Make some existing optimizations that would only trigger on scalars

also apply to vectors.  This allows us to compile this:

#include <emmintrin.h>
__m128i a(__m128 a, __m128 b) { return a==a & b==b; }
__m128i b(__m128 a, __m128 b) { return a!=a | b!=b; }

to:

_a:
	cmpordps	%xmm1, %xmm0
	ret
_b:
	cmpunordps	%xmm1, %xmm0
	ret

with clang instead of to a ton of horrible code.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76863 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 31 insertions, 2 deletions

diff --git a/test/Transforms/InstCombine/vector-casts.ll b/test/Transforms/InstCombine/vector-casts.ll
index c6d1eaa1ff..e617185206 100644
--- a/test/Transforms/InstCombine/vector-casts.ll
+++ b/test/Transforms/InstCombine/vector-casts.ll
@@ -5,7 +5,7 @@ define <2 x i1> @test1(<2 x i64> %a) {
   %t = trunc <2 x i64> %a to <2 x i1>
   ret <2 x i1> %t
 
-; CHECK: define <2 x i1> @test1
+; CHECK: @test1
 ; CHECK:   and <2 x i64> %a, <i64 1, i64 1>
 ; CHECK:   icmp ne <2 x i64> %tmp, zeroinitializer
 }
@@ -16,7 +16,36 @@ define <2 x i64> @test2(<2 x i64> %a) {
   %t = ashr <2 x i64> %b, <i64 1, i64 1>
   ret <2 x i64> %t
 
-; CHECK: define <2 x i64> @test2
+; CHECK: @test2
 ; CHECK:   and <2 x i64> %a, <i64 65535, i64 65535>
 ; CHECK:   lshr <2 x i64> %b, <i64 1, i64 1>
 }
+
+
+
+define <2 x i64> @test3(<4 x float> %a, <4 x float> %b) nounwind readnone {
+entry:
+	%cmp = fcmp ord <4 x float> %a, zeroinitializer	
+	%sext = sext <4 x i1> %cmp to <4 x i32>	
+	%cmp4 = fcmp ord <4 x float> %b, zeroinitializer
+	%sext5 = sext <4 x i1> %cmp4 to <4 x i32>
+	%and = and <4 x i32> %sext, %sext5
+	%conv = bitcast <4 x i32> %and to <2 x i64>
+	ret <2 x i64> %conv
+        
+; CHECK: @test3
+; CHECK:   fcmp ord <4 x float> %a, %b
+}
+
+define <2 x i64> @test4(<4 x float> %a, <4 x float> %b) nounwind readnone {
+entry:
+	%cmp = fcmp uno <4 x float> %a, zeroinitializer
+	%sext = sext <4 x i1> %cmp to <4 x i32>
+	%cmp4 = fcmp uno <4 x float> %b, zeroinitializer
+	%sext5 = sext <4 x i1> %cmp4 to <4 x i32>
+	%or = or <4 x i32> %sext, %sext5
+	%conv = bitcast <4 x i32> %or to <2 x i64>
+	ret <2 x i64> %conv
+; CHECK: @test4
+; CHECK:   fcmp uno <4 x float> %a, %b
+}