Remove sqrt from the volatile list

The LLVM IR spec now explicitly says that this intrinsic is required
to be rounded correctly. That means that without fasthmath flag
(which we do not set here), this intrinsic must have the bitwise
correctly rounded answer and should thus not differ between compile
and runtime. If there is still a case where it does differ, that is
likely some other underlying bug that we should fix instead.

Before:
```
julia> f() = sqrt(2)
f (generic function with 1 method)

julia> @code_typed f()
CodeInfo(
1 ─ %1 = Base.Math.sqrt_llvm(2.0)::Float64
└──      return %1
) => Float64
```

After:
```
julia> @code_typed f()
CodeInfo(
1 ─     return 1.4142135623730951
) => Float64
```

Author: Keno

base/compiler/optimize.jl

@@ -618,7 +618,6 @@ function is_pure_intrinsic_infer(f::IntrinsicFunction)
              f === Intrinsics.pointerset || # this one is never effect-free
              f === Intrinsics.llvmcall ||   # this one is never effect-free
              f === Intrinsics.arraylen ||   # this one is volatile
-             f === Intrinsics.sqrt_llvm ||  # this one may differ at runtime (by a few ulps)
              f === Intrinsics.sqrt_llvm_fast ||  # this one may differ at runtime (by a few ulps)
              f === Intrinsics.have_fma ||  # this one depends on the runtime environment
              f === Intrinsics.cglobal)  # cglobal lookup answer changes at runtime

test/compiler/inline.jl

@@ -891,3 +891,7 @@ let
     ftest(a) = (fcond(a, nothing); a)
     @test fully_eliminated(ftest, Tuple{Bool})
 end
+
+# sqrt not considered volatile
+f_sqrt() = sqrt(2)
+@test fully_eliminated(f_sqrt, Tuple{})

test/math.jl

@@ -1317,3 +1317,72 @@ end
         @test func(-1.9369631f13, 2.1513551f-7, -1.7354427f-24) == -4.1670958f6
     end
 end
+
+# Test that sqrt behaves correctly and doesn't exhibit fp80 double rounding.
+# This happened on old glibc versions.
+# Test case from https://sourceware.org/bugzilla/show_bug.cgi?id=14032.
+@testset "sqrt double rounding" begin
+    testdata = [
+        (0x1.fffffffffffffp+1023, 0x1.fffffffffffffp+511),
+        (0x1.ffffffffffffbp+1023, 0x1.ffffffffffffdp+511),
+        (0x1.ffffffffffff7p+1023, 0x1.ffffffffffffbp+511),
+        (0x1.ffffffffffff3p+1023, 0x1.ffffffffffff9p+511),
+        (0x1.fffffffffffefp+1023, 0x1.ffffffffffff7p+511),
+        (0x1.fffffffffffebp+1023, 0x1.ffffffffffff5p+511),
+        (0x1.fffffffffffe7p+1023, 0x1.ffffffffffff3p+511),
+        (0x1.fffffffffffe3p+1023, 0x1.ffffffffffff1p+511),
+        (0x1.fffffffffffdfp+1023, 0x1.fffffffffffefp+511),
+        (0x1.fffffffffffdbp+1023, 0x1.fffffffffffedp+511),
+        (0x1.fffffffffffd7p+1023, 0x1.fffffffffffebp+511),
+        (0x1.0000000000003p-1022, 0x1.0000000000001p-511),
+        (0x1.0000000000007p-1022, 0x1.0000000000003p-511),
+        (0x1.000000000000bp-1022, 0x1.0000000000005p-511),
+        (0x1.000000000000fp-1022, 0x1.0000000000007p-511),
+        (0x1.0000000000013p-1022, 0x1.0000000000009p-511),
+        (0x1.0000000000017p-1022, 0x1.000000000000bp-511),
+        (0x1.000000000001bp-1022, 0x1.000000000000dp-511),
+        (0x1.000000000001fp-1022, 0x1.000000000000fp-511),
+        (0x1.0000000000023p-1022, 0x1.0000000000011p-511),
+        (0x1.0000000000027p-1022, 0x1.0000000000013p-511),
+        (0x1.000000000002bp-1022, 0x1.0000000000015p-511),
+        (0x1.000000000002fp-1022, 0x1.0000000000017p-511),
+        (0x1.0000000000033p-1022, 0x1.0000000000019p-511),
+        (0x1.0000000000037p-1022, 0x1.000000000001bp-511),
+        (0x1.7167bc36eaa3bp+6, 0x1.3384c7db650cdp+3),
+        (0x1.7570994273ad7p+6, 0x1.353186e89b8ffp+3),
+        (0x1.7dae969442fe6p+6, 0x1.389640fb18b75p+3),
+        (0x1.7f8444fcf67e5p+6, 0x1.395659e94669fp+3),
+        (0x1.8364650e63a54p+6, 0x1.3aea9efe1a3d7p+3),
+        (0x1.85bedd274edd8p+6, 0x1.3bdf20c867057p+3),
+        (0x1.8609cf496ab77p+6, 0x1.3bfd7e14b5eabp+3),
+        (0x1.873849c70a375p+6, 0x1.3c77ed341d27fp+3),
+        (0x1.8919c962cbaaep+6, 0x1.3d3a7113ee82fp+3),
+        (0x1.8de4493e22dc6p+6, 0x1.3f27d448220c3p+3),
+        (0x1.924829a17a288p+6, 0x1.40e9552eec28fp+3),
+        (0x1.92702cd992f12p+6, 0x1.40f94a6fdfddfp+3),
+        (0x1.92b763a8311fdp+6, 0x1.4115af614695fp+3),
+        (0x1.947da013c7293p+6, 0x1.41ca91102940fp+3),
+        (0x1.9536091c494d2p+6, 0x1.4213e334c77adp+3),
+        (0x1.61b04c6p-1019, 0x1.a98b88f18b46dp-510),
+        (0x1.93789f1p-1018, 0x1.4162ae43d5821p-509),
+        (0x1.a1989b4p-1018, 0x1.46f6736eb44bbp-509),
+        (0x1.f93bc9p-1018, 0x1.67a36ec403bafp-509),
+        (0x1.2f675e3p-1017, 0x1.8a22ab6dcfee1p-509),
+        (0x1.a158508p-1017, 0x1.ce418a96cf589p-509),
+        (0x1.cd31f078p-1017, 0x1.e5ef1c65dccebp-509),
+        (0x1.33b43b08p-1016, 0x1.18a9f607e1701p-508),
+        (0x1.6e66a858p-1016, 0x1.324402a00b45fp-508),
+        (0x1.8661cbf8p-1016, 0x1.3c212046bfdffp-508),
+        (0x1.bbb221b4p-1016, 0x1.510681b939931p-508),
+        (0x1.c4942f3cp-1016, 0x1.5461e59227ab5p-508),
+        (0x1.dbb258c8p-1016, 0x1.5cf7b0f78d3afp-508),
+        (0x1.57103ea4p-1015, 0x1.a31ab946d340bp-508),
+        (0x1.9b294f88p-1015, 0x1.cad197e28e85bp-508),
+        (0x1.0000000000001p+0, 0x1p+0),
+        (0x1.fffffffffffffp-1, 0x1.fffffffffffffp-1),
+    ]
+    for (x,y) in testdata
+        @test sqrt(x) === y
+        @test Base.invokelatest((@eval ()->sqrt($x))) == y
+    end
+end