Subzero: Add fabs intrinsic support.

tests_lit/reader_tests/nacl-other-intrinsics.ll

 ; This tests parsing NaCl intrinsics not related to atomic operations.
 
 ; RUN: %p2i -i %s --insts | FileCheck %s
 ; RUN: %if --need=allow_disable_ir_gen --command \
 ; RUN:   %p2i -i %s --args -notranslate -timing -no-ir-gen \
 ; RUN: | %if --need=allow_disable_ir_gen --command \
 ; RUN:   FileCheck --check-prefix=NOIR %s
 
 declare i8* @llvm.nacl.read.tp()
 declare void @llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)
 declare void @llvm.memmove.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)
 declare void @llvm.memset.p0i8.i32(i8*, i8, i32, i32, i1)
 declare void @llvm.nacl.longjmp(i8*, i32)
 declare i32 @llvm.nacl.setjmp(i8*)
 declare float @llvm.sqrt.f32(float)
 declare double @llvm.sqrt.f64(double)
+declare float @llvm.fabs.f32(float)
+declare double @llvm.fabs.f64(double)
+declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
 declare void @llvm.trap()
 declare i16 @llvm.bswap.i16(i16)
 declare i32 @llvm.bswap.i32(i32)
 declare i64 @llvm.bswap.i64(i64)
 declare i32 @llvm.ctlz.i32(i32, i1)
 declare i64 @llvm.ctlz.i64(i64, i1)
 declare i32 @llvm.cttz.i32(i32, i1)
 declare i64 @llvm.cttz.i64(i64, i1)
 declare i32 @llvm.ctpop.i32(i32)
 declare i64 @llvm.ctpop.i64(i64)
@@ skipping 115 matching lines @@
 
 ; CHECK-NEXT: define double @test_sqrt_double(double %x, i32 %iptr) {
 ; CHECK-NEXT: entry:
 ; CHECK-NEXT:   %r = call double @llvm.sqrt.f64(double %x)
 ; CHECK-NEXT:   %r2 = call double @llvm.sqrt.f64(double %r)
 ; CHECK-NEXT:   %r3 = call double @llvm.sqrt.f64(double -0.000000e+00)
 ; CHECK-NEXT:   %r4 = fadd double %r2, %r3
 ; CHECK-NEXT:   ret double %r4
 ; CHECK-NEXT: }
 
+define float @test_fabs_float(float %x, i32 %iptr) {

[jvoung (off chromium), 2015/03/19 16:37:53]

similar -- iptr not needed?

[Jim Stichnoth, 2015/03/19 18:35:15]

Done.

+entry:
+  %r = call float @llvm.fabs.f32(float %x)
+  %r2 = call float @llvm.fabs.f32(float %r)
+  %r3 = call float @llvm.fabs.f32(float -0.0)
+  %r4 = fadd float %r2, %r3
+  ret float %r4
+}
+
+; CHECK-NEXT: define float @test_fabs_float(float %x, i32 %iptr) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT:   %r = call float @llvm.fabs.f32(float %x)
+; CHECK-NEXT:   %r2 = call float @llvm.fabs.f32(float %r)
+; CHECK-NEXT:   %r3 = call float @llvm.fabs.f32(float -0.000000e+00)
+; CHECK-NEXT:   %r4 = fadd float %r2, %r3
+; CHECK-NEXT:   ret float %r4
+; CHECK-NEXT: }
+
+define double @test_fabs_double(double %x, i32 %iptr) {
+entry:
+  %r = call double @llvm.fabs.f64(double %x)
+  %r2 = call double @llvm.fabs.f64(double %r)
+  %r3 = call double @llvm.fabs.f64(double -0.0)
+  %r4 = fadd double %r2, %r3
+  ret double %r4
+}
+
+; CHECK-NEXT: define double @test_fabs_double(double %x, i32 %iptr) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT:   %r = call double @llvm.fabs.f64(double %x)
+; CHECK-NEXT:   %r2 = call double @llvm.fabs.f64(double %r)
+; CHECK-NEXT:   %r3 = call double @llvm.fabs.f64(double -0.000000e+00)
+; CHECK-NEXT:   %r4 = fadd double %r2, %r3
+; CHECK-NEXT:   ret double %r4
+; CHECK-NEXT: }
+
+define <4 x float> @test_fabs_v4f32(<4 x float> %x, i32 %iptr) {
+entry:
+  %r = call <4 x float> @llvm.fabs.v4f32(<4 x float> %x)
+  %r2 = call <4 x float> @llvm.fabs.v4f32(<4 x float> %r)
+  %r3 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
+  %r4 = fadd <4 x float> %r2, %r3
+  ret <4 x float> %r4
+}
+
+; CHECK-NEXT: define <4 x float> @test_fabs_v4f32(<4 x float> %x, i32 %iptr) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT:   %r = call <4 x float> @llvm.fabs.v4f32(<4 x float> %x)
+; CHECK-NEXT:   %r2 = call <4 x float> @llvm.fabs.v4f32(<4 x float> %r)
+; CHECK-NEXT:   %r3 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
+; CHECK-NEXT:   %r4 = fadd <4 x float> %r2, %r3
+; CHECK-NEXT:   ret <4 x float> %r4
+; CHECK-NEXT: }
+
 define i32 @test_trap(i32 %br) {
 entry:
   %r1 = icmp eq i32 %br, 0
   br i1 %r1, label %Zero, label %NonZero
 Zero:
   call void @llvm.trap()
   unreachable
 NonZero:
   ret i32 1
 }
@@ skipping 175 matching lines @@
 ; CHECK-NEXT:   %sp3 = call i32 @llvm.stacksave()
 ; CHECK-NEXT:   %tmp3 = alloca i8, i32 %x_4, align 4
 ; CHECK-NEXT:   store i32 %y, i32* %tmp1, align 1
 ; CHECK-NEXT:   store i32 %x, i32* %tmp2, align 1
 ; CHECK-NEXT:   store i32 %x, i32* %tmp3, align 1
 ; CHECK-NEXT:   call void @llvm.stackrestore(i32 %sp1)
 ; CHECK-NEXT:   ret void
 ; CHECK-NEXT: }
 
 ; NOIR: Total across all functions