diff options
| author | James Molloy <james.molloy@arm.com> | 2017-02-13 12:32:47 +0000 |
|---|---|---|
| committer | James Molloy <james.molloy@arm.com> | 2017-02-13 12:32:47 +0000 |
| commit | 9b264f7915075aee21b4f57d80ce72ab29865144 (patch) | |
| tree | 50e6aa934c2024a5a31a5b7108aabf14d8bd433d /test/CodeGen/ARM/fp16-promote.ll | |
| parent | d9480271fe4af7388f3d0ed7fd56cb06d9326737 (diff) | |
[ARM] Use VCMP, not VCMPE, for floating point equality comparisons
When generating a floating point comparison we currently unconditionally
generate VCMPE. This has the sideeffect of setting the cumulative Invalid
bit in FPSCR if any of the operands are QNaN.
It is expected that use of a relational predicate on a QNaN value should
raise Invalid. Quoting from the C standard:
The relational and equality operators support the usual mathematical
relationships between numeric values. For any ordered pair of numeric
values exactly one of relationships the less, greater, equal and is true.
Relational operators may raise the floating-point exception when argument
values are NaNs.
The standard doesn't explicitly state the expectation for equality operators,
but the implication and obvious expectation is that equality operators
should not raise Invalid on a QNaN input, as those predicates are wholly
defined on unordered inputs (to return not equal).
Therefore, add a new operand to ARMISD::FPCMP and FPCMPZ indicating if
QNaN should raise Invalid, and pipe that through to TableGen.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@294945 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/ARM/fp16-promote.ll')
| -rw-r--r-- | test/CodeGen/ARM/fp16-promote.ll | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/test/CodeGen/ARM/fp16-promote.ll b/test/CodeGen/ARM/fp16-promote.ll index c9dafa8dfff..2f7dff70b9b 100644 --- a/test/CodeGen/ARM/fp16-promote.ll +++ b/test/CodeGen/ARM/fp16-promote.ll @@ -161,14 +161,14 @@ define void @test_select(half* %p, half* %q, i1 zeroext %c) #0 { ret void } -; Test only two variants of fcmp. These get translated to f32 vcmpe +; Test only two variants of fcmp. These get translated to f32 vcmp ; instructions anyway. ; CHECK-ALL-LABEL: test_fcmp_une: ; CHECK-FP16: vcvtb.f32.f16 ; CHECK-FP16: vcvtb.f32.f16 ; CHECK-LIBCALL: bl __aeabi_h2f ; CHECK-LIBCALL: bl __aeabi_h2f -; CHECK-VFP: vcmpe.f32 +; CHECK-VFP: vcmp.f32 ; CHECK-NOVFP: bl __aeabi_fcmpeq ; CHECK-FP16: vmrs APSR_nzcv, fpscr ; CHECK-ALL: movw{{ne|eq}} @@ -184,7 +184,7 @@ define i1 @test_fcmp_une(half* %p, half* %q) #0 { ; CHECK-FP16: vcvtb.f32.f16 ; CHECK-LIBCALL: bl __aeabi_h2f ; CHECK-LIBCALL: bl __aeabi_h2f -; CHECK-VFP: vcmpe.f32 +; CHECK-VFP: vcmp.f32 ; CHECK-NOVFP: bl __aeabi_fcmpeq ; CHECK-FP16: vmrs APSR_nzcv, fpscr ; CHECK-LIBCALL: movw{{ne|eq}} |