diff --git a/test/CodeGen/ARM/mult-alt-generic-arm.ll b/test/CodeGen/ARM/mult-alt-generic-arm.ll
new file mode 100644
index 0000000..a8104db
--- /dev/null
+++ b/test/CodeGen/ARM/mult-alt-generic-arm.ll
@@ -0,0 +1,323 @@
+; RUN: llc < %s -march=arm
+; ModuleID = 'mult-alt-generic.c'
+target datalayout = "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"
+target triple = "arm"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define arm_aapcscc void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define arm_aapcscc void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define arm_aapcscc void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define arm_aapcscc void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define arm_aapcscc void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+; No lowering support.
+;  %4 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+001) nounwind
+;  store i32 %4, i32* %out0, align 4
+;  %5 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+000) nounwind
+;  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define arm_aapcscc void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define arm_aapcscc void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define arm_aapcscc void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define arm_aapcscc void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+; No lowering support.
+;  %4 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+001) nounwind
+;  store i32 %4, i32* %out0, align 4
+;  %5 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+000) nounwind
+;  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define arm_aapcscc void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/MSP430/mult-alt-generic-msp430.ll b/test/CodeGen/MSP430/mult-alt-generic-msp430.ll
new file mode 100644
index 0000000..342afed
--- /dev/null
+++ b/test/CodeGen/MSP430/mult-alt-generic-msp430.ll
@@ -0,0 +1,323 @@
+; RUN: llc < %s -march=msp430
+; ModuleID = 'mult-alt-generic.c'
+target datalayout = "e-p:16:16:16-i8:8:8-i16:16:16-i32:16:32-n8:16"
+target triple = "msp430"
+
+@mout0 = common global i16 0, align 2
+@min1 = common global i16 0, align 2
+@marray = common global [2 x i16] zeroinitializer, align 2
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m"(i16* @mout0, i16* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %index = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %index, align 2
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,<r"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* %in1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r,r<"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,>r"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* %in1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r,r>"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,r"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,i"(i16 1) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,n"(i16 1) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,imr"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* @min1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r,imr"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  %2 = call i16 asm "foo $1,$0", "=r,imr"(i16 1) nounwind
+  store i16 %2, i16* %out0, align 2
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,X"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* @min1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r,X"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  %2 = call i16 asm "foo $1,$0", "=r,X"(i16 1) nounwind
+  store i16 %2, i16* %out0, align 2
+  %3 = call i16 asm "foo $1,$0", "=r,X"(i16* getelementptr inbounds ([2 x i16]* @marray, i32 0, i32 0)) nounwind
+  store i16 %3, i16* %out0, align 2
+; No lowering support.
+;  %4 = call i16 asm "foo $1,$0", "=r,X"(double 1.000000e+001) nounwind
+;  store i16 %4, i16* %out0, align 2
+;  %5 = call i16 asm "foo $1,$0", "=r,X"(double 1.000000e+000) nounwind
+;  store i16 %5, i16* %out0, align 2
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r,r"(i16* getelementptr inbounds ([2 x i16]* @marray, i32 0, i32 0)) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i16* @min1, align 2
+  call void asm "foo $1,$0", "=*m|r,m|r"(i16* @mout0, i16 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %index = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %index, align 2
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|<r"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* %in1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r|r,r|r<"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|>r"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* %in1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r|r,r|r>"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|m"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|i"(i16 1) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|n"(i16 1) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|imr"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* @min1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r|r,r|imr"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  %2 = call i16 asm "foo $1,$0", "=r|r,r|imr"(i16 1) nounwind
+  store i16 %2, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  %in1 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  store i16 1, i16* %in1, align 2
+  %tmp = load i16* %in1, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|X"(i16 %tmp) nounwind
+  store i16 %0, i16* %out0, align 2
+  %tmp1 = load i16* @min1, align 2
+  %1 = call i16 asm "foo $1,$0", "=r|r,r|X"(i16 %tmp1) nounwind
+  store i16 %1, i16* %out0, align 2
+  %2 = call i16 asm "foo $1,$0", "=r|r,r|X"(i16 1) nounwind
+  store i16 %2, i16* %out0, align 2
+  %3 = call i16 asm "foo $1,$0", "=r|r,r|X"(i16* getelementptr inbounds ([2 x i16]* @marray, i32 0, i32 0)) nounwind
+  store i16 %3, i16* %out0, align 2
+; No lowering support.
+;  %4 = call i16 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+001) nounwind
+;  store i16 %4, i16* %out0, align 2
+;  %5 = call i16 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+000) nounwind
+;  store i16 %5, i16* %out0, align 2
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i16, align 2
+  store i16 0, i16* %out0, align 2
+  %0 = call i16 asm "foo $1,$0", "=r|r,r|r"(i16* getelementptr inbounds ([2 x i16]* @marray, i32 0, i32 0)) nounwind
+  store i16 %0, i16* %out0, align 2
+  ret void
+}
diff --git a/test/CodeGen/PowerPC/mult-alt-generic-powerpc.ll b/test/CodeGen/PowerPC/mult-alt-generic-powerpc.ll
new file mode 100644
index 0000000..659cdf7
--- /dev/null
+++ b/test/CodeGen/PowerPC/mult-alt-generic-powerpc.ll
@@ -0,0 +1,321 @@
+; RUN: llc < %s -march=ppc32
+; ModuleID = 'mult-alt-generic.c'
+target datalayout = "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-v128:128:128-n32"
+target triple = "powerpc"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/PowerPC/mult-alt-generic-powerpc64.ll b/test/CodeGen/PowerPC/mult-alt-generic-powerpc64.ll
new file mode 100644
index 0000000..3da06f6
--- /dev/null
+++ b/test/CodeGen/PowerPC/mult-alt-generic-powerpc64.ll
@@ -0,0 +1,321 @@
+; RUN: llc < %s -march=ppc64
+; ModuleID = 'mult-alt-generic.c'
+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-v128:128:128-n32:64"
+target triple = "powerpc64"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/SPARC/mult-alt-generic-sparc.ll b/test/CodeGen/SPARC/mult-alt-generic-sparc.ll
new file mode 100644
index 0000000..6013b17
--- /dev/null
+++ b/test/CodeGen/SPARC/mult-alt-generic-sparc.ll
@@ -0,0 +1,323 @@
+; RUN: llc < %s -march=sparc
+; ModuleID = 'mult-alt-generic.c'
+target datalayout = "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-n32"
+target triple = "sparc"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+; No lowering support.
+;  %4 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+001) nounwind
+;  store i32 %4, i32* %out0, align 4
+;  %5 = call i32 asm "foo $1,$0", "=r,X"(double 1.000000e+000) nounwind
+;  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+; No lowering support.
+;  %4 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+001) nounwind
+;  store i32 %4, i32* %out0, align 4
+;  %5 = call i32 asm "foo $1,$0", "=r|r,r|X"(double 1.000000e+000) nounwind
+;  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|r"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/X86/mult-alt-generic-i686.ll b/test/CodeGen/X86/mult-alt-generic-i686.ll
new file mode 100644
index 0000000..7c3499f
--- /dev/null
+++ b/test/CodeGen/X86/mult-alt-generic-i686.ll
@@ -0,0 +1,321 @@
+; RUN: llc < %s -march=x86
+; ModuleID = 'mult-alt-generic.c'
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32-n8:16:32"
+target triple = "i686"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,im,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|im,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/X86/mult-alt-generic-x86_64.ll b/test/CodeGen/X86/mult-alt-generic-x86_64.ll
new file mode 100644
index 0000000..f35bb5e
--- /dev/null
+++ b/test/CodeGen/X86/mult-alt-generic-x86_64.ll
@@ -0,0 +1,321 @@
+; RUN: llc < %s -march=x86-64
+; ModuleID = 'mult-alt-generic.c'
+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"
+target triple = "x86_64"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_m() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,*m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32* @min1) nounwind
+  ret void
+}
+
+define void @single_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @single_V() nounwind {
+entry:
+  ret void
+}
+
+define void @single_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,<r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r<,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,>r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,r>,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @single_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,i,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,n,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,E,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r,F,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @single_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @single_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,imr,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @single_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r,X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @single_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r,im,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_m() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*m|r,m|r,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_o() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %index = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %index, align 4
+  ret void
+}
+
+define void @multi_V() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_lt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|<r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r<,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_gt() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|>r,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* %in1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|r>,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_r() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|m,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_i() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|i,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_n() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|n,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_E() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|E,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_F() nounwind {
+entry:
+  %out0 = alloca double, align 8
+  store double 0.000000e+000, double* %out0, align 8
+; No lowering support.
+;  %0 = call double asm "foo $1,$0", "=r|r,r|F,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+;  store double %0, double* %out0, align 8
+  ret void
+}
+
+define void @multi_s() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_g() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|imr,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_X() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  %in1 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  store i32 1, i32* %in1, align 4
+  %tmp = load i32* %in1, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* %out0, align 4
+  %tmp1 = load i32* @min1, align 4
+  %1 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 %tmp1) nounwind
+  store i32 %1, i32* %out0, align 4
+  %2 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32 1) nounwind
+  store i32 %2, i32* %out0, align 4
+  %3 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %3, i32* %out0, align 4
+  %4 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+001) nounwind
+  store i32 %4, i32* %out0, align 4
+  %5 = call i32 asm "foo $1,$0", "=r|r,r|X,~{dirflag},~{fpsr},~{flags}"(double 1.000000e+000) nounwind
+  store i32 %5, i32* %out0, align 4
+  ret void
+}
+
+define void @multi_p() nounwind {
+entry:
+  %out0 = alloca i32, align 4
+  store i32 0, i32* %out0, align 4
+  %0 = call i32 asm "foo $1,$0", "=r|r,r|im,~{dirflag},~{fpsr},~{flags}"(i32* getelementptr inbounds ([2 x i32]* @marray, i32 0, i32 0)) nounwind
+  store i32 %0, i32* %out0, align 4
+  ret void
+}
diff --git a/test/CodeGen/X86/mult-alt-x86.ll b/test/CodeGen/X86/mult-alt-x86.ll
new file mode 100644
index 0000000..c2c3e2d
--- /dev/null
+++ b/test/CodeGen/X86/mult-alt-x86.ll
@@ -0,0 +1,358 @@
+; RUN: llc < %s -march=x86
+; ModuleID = 'mult-alt-x86.c'
+target datalayout = "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-f80:128:128-v64:64:64-v128:128:128-a0:0:64-f80:32:32-n8:16:32"
+target triple = "i686-pc-win32"
+
+@mout0 = common global i32 0, align 4
+@min1 = common global i32 0, align 4
+@dout0 = common global double 0.000000e+000, align 8
+@din1 = common global double 0.000000e+000, align 8
+@marray = common global [2 x i32] zeroinitializer, align 4
+
+define void @single_R() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "=R,R,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_q() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "=q,q,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_Q() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "=Q,Q,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_a() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={ax},{ax},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_b() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={bx},{bx},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_c() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={cx},{cx},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_d() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={dx},{dx},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_S() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={si},{si},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_D() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "={di},{di},~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_A() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  %0 = call i32 asm "foo $1,$0", "=A,A,~{dirflag},~{fpsr},~{flags}"(i32 %tmp) nounwind
+  store i32 %0, i32* @mout0, align 4
+  ret void
+}
+
+define void @single_f() nounwind {
+entry:
+  ret void
+}
+
+define void @single_t() nounwind {
+entry:
+  ret void
+}
+
+define void @single_u() nounwind {
+entry:
+  ret void
+}
+
+define void @single_y() nounwind {
+entry:
+  %tmp = load double* @din1, align 8
+  %0 = call double asm "foo $1,$0", "=y,y,~{dirflag},~{fpsr},~{flags}"(double %tmp) nounwind
+  store double %0, double* @dout0, align 8
+  ret void
+}
+
+define void @single_x() nounwind {
+entry:
+  %tmp = load double* @din1, align 8
+  %0 = call double asm "foo $1,$0", "=x,x,~{dirflag},~{fpsr},~{flags}"(double %tmp) nounwind
+  store double %0, double* @dout0, align 8
+  ret void
+}
+
+define void @single_Y0() nounwind {
+entry:
+  ret void
+}
+
+define void @single_I() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,I,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_J() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,J,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_K() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,K,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_L() nounwind {
+entry:
+; Missing lowering support for 'L'.
+;  call void asm "foo $1,$0", "=*m,L,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_M() nounwind {
+entry:
+; Missing lowering support for 'M'.
+;  call void asm "foo $1,$0", "=*m,M,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_N() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,N,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_G() nounwind {
+entry:
+; Missing lowering support for 'G'.
+;  call void asm "foo $1,$0", "=*m,G,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, double 1.000000e+000) nounwind
+  ret void
+}
+
+define void @single_C() nounwind {
+entry:
+; Missing lowering support for 'C'.
+;  call void asm "foo $1,$0", "=*m,C,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, double 1.000000e+000) nounwind
+  ret void
+}
+
+define void @single_e() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,e,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @single_Z() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*m,Z,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_R() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|R|m,r|R|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_q() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|q|m,r|q|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_Q() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|Q|m,r|Q|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_a() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{ax}|m,r|{ax}|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_b() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{bx}|m,r|{bx}|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_c() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{cx}|m,r|{cx}|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_d() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{dx}|m,r|{dx},~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_S() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{si}|m,r|{si}|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_D() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|{di}|m,r|{di}|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_A() nounwind {
+entry:
+  %tmp = load i32* @min1, align 4
+  call void asm "foo $1,$0", "=*r|A|m,r|A|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 %tmp) nounwind
+  ret void
+}
+
+define void @multi_f() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_t() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_u() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_y() nounwind {
+entry:
+  %tmp = load double* @din1, align 8
+  call void asm "foo $1,$0", "=*r|y|m,r|y|m,~{dirflag},~{fpsr},~{flags}"(double* @dout0, double %tmp) nounwind
+  ret void
+}
+
+define void @multi_x() nounwind {
+entry:
+  %tmp = load double* @din1, align 8
+  call void asm "foo $1,$0", "=*r|x|m,r|x|m,~{dirflag},~{fpsr},~{flags}"(double* @dout0, double %tmp) nounwind
+  ret void
+}
+
+define void @multi_Y0() nounwind {
+entry:
+  ret void
+}
+
+define void @multi_I() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|I|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_J() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|J|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_K() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|K|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_L() nounwind {
+entry:
+; Missing lowering support for 'L'.
+;  call void asm "foo $1,$0", "=*r|m|m,r|L|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_M() nounwind {
+entry:
+; Missing lowering support for 'M'.
+;  call void asm "foo $1,$0", "=*r|m|m,r|M|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_N() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|N|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_G() nounwind {
+entry:
+; Missing lowering support for 'G'.
+;  call void asm "foo $1,$0", "=*r|m|m,r|G|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, double 1.000000e+000) nounwind
+  ret void
+}
+
+define void @multi_C() nounwind {
+entry:
+; Missing lowering support for 'C'.
+;  call void asm "foo $1,$0", "=*r|m|m,r|C|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, double 1.000000e+000) nounwind
+  ret void
+}
+
+define void @multi_e() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|e|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}
+
+define void @multi_Z() nounwind {
+entry:
+  call void asm "foo $1,$0", "=*r|m|m,r|Z|m,~{dirflag},~{fpsr},~{flags}"(i32* @mout0, i32 1) nounwind
+  ret void
+}