %220, i32 -1)
store volatile <16 x i32> %221, <16 x i32>* @Q6VecPredResult, align 64
%222 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%223 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %222, i32 -1)
%224 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%225 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%226 = call <64 x i1> @llvm.hexagon.V6.vgtuw.xor(<64 x i1> %223, <16 x i32> %224, <16 x i32> %225)
%227 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %226, i32 -1)
store volatile <16 x i32> %227, <16 x i32>* @Q6VecPredResult, align 64
%228 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%229 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %228, i32 -1)
%230 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%231 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%232 = call <64 x i1> @llvm.hexagon.V6.vgtw.xor(<64 x i1> %229, <16 x i32> %230, <16 x i32> %231)
%233 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %232, i32 -1)
store volatile <16 x i32> %233, <16 x i32>* @Q6VecPredResult, align 64
%234 = call <64 x i1> @llvm.hexagon.V6.pred.scalar2(i32 1)
%235 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %234, i32 -1)
store volatile <16 x i32> %235, <16 x i32>* @Q6VecPredResult, align 64
%236 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%237 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %236, i32 -1)
%238 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 1), align 64
%239 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %238, i32 -1)
%240 = call <64 x i1> @llvm.hexagon.V6.pred.xor(<64 x i1> %237, <64 x i1> %239)
%241 = tail call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %240, i32 -1)
store volatile <16 x i32> %241, <16 x i32>* @Q6VecPredResult, align 64
%242 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%243 = call <16 x i32> @llvm.hexagon.V6.vassign(<16 x i32> %242)
store volatile <16 x i32> %243, <16 x i32>* @VectorResult, align 64
%244 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%245 = call <16 x i32> @llvm.hexagon.V6.hi(<32 x i32> %244)
store volatile <16 x i32> %245, <16 x i32>* @VectorResult, align 64
%246 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%247 = call <16 x i32> @llvm.hexagon.V6.lo(<32 x i32> %246)
store volatile <16 x i32> %247, <16 x i32>* @VectorResult, align 64
%248 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%249 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%250 = call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> %248, <16 x i32> %249, i32 1)
store volatile <16 x i32> %250, <16 x i32>* @VectorResult, align 64
%251 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%252 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%253 = call <16 x i32> @llvm.hexagon.V6.valignb(<16 x i32> %251, <16 x i32> %252, i32 -1)
store volatile <16 x i32> %253, <16 x i32>* @VectorResult, align 64
%254 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%255 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %254, i32 -1)
%256 = call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %255, i32 -1)
store volatile <16 x i32> %256, <16 x i32>* @VectorResult, align 64
%257 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%258 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%259 = call <16 x i32> @llvm.hexagon.V6.vand(<16 x i32> %257, <16 x i32> %258)
store volatile <16 x i32> %259, <16 x i32>* @VectorResult, align 64
%260 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%261 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%262 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %261, i32 -1)
%263 = call <16 x i32> @llvm.hexagon.V6.vandqrt.acc(<16 x i32> %260, <64 x i1> %262, i32 -1)
store volatile <16 x i32> %263, <16 x i32>* @VectorResult, align 64
%264 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%265 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%266 = call <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32> %264, <16 x i32> %265)
store volatile <16 x i32> %266, <16 x i32>* @VectorResult, align 64
%267 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%268 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%269 = call <16 x i32> @llvm.hexagon.V6.vlalignbi(<16 x i32> %267, <16 x i32> %268, i32 1)
store volatile <16 x i32> %269, <16 x i32>* @VectorResult, align 64
%270 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%271 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%272 = call <16 x i32> @llvm.hexagon.V6.vlalignb(<16 x i32> %270, <16 x i32> %271, i32 -1)
store volatile <16 x i32> %272, <16 x i32>* @VectorResult, align 64
%273 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%274 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %273, i32 -1)
%275 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%276 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%277 = call <16 x i32> @llvm.hexagon.V6.vmux(<64 x i1> %274, <16 x i32> %275, <16 x i32> %276)
store volatile <16 x i32> %277, <16 x i32>* @VectorResult, align 64
%278 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%279 = call <16 x i32> @llvm.hexagon.V6.vnot(<16 x i32> %278)
store volatile <16 x i32> %279, <16 x i32>* @VectorResult, align 64
%280 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%281 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%282 = call <16 x i32> @llvm.hexagon.V6.vor(<16 x i32> %280, <16 x i32> %281)
store volatile <16 x i32> %282, <16 x i32>* @VectorResult, align 64
%283 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%284 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%285 = call <16 x i32> @llvm.hexagon.V6.vrdelta(<16 x i32> %283, <16 x i32> %284)
store volatile <16 x i32> %285, <16 x i32>* @VectorResult, align 64
%286 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%287 = call <16 x i32> @llvm.hexagon.V6.vror(<16 x i32> %286, i32 -1)
store volatile <16 x i32> %287, <16 x i32>* @VectorResult, align 64
%288 = call <16 x i32> @llvm.hexagon.V6.lvsplatw(i32 -1)
store volatile <16 x i32> %288, <16 x i32>* @VectorResult, align 64
%289 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%290 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%291 = call <16 x i32> @llvm.hexagon.V6.vxor(<16 x i32> %289, <16 x i32> %290)
store volatile <16 x i32> %291, <16 x i32>* @VectorResult, align 64
%292 = call <16 x i32> @llvm.hexagon.V6.vd0()
store volatile <16 x i32> %292, <16 x i32>* @VectorResult, align 64
%293 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%294 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %293, i32 -1)
%295 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%296 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%297 = call <16 x i32> @llvm.hexagon.V6.vaddbq(<64 x i1> %294, <16 x i32> %295, <16 x i32> %296)
store volatile <16 x i32> %297, <16 x i32>* @VectorResult, align 64
%298 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%299 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %298, i32 -1)
%300 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%301 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%302 = call <16 x i32> @llvm.hexagon.V6.vaddbnq(<64 x i1> %299, <16 x i32> %300, <16 x i32> %301)
store volatile <16 x i32> %302, <16 x i32>* @VectorResult, align 64
%303 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%304 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %303, i32 -1)
%305 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%306 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%307 = call <16 x i32> @llvm.hexagon.V6.vsubbq(<64 x i1> %304, <16 x i32> %305, <16 x i32> %306)
store volatile <16 x i32> %307, <16 x i32>* @VectorResult, align 64
%308 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%309 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %308, i32 -1)
%310 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%311 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%312 = call <16 x i32> @llvm.hexagon.V6.vsubbnq(<64 x i1> %309, <16 x i32> %310, <16 x i32> %311)
store volatile <16 x i32> %312, <16 x i32>* @VectorResult, align 64
%313 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%314 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%315 = call <16 x i32> @llvm.hexagon.V6.vaddb(<16 x i32> %313, <16 x i32> %314)
store volatile <16 x i32> %315, <16 x i32>* @VectorResult, align 64
%316 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%317 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%318 = call <16 x i32> @llvm.hexagon.V6.vasrhbrndsat(<16 x i32> %316, <16 x i32> %317, i32 -1)
store volatile <16 x i32> %318, <16 x i32>* @VectorResult, align 64
%319 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%320 = call <16 x i32> @llvm.hexagon.V6.vdealb(<16 x i32> %319)
store volatile <16 x i32> %320, <16 x i32>* @VectorResult, align 64
%321 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%322 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%323 = call <16 x i32> @llvm.hexagon.V6.vdealb4w(<16 x i32> %321, <16 x i32> %322)
store volatile <16 x i32> %323, <16 x i32>* @VectorResult, align 64
%324 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%325 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%326 = call <16 x i32> @llvm.hexagon.V6.vlutvvb(<16 x i32> %324, <16 x i32> %325, i32 -1)
store volatile <16 x i32> %326, <16 x i32>* @VectorResult, align 64
%327 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%328 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%329 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%330 = call <16 x i32> @llvm.hexagon.V6.vlutvvb.oracc(<16 x i32> %327, <16 x i32> %328, <16 x i32> %329, i32 -1)
store volatile <16 x i32> %330, <16 x i32>* @VectorResult, align 64
%331 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%332 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%333 = call <16 x i32> @llvm.hexagon.V6.vnavgub(<16 x i32> %331, <16 x i32> %332)
store volatile <16 x i32> %333, <16 x i32>* @VectorResult, align 64
%334 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%335 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%336 = call <16 x i32> @llvm.hexagon.V6.vpackhb.sat(<16 x i32> %334, <16 x i32> %335)
store volatile <16 x i32> %336, <16 x i32>* @VectorResult, align 64
%337 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%338 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%339 = call <16 x i32> @llvm.hexagon.V6.vpackeb(<16 x i32> %337, <16 x i32> %338)
store volatile <16 x i32> %339, <16 x i32>* @VectorResult, align 64
%340 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%341 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%342 = call <16 x i32> @llvm.hexagon.V6.vpackob(<16 x i32> %340, <16 x i32> %341)
store volatile <16 x i32> %342, <16 x i32>* @VectorResult, align 64
%343 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%344 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%345 = call <16 x i32> @llvm.hexagon.V6.vroundhb(<16 x i32> %343, <16 x i32> %344)
store volatile <16 x i32> %345, <16 x i32>* @VectorResult, align 64
%346 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%347 = call <16 x i32> @llvm.hexagon.V6.vshuffb(<16 x i32> %346)
store volatile <16 x i32> %347, <16 x i32>* @VectorResult, align 64
%348 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%349 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%350 = call <16 x i32> @llvm.hexagon.V6.vshuffeb(<16 x i32> %348, <16 x i32> %349)
store volatile <16 x i32> %350, <16 x i32>* @VectorResult, align 64
%351 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%352 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%353 = call <16 x i32> @llvm.hexagon.V6.vshuffob(<16 x i32> %351, <16 x i32> %352)
store volatile <16 x i32> %353, <16 x i32>* @VectorResult, align 64
%354 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%355 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%356 = call <16 x i32> @llvm.hexagon.V6.vsubb(<16 x i32> %354, <16 x i32> %355)
store volatile <16 x i32> %356, <16 x i32>* @VectorResult, align 64
%357 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%358 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %357, i32 -1)
%359 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%360 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%361 = call <16 x i32> @llvm.hexagon.V6.vaddhq(<64 x i1> %358, <16 x i32> %359, <16 x i32> %360)
store volatile <16 x i32> %361, <16 x i32>* @VectorResult, align 64
%362 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%363 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %362, i32 -1)
%364 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%365 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%366 = call <16 x i32> @llvm.hexagon.V6.vaddhnq(<64 x i1> %363, <16 x i32> %364, <16 x i32> %365)
store volatile <16 x i32> %366, <16 x i32>* @VectorResult, align 64
%367 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%368 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %367, i32 -1)
%369 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%370 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%371 = call <16 x i32> @llvm.hexagon.V6.vsubhq(<64 x i1> %368, <16 x i32> %369, <16 x i32> %370)
store volatile <16 x i32> %371, <16 x i32>* @VectorResult, align 64
%372 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%373 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %372, i32 -1)
%374 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%375 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%376 = call <16 x i32> @llvm.hexagon.V6.vsubhnq(<64 x i1> %373, <16 x i32> %374, <16 x i32> %375)
store volatile <16 x i32> %376, <16 x i32>* @VectorResult, align 64
%377 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%378 = call <16 x i32> @llvm.hexagon.V6.vabsh(<16 x i32> %377)
store volatile <16 x i32> %378, <16 x i32>* @VectorResult, align 64
%379 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%380 = call <16 x i32> @llvm.hexagon.V6.vabsh.sat(<16 x i32> %379)
store volatile <16 x i32> %380, <16 x i32>* @VectorResult, align 64
%381 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%382 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%383 = call <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32> %381, <16 x i32> %382)
store volatile <16 x i32> %383, <16 x i32>* @VectorResult, align 64
%384 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%385 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%386 = call <16 x i32> @llvm.hexagon.V6.vaddhsat(<16 x i32> %384, <16 x i32> %385)
store volatile <16 x i32> %386, <16 x i32>* @VectorResult, align 64
%387 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%388 = call <16 x i32> @llvm.hexagon.V6.vaslh(<16 x i32> %387, i32 -1)
store volatile <16 x i32> %388, <16 x i32>* @VectorResult, align 64
%389 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%390 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%391 = call <16 x i32> @llvm.hexagon.V6.vaslhv(<16 x i32> %389, <16 x i32> %390)
store volatile <16 x i32> %391, <16 x i32>* @VectorResult, align 64
%392 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%393 = call <16 x i32> @llvm.hexagon.V6.vasrh(<16 x i32> %392, i32 -1)
store volatile <16 x i32> %393, <16 x i32>* @VectorResult, align 64
%394 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%395 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%396 = call <16 x i32> @llvm.hexagon.V6.vasrhv(<16 x i32> %394, <16 x i32> %395)
store volatile <16 x i32> %396, <16 x i32>* @VectorResult, align 64
%397 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%398 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%399 = call <16 x i32> @llvm.hexagon.V6.vasrwh(<16 x i32> %397, <16 x i32> %398, i32 -1)
store volatile <16 x i32> %399, <16 x i32>* @VectorResult, align 64
%400 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%401 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%402 = call <16 x i32> @llvm.hexagon.V6.vasrwhrndsat(<16 x i32> %400, <16 x i32> %401, i32 -1)
store volatile <16 x i32> %402, <16 x i32>* @VectorResult, align 64
%403 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%404 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%405 = call <16 x i32> @llvm.hexagon.V6.vasrwhsat(<16 x i32> %403, <16 x i32> %404, i32 -1)
store volatile <16 x i32> %405, <16 x i32>* @VectorResult, align 64
%406 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%407 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%408 = call <16 x i32> @llvm.hexagon.V6.vavgh(<16 x i32> %406, <16 x i32> %407)
store volatile <16 x i32> %408, <16 x i32>* @VectorResult, align 64
%409 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%410 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%411 = call <16 x i32> @llvm.hexagon.V6.vavghrnd(<16 x i32> %409, <16 x i32> %410)
store volatile <16 x i32> %411, <16 x i32>* @VectorResult, align 64
%412 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%413 = call <16 x i32> @llvm.hexagon.V6.vdealh(<16 x i32> %412)
store volatile <16 x i32> %413, <16 x i32>* @VectorResult, align 64
%414 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%415 = call <16 x i32> @llvm.hexagon.V6.vdmpybus(<16 x i32> %414, i32 -1)
store volatile <16 x i32> %415, <16 x i32>* @VectorResult, align 64
%416 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%417 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%418 = call <16 x i32> @llvm.hexagon.V6.vdmpybus.acc(<16 x i32> %416, <16 x i32> %417, i32 -1)
store volatile <16 x i32> %418, <16 x i32>* @VectorResult, align 64
%419 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%420 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%421 = call <16 x i32> @llvm.hexagon.V6.vlsrhv(<16 x i32> %419, <16 x i32> %420)
store volatile <16 x i32> %421, <16 x i32>* @VectorResult, align 64
%422 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%423 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%424 = call <16 x i32> @llvm.hexagon.V6.vmaxh(<16 x i32> %422, <16 x i32> %423)
store volatile <16 x i32> %424, <16 x i32>* @VectorResult, align 64
%425 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%426 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%427 = call <16 x i32> @llvm.hexagon.V6.vminh(<16 x i32> %425, <16 x i32> %426)
store volatile <16 x i32> %427, <16 x i32>* @VectorResult, align 64
%428 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%429 = call <16 x i32> @llvm.hexagon.V6.vmpyhsrs(<16 x i32> %428, i32 -1)
store volatile <16 x i32> %429, <16 x i32>* @VectorResult, align 64
%430 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%431 = call <16 x i32> @llvm.hexagon.V6.vmpyhss(<16 x i32> %430, i32 -1)
store volatile <16 x i32> %431, <16 x i32>* @VectorResult, align 64
%432 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%433 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%434 = call <16 x i32> @llvm.hexagon.V6.vmpyhvsrs(<16 x i32> %432, <16 x i32> %433)
store volatile <16 x i32> %434, <16 x i32>* @VectorResult, align 64
%435 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%436 = call <16 x i32> @llvm.hexagon.V6.vmpyihb(<16 x i32> %435, i32 -1)
store volatile <16 x i32> %436, <16 x i32>* @VectorResult, align 64
%437 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%438 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%439 = call <16 x i32> @llvm.hexagon.V6.vmpyih(<16 x i32> %437, <16 x i32> %438)
store volatile <16 x i32> %439, <16 x i32>* @VectorResult, align 64
%440 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%441 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%442 = call <16 x i32> @llvm.hexagon.V6.vmpyihb.acc(<16 x i32> %440, <16 x i32> %441, i32 -1)
store volatile <16 x i32> %442, <16 x i32>* @VectorResult, align 64
%443 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%444 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%445 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%446 = call <16 x i32> @llvm.hexagon.V6.vmpyih.acc(<16 x i32> %443, <16 x i32> %444, <16 x i32> %445)
store volatile <16 x i32> %446, <16 x i32>* @VectorResult, align 64
%447 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%448 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%449 = call <16 x i32> @llvm.hexagon.V6.vnavgh(<16 x i32> %447, <16 x i32> %448)
store volatile <16 x i32> %449, <16 x i32>* @VectorResult, align 64
%450 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%451 = call <16 x i32> @llvm.hexagon.V6.vnormamth(<16 x i32> %450)
store volatile <16 x i32> %451, <16 x i32>* @VectorResult, align 64
%452 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%453 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%454 = call <16 x i32> @llvm.hexagon.V6.vpackwh.sat(<16 x i32> %452, <16 x i32> %453)
store volatile <16 x i32> %454, <16 x i32>* @VectorResult, align 64
%455 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%456 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%457 = call <16 x i32> @llvm.hexagon.V6.vpackeh(<16 x i32> %455, <16 x i32> %456)
store volatile <16 x i32> %457, <16 x i32>* @VectorResult, align 64
%458 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%459 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%460 = call <16 x i32> @llvm.hexagon.V6.vpackoh(<16 x i32> %458, <16 x i32> %459)
store volatile <16 x i32> %460, <16 x i32>* @VectorResult, align 64
%461 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%462 = call <16 x i32> @llvm.hexagon.V6.vpopcounth(<16 x i32> %461)
store volatile <16 x i32> %462, <16 x i32>* @VectorResult, align 64
%463 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%464 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%465 = call <16 x i32> @llvm.hexagon.V6.vroundwh(<16 x i32> %463, <16 x i32> %464)
store volatile <16 x i32> %465, <16 x i32>* @VectorResult, align 64
%466 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%467 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%468 = call <16 x i32> @llvm.hexagon.V6.vsatwh(<16 x i32> %466, <16 x i32> %467)
store volatile <16 x i32> %468, <16 x i32>* @VectorResult, align 64
%469 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%470 = call <16 x i32> @llvm.hexagon.V6.vshuffh(<16 x i32> %469)
store volatile <16 x i32> %470, <16 x i32>* @VectorResult, align 64
%471 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%472 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%473 = call <16 x i32> @llvm.hexagon.V6.vshufeh(<16 x i32> %471, <16 x i32> %472)
store volatile <16 x i32> %473, <16 x i32>* @VectorResult, align 64
%474 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%475 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%476 = call <16 x i32> @llvm.hexagon.V6.vshufoh(<16 x i32> %474, <16 x i32> %475)
store volatile <16 x i32> %476, <16 x i32>* @VectorResult, align 64
%477 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%478 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%479 = call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %477, <16 x i32> %478)
store volatile <16 x i32> %479, <16 x i32>* @VectorResult, align 64
%480 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%481 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%482 = call <16 x i32> @llvm.hexagon.V6.vsubhsat(<16 x i32> %480, <16 x i32> %481)
store volatile <16 x i32> %482, <16 x i32>* @VectorResult, align 64
%483 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%484 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%485 = call <16 x i32> @llvm.hexagon.V6.vabsdiffub(<16 x i32> %483, <16 x i32> %484)
store volatile <16 x i32> %485, <16 x i32>* @VectorResult, align 64
%486 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%487 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%488 = call <16 x i32> @llvm.hexagon.V6.vaddubsat(<16 x i32> %486, <16 x i32> %487)
store volatile <16 x i32> %488, <16 x i32>* @VectorResult, align 64
%489 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%490 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%491 = call <16 x i32> @llvm.hexagon.V6.vasrhubrndsat(<16 x i32> %489, <16 x i32> %490, i32 -1)
store volatile <16 x i32> %491, <16 x i32>* @VectorResult, align 64
%492 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%493 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%494 = call <16 x i32> @llvm.hexagon.V6.vasrhubsat(<16 x i32> %492, <16 x i32> %493, i32 -1)
store volatile <16 x i32> %494, <16 x i32>* @VectorResult, align 64
%495 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%496 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%497 = call <16 x i32> @llvm.hexagon.V6.vavgub(<16 x i32> %495, <16 x i32> %496)
store volatile <16 x i32> %497, <16 x i32>* @VectorResult, align 64
%498 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%499 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%500 = call <16 x i32> @llvm.hexagon.V6.vavgubrnd(<16 x i32> %498, <16 x i32> %499)
store volatile <16 x i32> %500, <16 x i32>* @VectorResult, align 64
%501 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%502 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%503 = call <16 x i32> @llvm.hexagon.V6.vmaxub(<16 x i32> %501, <16 x i32> %502)
store volatile <16 x i32> %503, <16 x i32>* @VectorResult, align 64
%504 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%505 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%506 = call <16 x i32> @llvm.hexagon.V6.vminub(<16 x i32> %504, <16 x i32> %505)
store volatile <16 x i32> %506, <16 x i32>* @VectorResult, align 64
%507 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%508 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%509 = call <16 x i32> @llvm.hexagon.V6.vpackhub.sat(<16 x i32> %507, <16 x i32> %508)
store volatile <16 x i32> %509, <16 x i32>* @VectorResult, align 64
%510 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%511 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%512 = call <16 x i32> @llvm.hexagon.V6.vroundhub(<16 x i32> %510, <16 x i32> %511)
store volatile <16 x i32> %512, <16 x i32>* @VectorResult, align 64
%513 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%514 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%515 = call <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32> %513, <16 x i32> %514)
store volatile <16 x i32> %515, <16 x i32>* @VectorResult, align 64
%516 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%517 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%518 = call <16 x i32> @llvm.hexagon.V6.vsububsat(<16 x i32> %516, <16 x i32> %517)
store volatile <16 x i32> %518, <16 x i32>* @VectorResult, align 64
%519 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%520 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%521 = call <16 x i32> @llvm.hexagon.V6.vabsdiffh(<16 x i32> %519, <16 x i32> %520)
store volatile <16 x i32> %521, <16 x i32>* @VectorResult, align 64
%522 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%523 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%524 = call <16 x i32> @llvm.hexagon.V6.vabsdiffuh(<16 x i32> %522, <16 x i32> %523)
store volatile <16 x i32> %524, <16 x i32>* @VectorResult, align 64
%525 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%526 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%527 = call <16 x i32> @llvm.hexagon.V6.vadduhsat(<16 x i32> %525, <16 x i32> %526)
store volatile <16 x i32> %527, <16 x i32>* @VectorResult, align 64
%528 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%529 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%530 = call <16 x i32> @llvm.hexagon.V6.vasrwuhsat(<16 x i32> %528, <16 x i32> %529, i32 -1)
store volatile <16 x i32> %530, <16 x i32>* @VectorResult, align 64
%531 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%532 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%533 = call <16 x i32> @llvm.hexagon.V6.vavguh(<16 x i32> %531, <16 x i32> %532)
store volatile <16 x i32> %533, <16 x i32>* @VectorResult, align 64
%534 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%535 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%536 = call <16 x i32> @llvm.hexagon.V6.vavguhrnd(<16 x i32> %534, <16 x i32> %535)
store volatile <16 x i32> %536, <16 x i32>* @VectorResult, align 64
%537 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%538 = call <16 x i32> @llvm.hexagon.V6.vcl0h(<16 x i32> %537)
store volatile <16 x i32> %538, <16 x i32>* @VectorResult, align 64
%539 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%540 = call <16 x i32> @llvm.hexagon.V6.vlsrh(<16 x i32> %539, i32 -1)
store volatile <16 x i32> %540, <16 x i32>* @VectorResult, align 64
%541 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%542 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%543 = call <16 x i32> @llvm.hexagon.V6.vmaxuh(<16 x i32> %541, <16 x i32> %542)
store volatile <16 x i32> %543, <16 x i32>* @VectorResult, align 64
%544 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%545 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%546 = call <16 x i32> @llvm.hexagon.V6.vminuh(<16 x i32> %544, <16 x i32> %545)
store volatile <16 x i32> %546, <16 x i32>* @VectorResult, align 64
%547 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%548 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%549 = call <16 x i32> @llvm.hexagon.V6.vpackwuh.sat(<16 x i32> %547, <16 x i32> %548)
store volatile <16 x i32> %549, <16 x i32>* @VectorResult, align 64
%550 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%551 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%552 = call <16 x i32> @llvm.hexagon.V6.vroundwuh(<16 x i32> %550, <16 x i32> %551)
store volatile <16 x i32> %552, <16 x i32>* @VectorResult, align 64
%553 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%554 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%555 = call <16 x i32> @llvm.hexagon.V6.vsubuhsat(<16 x i32> %553, <16 x i32> %554)
store volatile <16 x i32> %555, <16 x i32>* @VectorResult, align 64
%556 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%557 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%558 = call <16 x i32> @llvm.hexagon.V6.vabsdiffw(<16 x i32> %556, <16 x i32> %557)
store volatile <16 x i32> %558, <16 x i32>* @VectorResult, align 64
%559 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%560 = call <16 x i32> @llvm.hexagon.V6.vcl0w(<16 x i32> %559)
store volatile <16 x i32> %560, <16 x i32>* @VectorResult, align 64
%561 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%562 = call <16 x i32> @llvm.hexagon.V6.vlsrw(<16 x i32> %561, i32 -1)
store volatile <16 x i32> %562, <16 x i32>* @VectorResult, align 64
%563 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%564 = call <16 x i32> @llvm.hexagon.V6.vrmpyub(<16 x i32> %563, i32 -1)
store volatile <16 x i32> %564, <16 x i32>* @VectorResult, align 64
%565 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%566 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%567 = call <16 x i32> @llvm.hexagon.V6.vrmpyubv(<16 x i32> %565, <16 x i32> %566)
store volatile <16 x i32> %567, <16 x i32>* @VectorResult, align 64
%568 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%569 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%570 = call <16 x i32> @llvm.hexagon.V6.vrmpyub.acc(<16 x i32> %568, <16 x i32> %569, i32 -1)
store volatile <16 x i32> %570, <16 x i32>* @VectorResult, align 64
%571 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%572 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%573 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%574 = call <16 x i32> @llvm.hexagon.V6.vrmpyubv.acc(<16 x i32> %571, <16 x i32> %572, <16 x i32> %573)
store volatile <16 x i32> %574, <16 x i32>* @VectorResult, align 64
%575 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%576 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %575, i32 -1)
%577 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%578 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%579 = call <16 x i32> @llvm.hexagon.V6.vaddwq(<64 x i1> %576, <16 x i32> %577, <16 x i32> %578)
store volatile <16 x i32> %579, <16 x i32>* @VectorResult, align 64
%580 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%581 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %580, i32 -1)
%582 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%583 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%584 = call <16 x i32> @llvm.hexagon.V6.vaddwnq(<64 x i1> %581, <16 x i32> %582, <16 x i32> %583)
store volatile <16 x i32> %584, <16 x i32>* @VectorResult, align 64
%585 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%586 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %585, i32 -1)
%587 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%588 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%589 = call <16 x i32> @llvm.hexagon.V6.vsubwq(<64 x i1> %586, <16 x i32> %587, <16 x i32> %588)
store volatile <16 x i32> %589, <16 x i32>* @VectorResult, align 64
%590 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%591 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %590, i32 -1)
%592 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%593 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%594 = call <16 x i32> @llvm.hexagon.V6.vsubwnq(<64 x i1> %591, <16 x i32> %592, <16 x i32> %593)
store volatile <16 x i32> %594, <16 x i32>* @VectorResult, align 64
%595 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%596 = call <16 x i32> @llvm.hexagon.V6.vabsw(<16 x i32> %595)
store volatile <16 x i32> %596, <16 x i32>* @VectorResult, align 64
%597 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%598 = call <16 x i32> @llvm.hexagon.V6.vabsw.sat(<16 x i32> %597)
store volatile <16 x i32> %598, <16 x i32>* @VectorResult, align 64
%599 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%600 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%601 = call <16 x i32> @llvm.hexagon.V6.vaddw(<16 x i32> %599, <16 x i32> %600)
store volatile <16 x i32> %601, <16 x i32>* @VectorResult, align 64
%602 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%603 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%604 = call <16 x i32> @llvm.hexagon.V6.vaddwsat(<16 x i32> %602, <16 x i32> %603)
store volatile <16 x i32> %604, <16 x i32>* @VectorResult, align 64
%605 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%606 = call <16 x i32> @llvm.hexagon.V6.vaslw(<16 x i32> %605, i32 -1)
store volatile <16 x i32> %606, <16 x i32>* @VectorResult, align 64
%607 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%608 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%609 = call <16 x i32> @llvm.hexagon.V6.vaslwv(<16 x i32> %607, <16 x i32> %608)
store volatile <16 x i32> %609, <16 x i32>* @VectorResult, align 64
%610 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%611 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%612 = call <16 x i32> @llvm.hexagon.V6.vaslw.acc(<16 x i32> %610, <16 x i32> %611, i32 -1)
store volatile <16 x i32> %612, <16 x i32>* @VectorResult, align 64
%613 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%614 = call <16 x i32> @llvm.hexagon.V6.vasrw(<16 x i32> %613, i32 -1)
store volatile <16 x i32> %614, <16 x i32>* @VectorResult, align 64
%615 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%616 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%617 = call <16 x i32> @llvm.hexagon.V6.vasrwv(<16 x i32> %615, <16 x i32> %616)
store volatile <16 x i32> %617, <16 x i32>* @VectorResult, align 64
%618 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%619 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%620 = call <16 x i32> @llvm.hexagon.V6.vasrw.acc(<16 x i32> %618, <16 x i32> %619, i32 -1)
store volatile <16 x i32> %620, <16 x i32>* @VectorResult, align 64
%621 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%622 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%623 = call <16 x i32> @llvm.hexagon.V6.vavgw(<16 x i32> %621, <16 x i32> %622)
store volatile <16 x i32> %623, <16 x i32>* @VectorResult, align 64
%624 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%625 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%626 = call <16 x i32> @llvm.hexagon.V6.vavgwrnd(<16 x i32> %624, <16 x i32> %625)
store volatile <16 x i32> %626, <16 x i32>* @VectorResult, align 64
%627 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%628 = call <16 x i32> @llvm.hexagon.V6.vdmpyhb(<16 x i32> %627, i32 -1)
store volatile <16 x i32> %628, <16 x i32>* @VectorResult, align 64
%629 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%630 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsat(<16 x i32> %629, i32 -1)
store volatile <16 x i32> %630, <16 x i32>* @VectorResult, align 64
%631 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%632 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsusat(<16 x i32> %631, i32 -1)
store volatile <16 x i32> %632, <16 x i32>* @VectorResult, align 64
%633 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%634 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%635 = call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat(<16 x i32> %633, <16 x i32> %634)
store volatile <16 x i32> %635, <16 x i32>* @VectorResult, align 64
%636 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%637 = call <16 x i32> @llvm.hexagon.V6.vdmpyhisat(<32 x i32> %636, i32 -1)
store volatile <16 x i32> %637, <16 x i32>* @VectorResult, align 64
%638 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%639 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsuisat(<32 x i32> %638, i32 -1)
store volatile <16 x i32> %639, <16 x i32>* @VectorResult, align 64
%640 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%641 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%642 = call <16 x i32> @llvm.hexagon.V6.vdmpyhb.acc(<16 x i32> %640, <16 x i32> %641, i32 -1)
store volatile <16 x i32> %642, <16 x i32>* @VectorResult, align 64
%643 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%644 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%645 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsat.acc(<16 x i32> %643, <16 x i32> %644, i32 -1)
store volatile <16 x i32> %645, <16 x i32>* @VectorResult, align 64
%646 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%647 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%648 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsusat.acc(<16 x i32> %646, <16 x i32> %647, i32 -1)
store volatile <16 x i32> %648, <16 x i32>* @VectorResult, align 64
%649 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%650 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%651 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%652 = call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32> %649, <16 x i32> %650, <16 x i32> %651)
store volatile <16 x i32> %652, <16 x i32>* @VectorResult, align 64
%653 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%654 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%655 = call <16 x i32> @llvm.hexagon.V6.vdmpyhisat.acc(<16 x i32> %653, <32 x i32> %654, i32 -1)
store volatile <16 x i32> %655, <16 x i32>* @VectorResult, align 64
%656 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%657 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%658 = call <16 x i32> @llvm.hexagon.V6.vdmpyhsuisat.acc(<16 x i32> %656, <32 x i32> %657, i32 -1)
store volatile <16 x i32> %658, <16 x i32>* @VectorResult, align 64
%659 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%660 = call <16 x i32> @llvm.hexagon.V6.vinsertwr(<16 x i32> %659, i32 -1)
store volatile <16 x i32> %660, <16 x i32>* @VectorResult, align 64
%661 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%662 = call <16 x i32> @llvm.hexagon.V6.vinsertwr(<16 x i32> %661, i32 0)
store volatile <16 x i32> %662, <16 x i32>* @VectorResult, align 64
%663 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%664 = call <16 x i32> @llvm.hexagon.V6.vinsertwr(<16 x i32> %663, i32 1)
store volatile <16 x i32> %664, <16 x i32>* @VectorResult, align 64
%665 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%666 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%667 = call <16 x i32> @llvm.hexagon.V6.vlsrwv(<16 x i32> %665, <16 x i32> %666)
store volatile <16 x i32> %667, <16 x i32>* @VectorResult, align 64
%668 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%669 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%670 = call <16 x i32> @llvm.hexagon.V6.vmaxw(<16 x i32> %668, <16 x i32> %669)
store volatile <16 x i32> %670, <16 x i32>* @VectorResult, align 64
%671 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%672 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%673 = call <16 x i32> @llvm.hexagon.V6.vminw(<16 x i32> %671, <16 x i32> %672)
store volatile <16 x i32> %673, <16 x i32>* @VectorResult, align 64
%674 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%675 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%676 = call <16 x i32> @llvm.hexagon.V6.vmpyewuh(<16 x i32> %674, <16 x i32> %675)
store volatile <16 x i32> %676, <16 x i32>* @VectorResult, align 64
%677 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%678 = call <16 x i32> @llvm.hexagon.V6.vmpyiwb(<16 x i32> %677, i32 -1)
store volatile <16 x i32> %678, <16 x i32>* @VectorResult, align 64
%679 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%680 = call <16 x i32> @llvm.hexagon.V6.vmpyiwh(<16 x i32> %679, i32 -1)
store volatile <16 x i32> %680, <16 x i32>* @VectorResult, align 64
%681 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%682 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%683 = call <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32> %681, <16 x i32> %682, i32 -1)
store volatile <16 x i32> %683, <16 x i32>* @VectorResult, align 64
%684 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%685 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%686 = call <16 x i32> @llvm.hexagon.V6.vmpyiwh.acc(<16 x i32> %684, <16 x i32> %685, i32 -1)
store volatile <16 x i32> %686, <16 x i32>* @VectorResult, align 64
%687 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%688 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%689 = call <16 x i32> @llvm.hexagon.V6.vmpyiewuh(<16 x i32> %687, <16 x i32> %688)
store volatile <16 x i32> %689, <16 x i32>* @VectorResult, align 64
%690 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%691 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%692 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%693 = call <16 x i32> @llvm.hexagon.V6.vmpyiewh.acc(<16 x i32> %690, <16 x i32> %691, <16 x i32> %692)
store volatile <16 x i32> %693, <16 x i32>* @VectorResult, align 64
%694 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%695 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%696 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%697 = call <16 x i32> @llvm.hexagon.V6.vmpyiewuh.acc(<16 x i32> %694, <16 x i32> %695, <16 x i32> %696)
store volatile <16 x i32> %697, <16 x i32>* @VectorResult, align 64
%698 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%699 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%700 = call <16 x i32> @llvm.hexagon.V6.vmpyieoh(<16 x i32> %698, <16 x i32> %699)
store volatile <16 x i32> %700, <16 x i32>* @VectorResult, align 64
%701 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%702 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%703 = call <16 x i32> @llvm.hexagon.V6.vmpyiowh(<16 x i32> %701, <16 x i32> %702)
store volatile <16 x i32> %703, <16 x i32>* @VectorResult, align 64
%704 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%705 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%706 = call <16 x i32> @llvm.hexagon.V6.vmpyowh.rnd(<16 x i32> %704, <16 x i32> %705)
store volatile <16 x i32> %706, <16 x i32>* @VectorResult, align 64
%707 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%708 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%709 = call <16 x i32> @llvm.hexagon.V6.vmpyowh(<16 x i32> %707, <16 x i32> %708)
store volatile <16 x i32> %709, <16 x i32>* @VectorResult, align 64
%710 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%711 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%712 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%713 = call <16 x i32> @llvm.hexagon.V6.vmpyowh.rnd.sacc(<16 x i32> %710, <16 x i32> %711, <16 x i32> %712)
store volatile <16 x i32> %713, <16 x i32>* @VectorResult, align 64
%714 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%715 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%716 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%717 = call <16 x i32> @llvm.hexagon.V6.vmpyowh.sacc(<16 x i32> %714, <16 x i32> %715, <16 x i32> %716)
store volatile <16 x i32> %717, <16 x i32>* @VectorResult, align 64
%718 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%719 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%720 = call <16 x i32> @llvm.hexagon.V6.vnavgw(<16 x i32> %718, <16 x i32> %719)
store volatile <16 x i32> %720, <16 x i32>* @VectorResult, align 64
%721 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%722 = call <16 x i32> @llvm.hexagon.V6.vnormamtw(<16 x i32> %721)
store volatile <16 x i32> %722, <16 x i32>* @VectorResult, align 64
%723 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%724 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%725 = call <16 x i32> @llvm.hexagon.V6.vrmpybv(<16 x i32> %723, <16 x i32> %724)
store volatile <16 x i32> %725, <16 x i32>* @VectorResult, align 64
%726 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%727 = call <16 x i32> @llvm.hexagon.V6.vrmpybus(<16 x i32> %726, i32 -1)
store volatile <16 x i32> %727, <16 x i32>* @VectorResult, align 64
%728 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%729 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%730 = call <16 x i32> @llvm.hexagon.V6.vrmpybusv(<16 x i32> %728, <16 x i32> %729)
store volatile <16 x i32> %730, <16 x i32>* @VectorResult, align 64
%731 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%732 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%733 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%734 = call <16 x i32> @llvm.hexagon.V6.vrmpybv.acc(<16 x i32> %731, <16 x i32> %732, <16 x i32> %733)
store volatile <16 x i32> %734, <16 x i32>* @VectorResult, align 64
%735 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%736 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%737 = call <16 x i32> @llvm.hexagon.V6.vrmpybus.acc(<16 x i32> %735, <16 x i32> %736, i32 -1)
store volatile <16 x i32> %737, <16 x i32>* @VectorResult, align 64
%738 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%739 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%740 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 2), align 64
%741 = call <16 x i32> @llvm.hexagon.V6.vrmpybusv.acc(<16 x i32> %738, <16 x i32> %739, <16 x i32> %740)
store volatile <16 x i32> %741, <16 x i32>* @VectorResult, align 64
%742 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%743 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%744 = call <16 x i32> @llvm.hexagon.V6.vsubw(<16 x i32> %742, <16 x i32> %743)
store volatile <16 x i32> %744, <16 x i32>* @VectorResult, align 64
%745 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%746 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%747 = call <16 x i32> @llvm.hexagon.V6.vsubwsat(<16 x i32> %745, <16 x i32> %746)
store volatile <16 x i32> %747, <16 x i32>* @VectorResult, align 64
%748 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%749 = call <32 x i32> @llvm.hexagon.V6.vassignp(<32 x i32> %748)
store volatile <32 x i32> %749, <32 x i32>* @VectorPairResult, align 128
%750 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%751 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%752 = call <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32> %750, <16 x i32> %751)
store volatile <32 x i32> %752, <32 x i32>* @VectorPairResult, align 128
%753 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%754 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%755 = call <32 x i32> @llvm.hexagon.V6.vdealvdd(<16 x i32> %753, <16 x i32> %754, i32 -1)
store volatile <32 x i32> %755, <32 x i32>* @VectorPairResult, align 128
%756 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%757 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%758 = call <32 x i32> @llvm.hexagon.V6.vshuffvdd(<16 x i32> %756, <16 x i32> %757, i32 -1)
store volatile <32 x i32> %758, <32 x i32>* @VectorPairResult, align 128
%759 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%760 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%761 = call <32 x i32> @llvm.hexagon.V6.vshuffvdd(<16 x i32> %759, <16 x i32> %760, i32 0)
store volatile <32 x i32> %761, <32 x i32>* @VectorPairResult, align 128
%762 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%763 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%764 = call <32 x i32> @llvm.hexagon.V6.vshuffvdd(<16 x i32> %762, <16 x i32> %763, i32 1)
store volatile <32 x i32> %764, <32 x i32>* @VectorPairResult, align 128
%765 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vecpreds, i32 0, i32 0), align 64
%766 = tail call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %765, i32 -1)
%767 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%768 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%769 = call <32 x i32> @llvm.hexagon.V6.vswap(<64 x i1> %766, <16 x i32> %767, <16 x i32> %768)
store volatile <32 x i32> %769, <32 x i32>* @VectorPairResult, align 128
%770 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%771 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%772 = call <32 x i32> @llvm.hexagon.V6.vaddb.dv(<32 x i32> %770, <32 x i32> %771)
store volatile <32 x i32> %772, <32 x i32>* @VectorPairResult, align 128
%773 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%774 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%775 = call <32 x i32> @llvm.hexagon.V6.vshufoeb(<16 x i32> %773, <16 x i32> %774)
store volatile <32 x i32> %775, <32 x i32>* @VectorPairResult, align 128
%776 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%777 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%778 = call <32 x i32> @llvm.hexagon.V6.vsubb.dv(<32 x i32> %776, <32 x i32> %777)
store volatile <32 x i32> %778, <32 x i32>* @VectorPairResult, align 128
%779 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%780 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%781 = call <32 x i32> @llvm.hexagon.V6.vaddubh(<16 x i32> %779, <16 x i32> %780)
store volatile <32 x i32> %781, <32 x i32>* @VectorPairResult, align 128
%782 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%783 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%784 = call <32 x i32> @llvm.hexagon.V6.vaddh.dv(<32 x i32> %782, <32 x i32> %783)
store volatile <32 x i32> %784, <32 x i32>* @VectorPairResult, align 128
%785 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%786 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%787 = call <32 x i32> @llvm.hexagon.V6.vaddhsat.dv(<32 x i32> %785, <32 x i32> %786)
store volatile <32 x i32> %787, <32 x i32>* @VectorPairResult, align 128
%788 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%789 = call <32 x i32> @llvm.hexagon.V6.vdmpybus.dv(<32 x i32> %788, i32 -1)
store volatile <32 x i32> %789, <32 x i32>* @VectorPairResult, align 128
%790 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%791 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%792 = call <32 x i32> @llvm.hexagon.V6.vdmpybus.dv.acc(<32 x i32> %790, <32 x i32> %791, i32 -1)
store volatile <32 x i32> %792, <32 x i32>* @VectorPairResult, align 128
%793 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%794 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%795 = call <32 x i32> @llvm.hexagon.V6.vlutvwh(<16 x i32> %793, <16 x i32> %794, i32 -1)
store volatile <32 x i32> %795, <32 x i32>* @VectorPairResult, align 128
%796 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%797 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%798 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%799 = call <32 x i32> @llvm.hexagon.V6.vlutvwh.oracc(<32 x i32> %796, <16 x i32> %797, <16 x i32> %798, i32 -1)
store volatile <32 x i32> %799, <32 x i32>* @VectorPairResult, align 128
%800 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%801 = call <32 x i32> @llvm.hexagon.V6.vmpabus(<32 x i32> %800, i32 -1)
store volatile <32 x i32> %801, <32 x i32>* @VectorPairResult, align 128
%802 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%803 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%804 = call <32 x i32> @llvm.hexagon.V6.vmpabusv(<32 x i32> %802, <32 x i32> %803)
store volatile <32 x i32> %804, <32 x i32>* @VectorPairResult, align 128
%805 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%806 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%807 = call <32 x i32> @llvm.hexagon.V6.vmpabuuv(<32 x i32> %805, <32 x i32> %806)
store volatile <32 x i32> %807, <32 x i32>* @VectorPairResult, align 128
%808 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%809 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%810 = call <32 x i32> @llvm.hexagon.V6.vmpabus.acc(<32 x i32> %808, <32 x i32> %809, i32 -1)
store volatile <32 x i32> %810, <32 x i32>* @VectorPairResult, align 128
%811 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%812 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%813 = call <32 x i32> @llvm.hexagon.V6.vmpybv(<16 x i32> %811, <16 x i32> %812)
store volatile <32 x i32> %813, <32 x i32>* @VectorPairResult, align 128
%814 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%815 = call <32 x i32> @llvm.hexagon.V6.vmpybus(<16 x i32> %814, i32 -1)
store volatile <32 x i32> %815, <32 x i32>* @VectorPairResult, align 128
%816 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%817 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%818 = call <32 x i32> @llvm.hexagon.V6.vmpybusv(<16 x i32> %816, <16 x i32> %817)
store volatile <32 x i32> %818, <32 x i32>* @VectorPairResult, align 128
%819 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%820 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%821 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%822 = call <32 x i32> @llvm.hexagon.V6.vmpybv.acc(<32 x i32> %819, <16 x i32> %820, <16 x i32> %821)
store volatile <32 x i32> %822, <32 x i32>* @VectorPairResult, align 128
%823 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%824 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%825 = call <32 x i32> @llvm.hexagon.V6.vmpybus.acc(<32 x i32> %823, <16 x i32> %824, i32 -1)
store volatile <32 x i32> %825, <32 x i32>* @VectorPairResult, align 128
%826 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%827 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%828 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%829 = call <32 x i32> @llvm.hexagon.V6.vmpybusv.acc(<32 x i32> %826, <16 x i32> %827, <16 x i32> %828)
store volatile <32 x i32> %829, <32 x i32>* @VectorPairResult, align 128
%830 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%831 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%832 = call <32 x i32> @llvm.hexagon.V6.vshufoeh(<16 x i32> %830, <16 x i32> %831)
store volatile <32 x i32> %832, <32 x i32>* @VectorPairResult, align 128
%833 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%834 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%835 = call <32 x i32> @llvm.hexagon.V6.vsububh(<16 x i32> %833, <16 x i32> %834)
store volatile <32 x i32> %835, <32 x i32>* @VectorPairResult, align 128
%836 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%837 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%838 = call <32 x i32> @llvm.hexagon.V6.vsubh.dv(<32 x i32> %836, <32 x i32> %837)
store volatile <32 x i32> %838, <32 x i32>* @VectorPairResult, align 128
%839 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%840 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%841 = call <32 x i32> @llvm.hexagon.V6.vsubhsat.dv(<32 x i32> %839, <32 x i32> %840)
store volatile <32 x i32> %841, <32 x i32>* @VectorPairResult, align 128
%842 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%843 = call <32 x i32> @llvm.hexagon.V6.vsb(<16 x i32> %842)
store volatile <32 x i32> %843, <32 x i32>* @VectorPairResult, align 128
%844 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%845 = call <32 x i32> @llvm.hexagon.V6.vtmpyb(<32 x i32> %844, i32 -1)
store volatile <32 x i32> %845, <32 x i32>* @VectorPairResult, align 128
%846 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%847 = call <32 x i32> @llvm.hexagon.V6.vtmpybus(<32 x i32> %846, i32 -1)
store volatile <32 x i32> %847, <32 x i32>* @VectorPairResult, align 128
%848 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%849 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%850 = call <32 x i32> @llvm.hexagon.V6.vtmpyb.acc(<32 x i32> %848, <32 x i32> %849, i32 -1)
store volatile <32 x i32> %850, <32 x i32>* @VectorPairResult, align 128
%851 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%852 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%853 = call <32 x i32> @llvm.hexagon.V6.vtmpybus.acc(<32 x i32> %851, <32 x i32> %852, i32 -1)
store volatile <32 x i32> %853, <32 x i32>* @VectorPairResult, align 128
%854 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%855 = call <32 x i32> @llvm.hexagon.V6.vunpackb(<16 x i32> %854)
store volatile <32 x i32> %855, <32 x i32>* @VectorPairResult, align 128
%856 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%857 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%858 = call <32 x i32> @llvm.hexagon.V6.vunpackob(<32 x i32> %856, <16 x i32> %857)
store volatile <32 x i32> %858, <32 x i32>* @VectorPairResult, align 128
%859 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%860 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%861 = call <32 x i32> @llvm.hexagon.V6.vaddubsat.dv(<32 x i32> %859, <32 x i32> %860)
store volatile <32 x i32> %861, <32 x i32>* @VectorPairResult, align 128
%862 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%863 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%864 = call <32 x i32> @llvm.hexagon.V6.vsububsat.dv(<32 x i32> %862, <32 x i32> %863)
store volatile <32 x i32> %864, <32 x i32>* @VectorPairResult, align 128
%865 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%866 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%867 = call <32 x i32> @llvm.hexagon.V6.vadduhsat.dv(<32 x i32> %865, <32 x i32> %866)
store volatile <32 x i32> %867, <32 x i32>* @VectorPairResult, align 128
%868 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%869 = call <32 x i32> @llvm.hexagon.V6.vmpyub(<16 x i32> %868, i32 -1)
store volatile <32 x i32> %869, <32 x i32>* @VectorPairResult, align 128
%870 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%871 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%872 = call <32 x i32> @llvm.hexagon.V6.vmpyubv(<16 x i32> %870, <16 x i32> %871)
store volatile <32 x i32> %872, <32 x i32>* @VectorPairResult, align 128
%873 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%874 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%875 = call <32 x i32> @llvm.hexagon.V6.vmpyub.acc(<32 x i32> %873, <16 x i32> %874, i32 -1)
store volatile <32 x i32> %875, <32 x i32>* @VectorPairResult, align 128
%876 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%877 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%878 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%879 = call <32 x i32> @llvm.hexagon.V6.vmpyubv.acc(<32 x i32> %876, <16 x i32> %877, <16 x i32> %878)
store volatile <32 x i32> %879, <32 x i32>* @VectorPairResult, align 128
%880 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%881 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%882 = call <32 x i32> @llvm.hexagon.V6.vsubuhsat.dv(<32 x i32> %880, <32 x i32> %881)
store volatile <32 x i32> %882, <32 x i32>* @VectorPairResult, align 128
%883 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%884 = call <32 x i32> @llvm.hexagon.V6.vunpackub(<16 x i32> %883)
store volatile <32 x i32> %884, <32 x i32>* @VectorPairResult, align 128
%885 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%886 = call <32 x i32> @llvm.hexagon.V6.vzb(<16 x i32> %885)
store volatile <32 x i32> %886, <32 x i32>* @VectorPairResult, align 128
%887 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%888 = call <32 x i32> @llvm.hexagon.V6.vdsaduh(<32 x i32> %887, i32 -1)
store volatile <32 x i32> %888, <32 x i32>* @VectorPairResult, align 128
%889 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%890 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%891 = call <32 x i32> @llvm.hexagon.V6.vdsaduh.acc(<32 x i32> %889, <32 x i32> %890, i32 -1)
store volatile <32 x i32> %891, <32 x i32>* @VectorPairResult, align 128
%892 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%893 = call <32 x i32> @llvm.hexagon.V6.vmpyuh(<16 x i32> %892, i32 -1)
store volatile <32 x i32> %893, <32 x i32>* @VectorPairResult, align 128
%894 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%895 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%896 = call <32 x i32> @llvm.hexagon.V6.vmpyuhv(<16 x i32> %894, <16 x i32> %895)
store volatile <32 x i32> %896, <32 x i32>* @VectorPairResult, align 128
%897 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%898 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%899 = call <32 x i32> @llvm.hexagon.V6.vmpyuh.acc(<32 x i32> %897, <16 x i32> %898, i32 -1)
store volatile <32 x i32> %899, <32 x i32>* @VectorPairResult, align 128
%900 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%901 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%902 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%903 = call <32 x i32> @llvm.hexagon.V6.vmpyuhv.acc(<32 x i32> %900, <16 x i32> %901, <16 x i32> %902)
store volatile <32 x i32> %903, <32 x i32>* @VectorPairResult, align 128
%904 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%905 = call <32 x i32> @llvm.hexagon.V6.vrmpyubi(<32 x i32> %904, i32 -1, i32 0)
store volatile <32 x i32> %905, <32 x i32>* @VectorPairResult, align 128
%906 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%907 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%908 = call <32 x i32> @llvm.hexagon.V6.vrmpyubi.acc(<32 x i32> %906, <32 x i32> %907, i32 -1, i32 0)
store volatile <32 x i32> %908, <32 x i32>* @VectorPairResult, align 128
%909 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%910 = call <32 x i32> @llvm.hexagon.V6.vrsadubi(<32 x i32> %909, i32 -1, i32 0)
store volatile <32 x i32> %910, <32 x i32>* @VectorPairResult, align 128
%911 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%912 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%913 = call <32 x i32> @llvm.hexagon.V6.vrsadubi.acc(<32 x i32> %911, <32 x i32> %912, i32 -1, i32 0)
store volatile <32 x i32> %913, <32 x i32>* @VectorPairResult, align 128
%914 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%915 = call <32 x i32> @llvm.hexagon.V6.vunpackuh(<16 x i32> %914)
store volatile <32 x i32> %915, <32 x i32>* @VectorPairResult, align 128
%916 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%917 = call <32 x i32> @llvm.hexagon.V6.vzh(<16 x i32> %916)
store volatile <32 x i32> %917, <32 x i32>* @VectorPairResult, align 128
%918 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%919 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%920 = call <32 x i32> @llvm.hexagon.V6.vaddhw(<16 x i32> %918, <16 x i32> %919)
store volatile <32 x i32> %920, <32 x i32>* @VectorPairResult, align 128
%921 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%922 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%923 = call <32 x i32> @llvm.hexagon.V6.vadduhw(<16 x i32> %921, <16 x i32> %922)
store volatile <32 x i32> %923, <32 x i32>* @VectorPairResult, align 128
%924 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%925 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%926 = call <32 x i32> @llvm.hexagon.V6.vaddw.dv(<32 x i32> %924, <32 x i32> %925)
store volatile <32 x i32> %926, <32 x i32>* @VectorPairResult, align 128
%927 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%928 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%929 = call <32 x i32> @llvm.hexagon.V6.vaddwsat.dv(<32 x i32> %927, <32 x i32> %928)
store volatile <32 x i32> %929, <32 x i32>* @VectorPairResult, align 128
%930 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%931 = call <32 x i32> @llvm.hexagon.V6.vdmpyhb.dv(<32 x i32> %930, i32 -1)
store volatile <32 x i32> %931, <32 x i32>* @VectorPairResult, align 128
%932 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%933 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%934 = call <32 x i32> @llvm.hexagon.V6.vdmpyhb.dv.acc(<32 x i32> %932, <32 x i32> %933, i32 -1)
store volatile <32 x i32> %934, <32 x i32>* @VectorPairResult, align 128
%935 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%936 = call <32 x i32> @llvm.hexagon.V6.vmpahb(<32 x i32> %935, i32 -1)
store volatile <32 x i32> %936, <32 x i32>* @VectorPairResult, align 128
%937 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%938 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%939 = call <32 x i32> @llvm.hexagon.V6.vmpahb.acc(<32 x i32> %937, <32 x i32> %938, i32 -1)
store volatile <32 x i32> %939, <32 x i32>* @VectorPairResult, align 128
%940 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%941 = call <32 x i32> @llvm.hexagon.V6.vmpyh(<16 x i32> %940, i32 -1)
store volatile <32 x i32> %941, <32 x i32>* @VectorPairResult, align 128
%942 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%943 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%944 = call <32 x i32> @llvm.hexagon.V6.vmpyhv(<16 x i32> %942, <16 x i32> %943)
store volatile <32 x i32> %944, <32 x i32>* @VectorPairResult, align 128
%945 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%946 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%947 = call <32 x i32> @llvm.hexagon.V6.vmpyhus(<16 x i32> %945, <16 x i32> %946)
store volatile <32 x i32> %947, <32 x i32>* @VectorPairResult, align 128
%948 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%949 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%950 = call <32 x i32> @llvm.hexagon.V6.vmpyhsat.acc(<32 x i32> %948, <16 x i32> %949, i32 -1)
store volatile <32 x i32> %950, <32 x i32>* @VectorPairResult, align 128
%951 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%952 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%953 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%954 = call <32 x i32> @llvm.hexagon.V6.vmpyhv.acc(<32 x i32> %951, <16 x i32> %952, <16 x i32> %953)
store volatile <32 x i32> %954, <32 x i32>* @VectorPairResult, align 128
%955 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%956 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%957 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%958 = call <32 x i32> @llvm.hexagon.V6.vmpyhus.acc(<32 x i32> %955, <16 x i32> %956, <16 x i32> %957)
store volatile <32 x i32> %958, <32 x i32>* @VectorPairResult, align 128
%959 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%960 = call <32 x i32> @llvm.hexagon.V6.vrmpybusi(<32 x i32> %959, i32 -1, i32 0)
store volatile <32 x i32> %960, <32 x i32>* @VectorPairResult, align 128
%961 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%962 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%963 = call <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32> %961, <32 x i32> %962, i32 -1, i32 0)
store volatile <32 x i32> %963, <32 x i32>* @VectorPairResult, align 128
%964 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%965 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%966 = call <32 x i32> @llvm.hexagon.V6.vsubhw(<16 x i32> %964, <16 x i32> %965)
store volatile <32 x i32> %966, <32 x i32>* @VectorPairResult, align 128
%967 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%968 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 1), align 64
%969 = call <32 x i32> @llvm.hexagon.V6.vsubuhw(<16 x i32> %967, <16 x i32> %968)
store volatile <32 x i32> %969, <32 x i32>* @VectorPairResult, align 128
%970 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%971 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%972 = call <32 x i32> @llvm.hexagon.V6.vsubw.dv(<32 x i32> %970, <32 x i32> %971)
store volatile <32 x i32> %972, <32 x i32>* @VectorPairResult, align 128
%973 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%974 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%975 = call <32 x i32> @llvm.hexagon.V6.vsubwsat.dv(<32 x i32> %973, <32 x i32> %974)
store volatile <32 x i32> %975, <32 x i32>* @VectorPairResult, align 128
%976 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%977 = call <32 x i32> @llvm.hexagon.V6.vsh(<16 x i32> %976)
store volatile <32 x i32> %977, <32 x i32>* @VectorPairResult, align 128
%978 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%979 = call <32 x i32> @llvm.hexagon.V6.vtmpyhb(<32 x i32> %978, i32 -1)
store volatile <32 x i32> %979, <32 x i32>* @VectorPairResult, align 128
%980 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%981 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 1), align 128
%982 = call <32 x i32> @llvm.hexagon.V6.vtmpyhb.acc(<32 x i32> %980, <32 x i32> %981, i32 -1)
store volatile <32 x i32> %982, <32 x i32>* @VectorPairResult, align 128
%983 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%984 = call <32 x i32> @llvm.hexagon.V6.vunpackh(<16 x i32> %983)
store volatile <32 x i32> %984, <32 x i32>* @VectorPairResult, align 128
%985 = load volatile <32 x i32>, <32 x i32>* getelementptr inbounds ([15 x <32 x i32>], [15 x <32 x i32>]* @vector_pairs, i32 0, i32 0), align 128
%986 = load volatile <16 x i32>, <16 x i32>* getelementptr inbounds ([15 x <16 x i32>], [15 x <16 x i32>]* @vectors, i32 0, i32 0), align 64
%987 = call <32 x i32> @llvm.hexagon.V6.vunpackoh(<32 x i32> %985, <16 x i32> %986)
store volatile <32 x i32> %987, <32 x i32>* @VectorPairResult, align 128
ret i32 0
}
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.and(<64 x i1>, <64 x i1>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.and.n(<64 x i1>, <64 x i1>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.not(<64 x i1>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.or(<64 x i1>, <64 x i1>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.or.n(<64 x i1>, <64 x i1>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vandvrt.acc(<64 x i1>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqb.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqh.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqw.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqb.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqh.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqw.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqb.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqh.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.veqw.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgth(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtb.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgth.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtub.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuh.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuw.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtw.and(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtb.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgth.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtub.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuh.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuw.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtw.or(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtb.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgth.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtub.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuh.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtuw.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.vgtw.xor(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.scalar2(i32) #1
; Function Attrs: nounwind readnone
declare <64 x i1> @llvm.hexagon.V6.pred.xor(<64 x i1>, <64 x i1>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vassign(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.hi(<32 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.lo(<32 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.valignb(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vand(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vandqrt.acc(<16 x i32>, <64 x i1>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlalignbi(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlalignb(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmux(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnot(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vor(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrdelta(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vror(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.lvsplatw(i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vxor(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vd0() #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddbq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddbnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubbq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubbnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrhbrndsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdealb(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdealb4w(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlutvvb(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlutvvb.oracc(<16 x i32>, <16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnavgub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackhb.sat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackeb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackob(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vroundhb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshuffb(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshuffeb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshuffob(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddhq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddhnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubhq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubhnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsh.sat(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddhsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaslh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaslhv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrhv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrwh(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrwhrndsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrwhsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavghrnd(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdealh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpybus(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpybus.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlsrhv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmaxh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vminh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyhsrs(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyhss(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyhvsrs(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyihb(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyih(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyihb.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyih.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnavgh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnormamth(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackwh.sat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackeh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackoh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpopcounth(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vroundwh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsatwh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshuffh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshufeh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshufoh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubhsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsdiffub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddubsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrhubrndsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrhubsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgubrnd(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmaxub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vminub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackhub.sat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vroundhub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsububsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsdiffh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsdiffuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vadduhsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrwuhsat(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavguh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavguhrnd(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vcl0h(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlsrh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmaxuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vminuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackwuh.sat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vroundwuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubuhsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsdiffw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vcl0w(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlsrw(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpyub(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpyubv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpyub.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpyubv.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddwq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddwnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubwq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubwnq(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsw(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vabsw.sat(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddwsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaslw(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaslwv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaslw.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrw(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrwv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrw.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgwrnd(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhb(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsat(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsusat(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhvsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhisat(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsuisat(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhb.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsat.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsusat.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhisat.acc(<16 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhsuisat.acc(<16 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vinsertwr(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlsrwv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmaxw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vminw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyewuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiwb(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiwh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiwh.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiewuh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiewh.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiewuh.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyieoh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiowh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyowh.rnd(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyowh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyowh.rnd.sacc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyowh.sacc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnavgw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnormamtw(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybus(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybusv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybv.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybus.acc(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vrmpybusv.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubwsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vassignp(<32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdealvdd(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vshuffvdd(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vswap(<64 x i1>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddb.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vshufoeb(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubb.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddubh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddh.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddhsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdmpybus.dv(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdmpybus.dv.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vlutvwh(<16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vlutvwh.oracc(<32 x i32>, <16 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpabus(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpabusv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpabuuv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpabus.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybus(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybusv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybv.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybus.acc(<32 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpybusv.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vshufoeh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsububh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubh.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubhsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsb(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpyb(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpybus(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpyb.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpybus.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackb(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackob(<32 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddubsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsububsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vadduhsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyub(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyubv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyub.acc(<32 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyubv.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubuhsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackub(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vzb(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdsaduh(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdsaduh.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyuh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyuhv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyuh.acc(<32 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyuhv.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrmpyubi(<32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrmpyubi.acc(<32 x i32>, <32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrsadubi(<32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrsadubi.acc(<32 x i32>, <32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackuh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vzh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddhw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vadduhw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddw.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vaddwsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdmpyhb.dv(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vdmpyhb.dv.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpahb(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpahb.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyh(<16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyhv(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyhus(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyhsat.acc(<32 x i32>, <16 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyhv.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpyhus.acc(<32 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrmpybusi(<32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32>, <32 x i32>, i32, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubhw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubuhw(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubw.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsubwsat.dv(<32 x i32>, <32 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vsh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpyhb(<32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vtmpyhb.acc(<32 x i32>, <32 x i32>, i32) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackh(<16 x i32>) #1
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vunpackoh(<32 x i32>, <16 x i32>) #1
attributes #0 = { nounwind "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length64b" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nounwind readnone }