Move the remaining simplify-libcalls tests to instcombine, merging most of them into a single file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182211 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 603 insertions, 0 deletions

diff --git a/test/Transforms/InstCombine/PR7357.ll b/test/Transforms/InstCombine/PR7357.ll
new file mode 100644
index 0000000000..c397c5ee05
--- /dev/null
+++ b/test/Transforms/InstCombine/PR7357.ll
@@ -0,0 +1,17 @@
+; RUN: opt < %s "-default-data-layout=e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32" -instcombine -S | FileCheck %s
+@.str1 = private constant [11 x i8] c"(){};[]&|:\00", align 4
+
+; check that simplify libcalls will not replace a call with one calling
+; convention with a new call with a different calling convention.
+
+; CHECK: define arm_aapcscc i32 @foo(i32 %argc)
+; CHECK: call arm_aapcscc  i8* @strchr
+define arm_aapcscc i32 @foo(i32 %argc) nounwind {
+bb.nph:
+  %c = call arm_aapcscc  i8* @strchr(i8* getelementptr ([11 x i8]* @.str1, i32 0,
+i32 0), i32 %argc) nounwind readonly
+  %p = ptrtoint i8* %c to i32
+  ret i32 %p
+}
+
+declare arm_aapcscc i8* @strchr(i8*, i32) nounwind readonly
diff --git a/test/Transforms/InstCombine/float-shrink-compare.ll b/test/Transforms/InstCombine/float-shrink-compare.ll
new file mode 100644
index 0000000000..0b93e841dc
--- /dev/null
+++ b/test/Transforms/InstCombine/float-shrink-compare.ll
@@ -0,0 +1,179 @@
+; RUN: opt -S -instcombine < %s | FileCheck %s
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+define i32 @test1(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @ceil(double %1) nounwind readnone
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test1
+; CHECK-NEXT: %ceilf = call float @ceilf(float %x)
+; CHECK-NEXT: fcmp oeq float %ceilf, %y
+}
+
+define i32 @test2(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @fabs(double %1) nounwind readnone
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test2
+; CHECK-NEXT: %fabsf = call float @fabsf(float %x)
+; CHECK-NEXT: fcmp oeq float %fabsf, %y
+}
+
+define i32 @test3(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @floor(double %1) nounwind readnone
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test3
+; CHECK-NEXT: %floorf = call float @floorf(float %x)
+; CHECK-NEXT: fcmp oeq float %floorf, %y
+}
+
+define i32 @test4(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @nearbyint(double %1) nounwind
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test4
+; CHECK-NEXT: %nearbyintf = call float @nearbyintf(float %x)
+; CHECK-NEXT: fcmp oeq float %nearbyintf, %y
+}
+
+define i32 @test5(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @rint(double %1) nounwind
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test5
+; CHECK-NEXT: %rintf = call float @rintf(float %x)
+; CHECK-NEXT: fcmp oeq float %rintf, %y
+}
+
+define i32 @test6(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @round(double %1) nounwind readnone
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test6
+; CHECK-NEXT: %roundf = call float @roundf(float %x)
+; CHECK-NEXT: fcmp oeq float %roundf, %y
+}
+
+define i32 @test7(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %x to double
+  %2 = call double @trunc(double %1) nounwind
+  %3 = fpext float %y to double
+  %4 = fcmp oeq double %2, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test7
+; CHECK-NEXT: %truncf = call float @truncf(float %x)
+; CHECK-NEXT: fcmp oeq float %truncf, %y
+}
+
+define i32 @test8(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @ceil(double %2) nounwind readnone
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test8
+; CHECK-NEXT: %ceilf = call float @ceilf(float %x)
+; CHECK-NEXT: fcmp oeq float %ceilf, %y
+}
+
+define i32 @test9(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @fabs(double %2) nounwind readnone
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test9
+; CHECK-NEXT: %fabsf = call float @fabsf(float %x)
+; CHECK-NEXT: fcmp oeq float %fabsf, %y
+}
+
+define i32 @test10(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @floor(double %2) nounwind readnone
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test10
+; CHECK-NEXT: %floorf = call float @floorf(float %x)
+; CHECK-NEXT: fcmp oeq float %floorf, %y
+}
+
+define i32 @test11(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @nearbyint(double %2) nounwind
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test11
+; CHECK-NEXT: %nearbyintf = call float @nearbyintf(float %x)
+; CHECK-NEXT: fcmp oeq float %nearbyintf, %y
+}
+
+define i32 @test12(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @rint(double %2) nounwind
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test12
+; CHECK-NEXT: %rintf = call float @rintf(float %x)
+; CHECK-NEXT: fcmp oeq float %rintf, %y
+}
+
+define i32 @test13(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @round(double %2) nounwind readnone
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test13
+; CHECK-NEXT: %roundf = call float @roundf(float %x)
+; CHECK-NEXT: fcmp oeq float %roundf, %y
+}
+
+define i32 @test14(float %x, float %y) nounwind uwtable {
+  %1 = fpext float %y to double
+  %2 = fpext float %x to double
+  %3 = call double @trunc(double %2) nounwind
+  %4 = fcmp oeq double %1, %3
+  %5 = zext i1 %4 to i32
+  ret i32 %5
+; CHECK: @test14
+; CHECK-NEXT: %truncf = call float @truncf(float %x)
+; CHECK-NEXT: fcmp oeq float %truncf, %y
+}
+
+declare double @fabs(double) nounwind readnone
+declare double @ceil(double) nounwind readnone
+declare double @floor(double) nounwind readnone
+declare double @nearbyint(double) nounwind readnone
+declare double @rint(double) nounwind readnone
+declare double @round(double) nounwind readnone
+declare double @trunc(double) nounwind readnone
diff --git a/test/Transforms/InstCombine/simplify-libcalls.ll b/test/Transforms/InstCombine/simplify-libcalls.ll
new file mode 100644
index 0000000000..4f3a506bee
--- /dev/null
+++ b/test/Transforms/InstCombine/simplify-libcalls.ll
@@ -0,0 +1,132 @@
+; RUN: opt -S < %s -instcombine | FileCheck %s
+
+@G = constant [3 x i8] c"%s\00"		; <[3 x i8]*> [#uses=1]
+
+declare i32 @sprintf(i8*, i8*, ...)
+
+define void @foo(i8* %P, i32* %X) {
+	call i32 (i8*, i8*, ...)* @sprintf( i8* %P, i8* getelementptr ([3 x i8]* @G, i32 0, i32 0), i32* %X )		; <i32>:1 [#uses=0]
+	ret void
+}
+
+; PR1307
+@str = internal constant [5 x i8] c"foog\00"
+@str1 = internal constant [8 x i8] c"blahhh!\00"
+@str2 = internal constant [5 x i8] c"Ponk\00"
+
+define i8* @test1() {
+        %tmp3 = tail call i8* @strchr( i8* getelementptr ([5 x i8]* @str, i32 0, i32 2), i32 103 )              ; <i8*> [#uses=1]
+        ret i8* %tmp3
+
+; CHECK: @test1
+; CHECK: ret i8* getelementptr inbounds ([5 x i8]* @str, i32 0, i64 3)
+}
+
+declare i8* @strchr(i8*, i32)
+
+define i8* @test2() {
+        %tmp3 = tail call i8* @strchr( i8* getelementptr ([8 x i8]* @str1, i32 0, i32 2), i32 0 )               ; <i8*> [#uses=1]
+        ret i8* %tmp3
+
+; CHECK: @test2
+; CHECK: ret i8* getelementptr inbounds ([8 x i8]* @str1, i32 0, i64 7)
+}
+
+define i8* @test3() {
+entry:
+        %tmp3 = tail call i8* @strchr( i8* getelementptr ([5 x i8]* @str2, i32 0, i32 1), i32 80 )              ; <i8*> [#uses=1]
+        ret i8* %tmp3
+
+; CHECK: @test3
+; CHECK: ret i8* null
+}
+
+@_2E_str = external constant [5 x i8]		; <[5 x i8]*> [#uses=1]
+
+declare i32 @memcmp(i8*, i8*, i32) nounwind readonly
+
+define i1 @PR2341(i8** %start_addr) {
+entry:
+	%tmp4 = load i8** %start_addr, align 4		; <i8*> [#uses=1]
+	%tmp5 = call i32 @memcmp( i8* %tmp4, i8* getelementptr ([5 x i8]* @_2E_str, i32 0, i32 0), i32 4 ) nounwind readonly 		; <i32> [#uses=1]
+	%tmp6 = icmp eq i32 %tmp5, 0		; <i1> [#uses=1]
+	ret i1 %tmp6
+
+; CHECK: @PR2341
+; CHECK: i32
+}
+
+define i32 @PR4284() nounwind {
+entry:
+	%c0 = alloca i8, align 1		; <i8*> [#uses=2]
+	%c2 = alloca i8, align 1		; <i8*> [#uses=2]
+	store i8 64, i8* %c0
+	store i8 -127, i8* %c2
+	%call = call i32 @memcmp(i8* %c0, i8* %c2, i32 1)		; <i32> [#uses=1]
+	ret i32 %call
+
+; CHECK: @PR4284
+; CHECK: ret i32 -65
+}
+
+%struct.__sFILE = type { i8*, i32, i32, i16, i16, %struct.__sbuf, i32, i8*, i32 (i8*)*, i32 (i8*, i8*, i32)*, i64 (i8*, i64, i32)*, i32 (i8*, i8*, i32)*, %struct.__sbuf, i8*, i32, [3 x i8], [1 x i8], %struct.__sbuf, i32, i64, %struct.pthread_mutex*, %struct.pthread*, i32, i32, %union.anon }
+%struct.__sbuf = type { i8*, i32, [4 x i8] }
+%struct.pthread = type opaque
+%struct.pthread_mutex = type opaque
+%union.anon = type { i64, [120 x i8] }
+@.str13 = external constant [2 x i8]		; <[2 x i8]*> [#uses=1]
+@.str14 = external constant [2 x i8]		; <[2 x i8]*> [#uses=1]
+
+define i32 @PR4641(i32 %argc, i8** %argv) nounwind {
+entry:
+	call void @exit(i32 0) nounwind
+	%cond392 = select i1 undef, i8* getelementptr ([2 x i8]* @.str13, i32 0, i32 0), i8* getelementptr ([2 x i8]* @.str14, i32 0, i32 0)		; <i8*> [#uses=1]
+	%call393 = call %struct.__sFILE* @fopen(i8* undef, i8* %cond392) nounwind		; <%struct.__sFILE*> [#uses=0]
+	unreachable
+}
+
+declare %struct.__sFILE* @fopen(i8*, i8*)
+
+declare void @exit(i32)
+
+define i32 @PR4645() {
+entry:
+	br label %if.then
+
+lor.lhs.false:		; preds = %while.body
+	br i1 undef, label %if.then, label %for.cond
+
+if.then:		; preds = %lor.lhs.false, %while.body
+	call void @exit(i32 1)
+	br label %for.cond
+
+for.cond:		; preds = %for.end, %if.then, %lor.lhs.false
+	%j.0 = phi i32 [ %inc47, %for.end ], [ 0, %if.then ], [ 0, %lor.lhs.false ]		; <i32> [#uses=1]
+	unreachable
+
+for.end:		; preds = %for.cond20
+	%inc47 = add i32 %j.0, 1		; <i32> [#uses=1]
+	br label %for.cond
+}
+
+@h = constant [2 x i8] c"h\00"		; <[2 x i8]*> [#uses=1]
+@hel = constant [4 x i8] c"hel\00"		; <[4 x i8]*> [#uses=1]
+@hello_u = constant [8 x i8] c"hello_u\00"		; <[8 x i8]*> [#uses=1]
+
+define i32 @MemCpy() {
+  %h_p = getelementptr [2 x i8]* @h, i32 0, i32 0
+  %hel_p = getelementptr [4 x i8]* @hel, i32 0, i32 0
+  %hello_u_p = getelementptr [8 x i8]* @hello_u, i32 0, i32 0
+  %target = alloca [1024 x i8]
+  %target_p = getelementptr [1024 x i8]* %target, i32 0, i32 0
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %target_p, i8* %h_p, i32 2, i32 2, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %target_p, i8* %hel_p, i32 4, i32 4, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %target_p, i8* %hello_u_p, i32 8, i32 8, i1 false)
+  ret i32 0
+
+; CHECK: @MemCpy
+; CHECK-NOT: llvm.memcpy
+; CHECK: ret i32 0
+}
+
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
diff --git a/test/Transforms/InstCombine/win-math.ll b/test/Transforms/InstCombine/win-math.ll
new file mode 100644
index 0000000000..367e5b8072
--- /dev/null
+++ b/test/Transforms/InstCombine/win-math.ll
@@ -0,0 +1,275 @@
+; RUN: opt -O2 -S -mtriple=i386-pc-win32 < %s | FileCheck %s -check-prefix=WIN32
+; RUN: opt -O2 -S -mtriple=x86_64-pc-win32 < %s | FileCheck %s -check-prefix=WIN64
+; RUN: opt -O2 -S -mtriple=i386-pc-mingw32 < %s | FileCheck %s -check-prefix=MINGW32
+; RUN: opt -O2 -S -mtriple=x86_64-pc-mingw32 < %s | FileCheck %s -check-prefix=MINGW64
+
+; x86 win32 msvcrt does not provide entry points for single-precision libm.
+; x86-64 win32 msvcrt does (except for fabsf)
+; msvcrt does not provide C99 math, but mingw32 does.
+
+declare double @acos(double %x)
+define float @float_acos(float %x) nounwind readnone {
+; WIN32: @float_acos
+; WIN32-NOT: float @acosf
+; WIN32: double @acos
+    %1 = fpext float %x to double
+    %2 = call double @acos(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @asin(double %x)
+define float @float_asin(float %x) nounwind readnone {
+; WIN32: @float_asin
+; WIN32-NOT: float @asinf
+; WIN32: double @asin
+    %1 = fpext float %x to double
+    %2 = call double @asin(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @atan(double %x)
+define float @float_atan(float %x) nounwind readnone {
+; WIN32: @float_atan
+; WIN32-NOT: float @atanf
+; WIN32: double @atan
+    %1 = fpext float %x to double
+    %2 = call double @atan(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @atan2(double %x, double %y)
+define float @float_atan2(float %x, float %y) nounwind readnone {
+; WIN32: @float_atan2
+; WIN32-NOT: float @atan2f
+; WIN32: double @atan2
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @atan2(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @ceil(double %x)
+define float @float_ceil(float %x) nounwind readnone {
+; WIN32: @float_ceil
+; WIN32-NOT: float @ceilf
+; WIN32: double @ceil
+; WIN64: @float_ceil
+; WIN64: float @ceilf
+; WIN64-NOT: double @ceil
+; MINGW32: @float_ceil
+; MINGW32: float @ceilf
+; MINGW32-NOT: double @ceil
+; MINGW64: @float_ceil
+; MINGW64: float @ceilf
+; MINGW64-NOT: double @ceil
+    %1 = fpext float %x to double
+    %2 = call double @ceil(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @_copysign(double %x)
+define float @float_copysign(float %x) nounwind readnone {
+; WIN32: @float_copysign
+; WIN32-NOT: float @copysignf
+; WIN32-NOT: float @_copysignf
+; WIN32: double @_copysign
+    %1 = fpext float %x to double
+    %2 = call double @_copysign(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @cos(double %x)
+define float @float_cos(float %x) nounwind readnone {
+; WIN32: @float_cos
+; WIN32-NOT: float @cosf
+; WIN32: double @cos
+    %1 = fpext float %x to double
+    %2 = call double @cos(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @cosh(double %x)
+define float @float_cosh(float %x) nounwind readnone {
+; WIN32: @float_cosh
+; WIN32-NOT: float @coshf
+; WIN32: double @cosh
+    %1 = fpext float %x to double
+    %2 = call double @cosh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @exp(double %x, double %y)
+define float @float_exp(float %x, float %y) nounwind readnone {
+; WIN32: @float_exp
+; WIN32-NOT: float @expf
+; WIN32: double @exp
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @exp(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @fabs(double %x, double %y)
+define float @float_fabs(float %x, float %y) nounwind readnone {
+; WIN32: @float_fabs
+; WIN32-NOT: float @fabsf
+; WIN32: double @fabs
+; WIN64: @float_fabs
+; WIN64-NOT: float @fabsf
+; WIN64: double @fabs
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @fabs(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @floor(double %x)
+define float @float_floor(float %x) nounwind readnone {
+; WIN32: @float_floor
+; WIN32-NOT: float @floorf
+; WIN32: double @floor
+; WIN64: @float_floor
+; WIN64: float @floorf
+; WIN64-NOT: double @floor
+; MINGW32: @float_floor
+; MINGW32: float @floorf
+; MINGW32-NOT: double @floor
+; MINGW64: @float_floor
+; MINGW64: float @floorf
+; MINGW64-NOT: double @floor
+    %1 = fpext float %x to double
+    %2 = call double @floor(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @fmod(double %x, double %y)
+define float @float_fmod(float %x, float %y) nounwind readnone {
+; WIN32: @float_fmod
+; WIN32-NOT: float @fmodf
+; WIN32: double @fmod
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @fmod(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @log(double %x)
+define float @float_log(float %x) nounwind readnone {
+; WIN32: @float_log
+; WIN32-NOT: float @logf
+; WIN32: double @log
+    %1 = fpext float %x to double
+    %2 = call double @log(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @pow(double %x, double %y)
+define float @float_pow(float %x, float %y) nounwind readnone {
+; WIN32: @float_pow
+; WIN32-NOT: float @powf
+; WIN32: double @pow
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @pow(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @sin(double %x)
+define float @float_sin(float %x) nounwind readnone {
+; WIN32: @float_sin
+; WIN32-NOT: float @sinf
+; WIN32: double @sin
+    %1 = fpext float %x to double
+    %2 = call double @sin(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @sinh(double %x)
+define float @float_sinh(float %x) nounwind readnone {
+; WIN32: @float_sinh
+; WIN32-NOT: float @sinhf
+; WIN32: double @sinh
+    %1 = fpext float %x to double
+    %2 = call double @sinh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @sqrt(double %x)
+define float @float_sqrt(float %x) nounwind readnone {
+; WIN32: @float_sqrt
+; WIN32-NOT: float @sqrtf
+; WIN32: double @sqrt
+; WIN64: @float_sqrt
+; WIN64: float @sqrtf
+; WIN64-NOT: double @sqrt
+; MINGW32: @float_sqrt
+; MINGW32: float @sqrtf
+; MINGW32-NOT: double @sqrt
+; MINGW64: @float_sqrt
+; MINGW64: float @sqrtf
+; MINGW64-NOT: double @sqrt
+    %1 = fpext float %x to double
+    %2 = call double @sqrt(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @tan(double %x)
+define float @float_tan(float %x) nounwind readnone {
+; WIN32: @float_tan
+; WIN32-NOT: float @tanf
+; WIN32: double @tan
+    %1 = fpext float %x to double
+    %2 = call double @tan(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @tanh(double %x)
+define float @float_tanh(float %x) nounwind readnone {
+; WIN32: @float_tanh
+; WIN32-NOT: float @tanhf
+; WIN32: double @tanh
+    %1 = fpext float %x to double
+    %2 = call double @tanh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+; win32 does not have round; mingw32 does
+declare double @round(double %x)
+define float @float_round(float %x) nounwind readnone {
+; WIN32: @float_round
+; WIN32-NOT: float @roundf
+; WIN32: double @round
+; WIN64: @float_round
+; WIN64-NOT: float @roundf
+; WIN64: double @round
+; MINGW32: @float_round
+; MINGW32: float @roundf
+; MINGW32-NOT: double @round
+; MINGW64: @float_round
+; MINGW64: float @roundf
+; MINGW64-NOT: double @round
+    %1 = fpext float %x to double
+    %2 = call double @round(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+