20 files changed, 1563 insertions, 85 deletions

diff --git a/test/CodeGen/PTX/add.ll b/test/CodeGen/PTX/add.ll
index 1259d03e96c9..235b00e8782f 100644
--- a/test/CodeGen/PTX/add.ll
+++ b/test/CodeGen/PTX/add.ll
@@ -1,15 +1,71 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
-define ptx_device i32 @t1(i32 %x, i32 %y) {
-; CHECK: add.s32 r0, r1, r2;
+define ptx_device i16 @t1_u16(i16 %x, i16 %y) {
+; CHECK: add.u16 rh0, rh1, rh2;
+; CHECK-NEXT: ret;
+	%z = add i16 %x, %y
+	ret i16 %z
+}
+
+define ptx_device i32 @t1_u32(i32 %x, i32 %y) {
+; CHECK: add.u32 r0, r1, r2;
+; CHECK-NEXT: ret;
 	%z = add i32 %x, %y
-; CHECK: ret;
 	ret i32 %z
 }
 
-define ptx_device i32 @t2(i32 %x) {
-; CHECK: add.s32 r0, r1, 1;
+define ptx_device i64 @t1_u64(i64 %x, i64 %y) {
+; CHECK: add.u64 rd0, rd1, rd2;
+; CHECK-NEXT: ret;
+	%z = add i64 %x, %y
+	ret i64 %z
+}
+
+define ptx_device float @t1_f32(float %x, float %y) {
+; CHECK: add.f32 f0, f1, f2
+; CHECK-NEXT: ret;
+  %z = fadd float %x, %y
+  ret float %z
+}
+
+define ptx_device double @t1_f64(double %x, double %y) {
+; CHECK: add.f64 fd0, fd1, fd2
+; CHECK-NEXT: ret;
+  %z = fadd double %x, %y
+  ret double %z
+}
+
+define ptx_device i16 @t2_u16(i16 %x) {
+; CHECK: add.u16 rh0, rh1, 1;
+; CHECK-NEXT: ret;
+	%z = add i16 %x, 1
+	ret i16 %z
+}
+
+define ptx_device i32 @t2_u32(i32 %x) {
+; CHECK: add.u32 r0, r1, 1;
+; CHECK-NEXT: ret;
 	%z = add i32 %x, 1
-; CHECK: ret;
 	ret i32 %z
 }
+
+define ptx_device i64 @t2_u64(i64 %x) {
+; CHECK: add.u64 rd0, rd1, 1;
+; CHECK-NEXT: ret;
+	%z = add i64 %x, 1
+	ret i64 %z
+}
+
+define ptx_device float @t2_f32(float %x) {
+; CHECK: add.f32 f0, f1, 0F3F800000;
+; CHECK-NEXT: ret;
+  %z = fadd float %x, 1.0
+  ret float %z
+}
+
+define ptx_device double @t2_f64(double %x) {
+; CHECK: add.f64 fd0, fd1, 0D3FF0000000000000;
+; CHECK-NEXT: ret;
+  %z = fadd double %x, 1.0
+  ret double %z
+}
diff --git a/test/CodeGen/PTX/bitwise.ll b/test/CodeGen/PTX/bitwise.ll
new file mode 100644
index 000000000000..dbc77e53330b
--- /dev/null
+++ b/test/CodeGen/PTX/bitwise.ll
@@ -0,0 +1,24 @@
+; RUN: llc < %s -march=ptx32 | FileCheck %s
+
+; preds
+
+define ptx_device i32 @t1_and_preds(i1 %x, i1 %y) {
+; CHECK: and.pred p0, p1, p2
+  %c = and i1 %x, %y
+  %d = zext i1 %c to i32 
+  ret i32 %d
+}
+
+define ptx_device i32 @t1_or_preds(i1 %x, i1 %y) {
+; CHECK: or.pred p0, p1, p2
+  %a = or i1 %x, %y
+  %b = zext i1 %a to i32 
+  ret i32 %b
+}
+
+define ptx_device i32 @t1_xor_preds(i1 %x, i1 %y) {
+; CHECK: xor.pred p0, p1, p2
+  %a = xor i1 %x, %y
+  %b = zext i1 %a to i32 
+  ret i32 %b
+}
diff --git a/test/CodeGen/PTX/bra.ll b/test/CodeGen/PTX/bra.ll
new file mode 100644
index 000000000000..49383eb3cf96
--- /dev/null
+++ b/test/CodeGen/PTX/bra.ll
@@ -0,0 +1,24 @@
+; RUN: llc < %s -march=ptx32 | FileCheck %s
+
+define ptx_device void @test_bra_direct() {
+; CHECK: bra $L__BB0_1;
+entry:
+	br label %loop
+loop:
+	br label %loop
+}
+
+define ptx_device i32 @test_bra_cond_direct(i32 %x, i32 %y) {
+entry:
+; CHECK: setp.le.u32 p0, r1, r2
+	%p = icmp ugt i32 %x, %y
+; CHECK-NEXT: @p0 bra
+; CHECK-NOT: bra
+	br i1 %p, label %clause.if, label %clause.else
+clause.if:
+; CHECK: mov.u32 r0, r1
+	ret i32 %x
+clause.else:
+; CHECK: mov.u32 r0, r2
+	ret i32 %y
+}
diff --git a/test/CodeGen/PTX/exit.ll b/test/CodeGen/PTX/exit.ll
index 4071babb80ce..7816c801728f 100644
--- a/test/CodeGen/PTX/exit.ll
+++ b/test/CodeGen/PTX/exit.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
 define ptx_kernel void @t1() {
 ; CHECK: exit;
diff --git a/test/CodeGen/PTX/fdiv-sm10.ll b/test/CodeGen/PTX/fdiv-sm10.ll
new file mode 100644
index 000000000000..121360ce9be3
--- /dev/null
+++ b/test/CodeGen/PTX/fdiv-sm10.ll
@@ -0,0 +1,15 @@
+; RUN: llc < %s -march=ptx32 -mattr=+sm10 | FileCheck %s
+
+define ptx_device float @t1_f32(float %x, float %y) {
+; CHECK: div.approx.f32 f0, f1, f2;
+; CHECK-NEXT: ret;
+	%a = fdiv float %x, %y
+	ret float %a
+}
+
+define ptx_device double @t1_f64(double %x, double %y) {
+; CHECK: div.f64 fd0, fd1, fd2;
+; CHECK-NEXT: ret;
+	%a = fdiv double %x, %y
+	ret double %a
+}
diff --git a/test/CodeGen/PTX/fdiv-sm13.ll b/test/CodeGen/PTX/fdiv-sm13.ll
new file mode 100644
index 000000000000..0ec7bae8030e
--- /dev/null
+++ b/test/CodeGen/PTX/fdiv-sm13.ll
@@ -0,0 +1,15 @@
+; RUN: llc < %s -march=ptx32 -mattr=+sm13 | FileCheck %s
+
+define ptx_device float @t1_f32(float %x, float %y) {
+; CHECK: div.approx.f32 f0, f1, f2;
+; CHECK-NEXT: ret;
+	%a = fdiv float %x, %y
+	ret float %a
+}
+
+define ptx_device double @t1_f64(double %x, double %y) {
+; CHECK: div.rn.f64 fd0, fd1, fd2;
+; CHECK-NEXT: ret;
+	%a = fdiv double %x, %y
+	ret double %a
+}
diff --git a/test/CodeGen/PTX/intrinsic.ll b/test/CodeGen/PTX/intrinsic.ll
new file mode 100644
index 000000000000..cea41827ca47
--- /dev/null
+++ b/test/CodeGen/PTX/intrinsic.ll
@@ -0,0 +1,281 @@
+; RUN: llc < %s -march=ptx32 -mattr=+ptx20,+sm20 | FileCheck %s
+
+define ptx_device i32 @test_tid_x() {
+; CHECK: mov.u32 r0, %tid.x;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.tid.x()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_tid_y() {
+; CHECK: mov.u32 r0, %tid.y;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.tid.y()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_tid_z() {
+; CHECK: mov.u32 r0, %tid.z;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.tid.z()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_tid_w() {
+; CHECK: mov.u32 r0, %tid.w;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.tid.w()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ntid_x() {
+; CHECK: mov.u32 r0, %ntid.x;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ntid.x()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ntid_y() {
+; CHECK: mov.u32 r0, %ntid.y;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ntid.y()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ntid_z() {
+; CHECK: mov.u32 r0, %ntid.z;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ntid.z()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ntid_w() {
+; CHECK: mov.u32 r0, %ntid.w;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ntid.w()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_laneid() {
+; CHECK: mov.u32 r0, %laneid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.laneid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_warpid() {
+; CHECK: mov.u32 r0, %warpid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.warpid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nwarpid() {
+; CHECK: mov.u32 r0, %nwarpid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nwarpid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ctaid_x() {
+; CHECK: mov.u32 r0, %ctaid.x;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ctaid.x()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ctaid_y() {
+; CHECK: mov.u32 r0, %ctaid.y;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ctaid.y()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ctaid_z() {
+; CHECK: mov.u32 r0, %ctaid.z;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ctaid.z()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_ctaid_w() {
+; CHECK: mov.u32 r0, %ctaid.w;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.ctaid.w()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nctaid_x() {
+; CHECK: mov.u32 r0, %nctaid.x;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nctaid.x()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nctaid_y() {
+; CHECK: mov.u32 r0, %nctaid.y;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nctaid.y()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nctaid_z() {
+; CHECK: mov.u32 r0, %nctaid.z;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nctaid.z()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nctaid_w() {
+; CHECK: mov.u32 r0, %nctaid.w;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nctaid.w()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_smid() {
+; CHECK: mov.u32 r0, %smid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.smid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_nsmid() {
+; CHECK: mov.u32 r0, %nsmid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.nsmid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_gridid() {
+; CHECK: mov.u32 r0, %gridid;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.gridid()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_lanemask_eq() {
+; CHECK: mov.u32 r0, %lanemask_eq;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.lanemask.eq()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_lanemask_le() {
+; CHECK: mov.u32 r0, %lanemask_le;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.lanemask.le()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_lanemask_lt() {
+; CHECK: mov.u32 r0, %lanemask_lt;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.lanemask.lt()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_lanemask_ge() {
+; CHECK: mov.u32 r0, %lanemask_ge;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.lanemask.ge()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_lanemask_gt() {
+; CHECK: mov.u32 r0, %lanemask_gt;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.lanemask.gt()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_clock() {
+; CHECK: mov.u32 r0, %clock;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.clock()
+	ret i32 %x
+}
+
+define ptx_device i64 @test_clock64() {
+; CHECK: mov.u64 rd0, %clock64;
+; CHECK-NEXT: ret;
+	%x = call i64 @llvm.ptx.read.clock64()
+	ret i64 %x
+}
+
+define ptx_device i32 @test_pm0() {
+; CHECK: mov.u32 r0, %pm0;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.pm0()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_pm1() {
+; CHECK: mov.u32 r0, %pm1;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.pm1()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_pm2() {
+; CHECK: mov.u32 r0, %pm2;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.pm2()
+	ret i32 %x
+}
+
+define ptx_device i32 @test_pm3() {
+; CHECK: mov.u32 r0, %pm3;
+; CHECK-NEXT: ret;
+	%x = call i32 @llvm.ptx.read.pm3()
+	ret i32 %x
+}
+
+define ptx_device void @test_bar_sync() {
+; CHECK: bar.sync 0
+; CHECK-NEXT: ret;
+	call void @llvm.ptx.bar.sync(i32 0)
+	ret void
+}
+
+declare i32 @llvm.ptx.read.tid.x()
+declare i32 @llvm.ptx.read.tid.y()
+declare i32 @llvm.ptx.read.tid.z()
+declare i32 @llvm.ptx.read.tid.w()
+declare i32 @llvm.ptx.read.ntid.x()
+declare i32 @llvm.ptx.read.ntid.y()
+declare i32 @llvm.ptx.read.ntid.z()
+declare i32 @llvm.ptx.read.ntid.w()
+
+declare i32 @llvm.ptx.read.laneid()
+declare i32 @llvm.ptx.read.warpid()
+declare i32 @llvm.ptx.read.nwarpid()
+
+declare i32 @llvm.ptx.read.ctaid.x()
+declare i32 @llvm.ptx.read.ctaid.y()
+declare i32 @llvm.ptx.read.ctaid.z()
+declare i32 @llvm.ptx.read.ctaid.w()
+declare i32 @llvm.ptx.read.nctaid.x()
+declare i32 @llvm.ptx.read.nctaid.y()
+declare i32 @llvm.ptx.read.nctaid.z()
+declare i32 @llvm.ptx.read.nctaid.w()
+
+declare i32 @llvm.ptx.read.smid()
+declare i32 @llvm.ptx.read.nsmid()
+declare i32 @llvm.ptx.read.gridid()
+
+declare i32 @llvm.ptx.read.lanemask.eq()
+declare i32 @llvm.ptx.read.lanemask.le()
+declare i32 @llvm.ptx.read.lanemask.lt()
+declare i32 @llvm.ptx.read.lanemask.ge()
+declare i32 @llvm.ptx.read.lanemask.gt()
+
+declare i32 @llvm.ptx.read.clock()
+declare i64 @llvm.ptx.read.clock64()
+
+declare i32 @llvm.ptx.read.pm0()
+declare i32 @llvm.ptx.read.pm1()
+declare i32 @llvm.ptx.read.pm2()
+declare i32 @llvm.ptx.read.pm3()
+
+declare void @llvm.ptx.bar.sync(i32 %i)
diff --git a/test/CodeGen/PTX/ld.ll b/test/CodeGen/PTX/ld.ll
index 836c4d41045a..377a95abe3db 100644
--- a/test/CodeGen/PTX/ld.ll
+++ b/test/CodeGen/PTX/ld.ll
@@ -1,78 +1,447 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
-;CHECK: .extern .global .s32 array[];
-@array = external global [10 x i32]
+;CHECK: .extern .global .b8 array_i16[20];
+@array_i16 = external global [10 x i16]
 
-;CHECK: .extern .const .s32 array_constant[];
-@array_constant = external addrspace(1) constant [10 x i32]
+;CHECK: .extern .const .b8 array_constant_i16[20];
+@array_constant_i16 = external addrspace(1) constant [10 x i16]
 
-;CHECK: .extern .local .s32 array_local[];
-@array_local = external addrspace(2) global [10 x i32]
+;CHECK: .extern .local .b8 array_local_i16[20];
+@array_local_i16 = external addrspace(2) global [10 x i16]
 
-;CHECK: .extern .shared .s32 array_shared[];
-@array_shared = external addrspace(4) global [10 x i32]
+;CHECK: .extern .shared .b8 array_shared_i16[20];
+@array_shared_i16 = external addrspace(4) global [10 x i16]
 
-define ptx_device i32 @t1(i32* %p) {
+;CHECK: .extern .global .b8 array_i32[40];
+@array_i32 = external global [10 x i32]
+
+;CHECK: .extern .const .b8 array_constant_i32[40];
+@array_constant_i32 = external addrspace(1) constant [10 x i32]
+
+;CHECK: .extern .local .b8 array_local_i32[40];
+@array_local_i32 = external addrspace(2) global [10 x i32]
+
+;CHECK: .extern .shared .b8 array_shared_i32[40];
+@array_shared_i32 = external addrspace(4) global [10 x i32]
+
+;CHECK: .extern .global .b8 array_i64[80];
+@array_i64 = external global [10 x i64]
+
+;CHECK: .extern .const .b8 array_constant_i64[80];
+@array_constant_i64 = external addrspace(1) constant [10 x i64]
+
+;CHECK: .extern .local .b8 array_local_i64[80];
+@array_local_i64 = external addrspace(2) global [10 x i64]
+
+;CHECK: .extern .shared .b8 array_shared_i64[80];
+@array_shared_i64 = external addrspace(4) global [10 x i64]
+
+;CHECK: .extern .global .b8 array_float[40];
+@array_float = external global [10 x float]
+
+;CHECK: .extern .const .b8 array_constant_float[40];
+@array_constant_float = external addrspace(1) constant [10 x float]
+
+;CHECK: .extern .local .b8 array_local_float[40];
+@array_local_float = external addrspace(2) global [10 x float]
+
+;CHECK: .extern .shared .b8 array_shared_float[40];
+@array_shared_float = external addrspace(4) global [10 x float]
+
+;CHECK: .extern .global .b8 array_double[80];
+@array_double = external global [10 x double]
+
+;CHECK: .extern .const .b8 array_constant_double[80];
+@array_constant_double = external addrspace(1) constant [10 x double]
+
+;CHECK: .extern .local .b8 array_local_double[80];
+@array_local_double = external addrspace(2) global [10 x double]
+
+;CHECK: .extern .shared .b8 array_shared_double[80];
+@array_shared_double = external addrspace(4) global [10 x double]
+
+
+define ptx_device i16 @t1_u16(i16* %p) {
 entry:
-;CHECK: ld.global.s32 r0, [r1];
+;CHECK: ld.global.u16 rh0, [r1];
+;CHECK-NEXT: ret;
+  %x = load i16* %p
+  ret i16 %x
+}
+
+define ptx_device i32 @t1_u32(i32* %p) {
+entry:
+;CHECK: ld.global.u32 r0, [r1];
+;CHECK-NEXT: ret;
   %x = load i32* %p
   ret i32 %x
 }
 
-define ptx_device i32 @t2(i32* %p) {
+define ptx_device i64 @t1_u64(i64* %p) {
+entry:
+;CHECK: ld.global.u64 rd0, [r1];
+;CHECK-NEXT: ret;
+  %x = load i64* %p
+  ret i64 %x
+}
+
+define ptx_device float @t1_f32(float* %p) {
+entry:
+;CHECK: ld.global.f32 f0, [r1];
+;CHECK-NEXT: ret;
+  %x = load float* %p
+  ret float %x
+}
+
+define ptx_device double @t1_f64(double* %p) {
+entry:
+;CHECK: ld.global.f64 fd0, [r1];
+;CHECK-NEXT: ret;
+  %x = load double* %p
+  ret double %x
+}
+
+define ptx_device i16 @t2_u16(i16* %p) {
 entry:
-;CHECK: ld.global.s32 r0, [r1+4];
+;CHECK: ld.global.u16 rh0, [r1+2];
+;CHECK-NEXT: ret;
+  %i = getelementptr i16* %p, i32 1
+  %x = load i16* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t2_u32(i32* %p) {
+entry:
+;CHECK: ld.global.u32 r0, [r1+4];
+;CHECK-NEXT: ret;
   %i = getelementptr i32* %p, i32 1
   %x = load i32* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t3(i32* %p, i32 %q) {
+define ptx_device i64 @t2_u64(i64* %p) {
+entry:
+;CHECK: ld.global.u64 rd0, [r1+8];
+;CHECK-NEXT: ret;
+  %i = getelementptr i64* %p, i32 1
+  %x = load i64* %i
+  ret i64 %x
+}
+
+define ptx_device float @t2_f32(float* %p) {
+entry:
+;CHECK: ld.global.f32 f0, [r1+4];
+;CHECK-NEXT: ret;
+  %i = getelementptr float* %p, i32 1
+  %x = load float* %i
+  ret float %x
+}
+
+define ptx_device double @t2_f64(double* %p) {
+entry:
+;CHECK: ld.global.f64 fd0, [r1+8];
+;CHECK-NEXT: ret;
+  %i = getelementptr double* %p, i32 1
+  %x = load double* %i
+  ret double %x
+}
+
+define ptx_device i16 @t3_u16(i16* %p, i32 %q) {
+entry:
+;CHECK: shl.b32 r0, r2, 1;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: ld.global.u16 rh0, [r0];
+  %i = getelementptr i16* %p, i32 %q
+  %x = load i16* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t3_u32(i32* %p, i32 %q) {
 entry:
 ;CHECK: shl.b32 r0, r2, 2;
-;CHECK: add.s32 r0, r1, r0;
-;CHECK: ld.global.s32 r0, [r0];
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: ld.global.u32 r0, [r0];
   %i = getelementptr i32* %p, i32 %q
   %x = load i32* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t4_global() {
+define ptx_device i64 @t3_u64(i64* %p, i32 %q) {
+entry:
+;CHECK: shl.b32 r0, r2, 3;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: ld.global.u64 rd0, [r0];
+  %i = getelementptr i64* %p, i32 %q
+  %x = load i64* %i
+  ret i64 %x
+}
+
+define ptx_device float @t3_f32(float* %p, i32 %q) {
+entry:
+;CHECK: shl.b32 r0, r2, 2;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: ld.global.f32 f0, [r0];
+  %i = getelementptr float* %p, i32 %q
+  %x = load float* %i
+  ret float %x
+}
+
+define ptx_device double @t3_f64(double* %p, i32 %q) {
+entry:
+;CHECK: shl.b32 r0, r2, 3;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: ld.global.f64 fd0, [r0];
+  %i = getelementptr double* %p, i32 %q
+  %x = load double* %i
+  ret double %x
+}
+
+define ptx_device i16 @t4_global_u16() {
+entry:
+;CHECK: mov.u32 r0, array_i16;
+;CHECK-NEXT: ld.global.u16 rh0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16]* @array_i16, i32 0, i32 0
+  %x = load i16* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t4_global_u32() {
 entry:
-;CHECK: ld.global.s32 r0, [array];
-  %i = getelementptr [10 x i32]* @array, i32 0, i32 0
+;CHECK: mov.u32 r0, array_i32;
+;CHECK-NEXT: ld.global.u32 r0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32]* @array_i32, i32 0, i32 0
   %x = load i32* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t4_const() {
+define ptx_device i64 @t4_global_u64() {
 entry:
-;CHECK: ld.const.s32 r0, [array_constant];
-  %i = getelementptr [10 x i32] addrspace(1)* @array_constant, i32 0, i32 0
+;CHECK: mov.u32 r0, array_i64;
+;CHECK-NEXT: ld.global.u64 rd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64]* @array_i64, i32 0, i32 0
+  %x = load i64* %i
+  ret i64 %x
+}
+
+define ptx_device float @t4_global_f32() {
+entry:
+;CHECK: mov.u32 r0, array_float;
+;CHECK-NEXT: ld.global.f32 f0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float]* @array_float, i32 0, i32 0
+  %x = load float* %i
+  ret float %x
+}
+
+define ptx_device double @t4_global_f64() {
+entry:
+;CHECK: mov.u32 r0, array_double;
+;CHECK-NEXT: ld.global.f64 fd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double]* @array_double, i32 0, i32 0
+  %x = load double* %i
+  ret double %x
+}
+
+define ptx_device i16 @t4_const_u16() {
+entry:
+;CHECK: mov.u32 r0, array_constant_i16;
+;CHECK-NEXT: ld.const.u16 rh0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16] addrspace(1)* @array_constant_i16, i32 0, i32 0
+  %x = load i16 addrspace(1)* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t4_const_u32() {
+entry:
+;CHECK: mov.u32 r0, array_constant_i32;
+;CHECK-NEXT: ld.const.u32 r0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32] addrspace(1)* @array_constant_i32, i32 0, i32 0
   %x = load i32 addrspace(1)* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t4_local() {
+define ptx_device i64 @t4_const_u64() {
+entry:
+;CHECK: mov.u32 r0, array_constant_i64;
+;CHECK-NEXT: ld.const.u64 rd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64] addrspace(1)* @array_constant_i64, i32 0, i32 0
+  %x = load i64 addrspace(1)* %i
+  ret i64 %x
+}
+
+define ptx_device float @t4_const_f32() {
 entry:
-;CHECK: ld.local.s32 r0, [array_local];
-  %i = getelementptr [10 x i32] addrspace(2)* @array_local, i32 0, i32 0
+;CHECK: mov.u32 r0, array_constant_float;
+;CHECK-NEXT: ld.const.f32 f0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float] addrspace(1)* @array_constant_float, i32 0, i32 0
+  %x = load float addrspace(1)* %i
+  ret float %x
+}
+
+define ptx_device double @t4_const_f64() {
+entry:
+;CHECK: mov.u32 r0, array_constant_double;
+;CHECK-NEXT: ld.const.f64 fd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double] addrspace(1)* @array_constant_double, i32 0, i32 0
+  %x = load double addrspace(1)* %i
+  ret double %x
+}
+
+define ptx_device i16 @t4_local_u16() {
+entry:
+;CHECK: mov.u32 r0, array_local_i16;
+;CHECK-NEXT: ld.local.u16 rh0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16] addrspace(2)* @array_local_i16, i32 0, i32 0
+  %x = load i16 addrspace(2)* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t4_local_u32() {
+entry:
+;CHECK: mov.u32 r0, array_local_i32;
+;CHECK-NEXT: ld.local.u32 r0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32] addrspace(2)* @array_local_i32, i32 0, i32 0
   %x = load i32 addrspace(2)* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t4_shared() {
+define ptx_device i64 @t4_local_u64() {
 entry:
-;CHECK: ld.shared.s32 r0, [array_shared];
-  %i = getelementptr [10 x i32] addrspace(4)* @array_shared, i32 0, i32 0
+;CHECK: mov.u32 r0, array_local_i64;
+;CHECK-NEXT: ld.local.u64 rd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64] addrspace(2)* @array_local_i64, i32 0, i32 0
+  %x = load i64 addrspace(2)* %i
+  ret i64 %x
+}
+
+define ptx_device float @t4_local_f32() {
+entry:
+;CHECK: mov.u32 r0, array_local_float;
+;CHECK-NEXT: ld.local.f32 f0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float] addrspace(2)* @array_local_float, i32 0, i32 0
+  %x = load float addrspace(2)* %i
+  ret float %x
+}
+
+define ptx_device double @t4_local_f64() {
+entry:
+;CHECK: mov.u32 r0, array_local_double;
+;CHECK-NEXT: ld.local.f64 fd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double] addrspace(2)* @array_local_double, i32 0, i32 0
+  %x = load double addrspace(2)* %i
+  ret double %x
+}
+
+define ptx_device i16 @t4_shared_u16() {
+entry:
+;CHECK: mov.u32 r0, array_shared_i16;
+;CHECK-NEXT: ld.shared.u16 rh0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16] addrspace(4)* @array_shared_i16, i32 0, i32 0
+  %x = load i16 addrspace(4)* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t4_shared_u32() {
+entry:
+;CHECK: mov.u32 r0, array_shared_i32;
+;CHECK-NEXT: ld.shared.u32 r0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32] addrspace(4)* @array_shared_i32, i32 0, i32 0
   %x = load i32 addrspace(4)* %i
   ret i32 %x
 }
 
-define ptx_device i32 @t5() {
+define ptx_device i64 @t4_shared_u64() {
+entry:
+;CHECK: mov.u32 r0, array_shared_i64;
+;CHECK-NEXT: ld.shared.u64 rd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64] addrspace(4)* @array_shared_i64, i32 0, i32 0
+  %x = load i64 addrspace(4)* %i
+  ret i64 %x
+}
+
+define ptx_device float @t4_shared_f32() {
+entry:
+;CHECK: mov.u32 r0, array_shared_float;
+;CHECK-NEXT: ld.shared.f32 f0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float] addrspace(4)* @array_shared_float, i32 0, i32 0
+  %x = load float addrspace(4)* %i
+  ret float %x
+}
+
+define ptx_device double @t4_shared_f64() {
+entry:
+;CHECK: mov.u32 r0, array_shared_double;
+;CHECK-NEXT: ld.shared.f64 fd0, [r0];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double] addrspace(4)* @array_shared_double, i32 0, i32 0
+  %x = load double addrspace(4)* %i
+  ret double %x
+}
+
+define ptx_device i16 @t5_u16() {
 entry:
-;CHECK: ld.global.s32 r0, [array+4];
-  %i = getelementptr [10 x i32]* @array, i32 0, i32 1
+;CHECK: mov.u32 r0, array_i16;
+;CHECK-NEXT: ld.global.u16 rh0, [r0+2];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16]* @array_i16, i32 0, i32 1
+  %x = load i16* %i
+  ret i16 %x
+}
+
+define ptx_device i32 @t5_u32() {
+entry:
+;CHECK: mov.u32 r0, array_i32;
+;CHECK-NEXT: ld.global.u32 r0, [r0+4];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32]* @array_i32, i32 0, i32 1
   %x = load i32* %i
   ret i32 %x
 }
+
+define ptx_device i64 @t5_u64() {
+entry:
+;CHECK: mov.u32 r0, array_i64;
+;CHECK-NEXT: ld.global.u64 rd0, [r0+8];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64]* @array_i64, i32 0, i32 1
+  %x = load i64* %i
+  ret i64 %x
+}
+
+define ptx_device float @t5_f32() {
+entry:
+;CHECK: mov.u32 r0, array_float;
+;CHECK-NEXT: ld.global.f32 f0, [r0+4];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float]* @array_float, i32 0, i32 1
+  %x = load float* %i
+  ret float %x
+}
+
+define ptx_device double @t5_f64() {
+entry:
+;CHECK: mov.u32 r0, array_double;
+;CHECK-NEXT: ld.global.f64 fd0, [r0+8];
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double]* @array_double, i32 0, i32 1
+  %x = load double* %i
+  ret double %x
+}
diff --git a/test/CodeGen/PTX/llvm-intrinsic.ll b/test/CodeGen/PTX/llvm-intrinsic.ll
new file mode 100644
index 000000000000..1e265f5b7b3a
--- /dev/null
+++ b/test/CodeGen/PTX/llvm-intrinsic.ll
@@ -0,0 +1,56 @@
+; RUN: llc < %s -march=ptx32 -mattr=+ptx20,+sm20 | FileCheck %s
+
+define ptx_device float @test_sqrt_f32(float %x) {
+entry:
+; CHECK: sqrt.rn.f32 f0, f1;
+; CHECK-NEXT: ret;
+  %y = call float @llvm.sqrt.f32(float %x)
+  ret float %y
+}
+
+define ptx_device double @test_sqrt_f64(double %x) {
+entry:
+; CHECK: sqrt.rn.f64 fd0, fd1;
+; CHECK-NEXT: ret;
+  %y = call double @llvm.sqrt.f64(double %x)
+  ret double %y
+}
+
+define ptx_device float @test_sin_f32(float %x) {
+entry:
+; CHECK: sin.approx.f32 f0, f1;
+; CHECK-NEXT: ret;
+  %y = call float @llvm.sin.f32(float %x)
+  ret float %y
+}
+
+define ptx_device double @test_sin_f64(double %x) {
+entry:
+; CHECK: sin.approx.f64 fd0, fd1;
+; CHECK-NEXT: ret;
+  %y = call double @llvm.sin.f64(double %x)
+  ret double %y
+}
+
+define ptx_device float @test_cos_f32(float %x) {
+entry:
+; CHECK: cos.approx.f32 f0, f1;
+; CHECK-NEXT: ret;
+  %y = call float @llvm.cos.f32(float %x)
+  ret float %y
+}
+
+define ptx_device double @test_cos_f64(double %x) {
+entry:
+; CHECK: cos.approx.f64 fd0, fd1;
+; CHECK-NEXT: ret;
+  %y = call double @llvm.cos.f64(double %x)
+  ret double %y
+}
+
+declare float  @llvm.sqrt.f32(float)
+declare double @llvm.sqrt.f64(double)
+declare float  @llvm.sin.f32(float)
+declare double @llvm.sin.f64(double)
+declare float  @llvm.cos.f32(float)
+declare double @llvm.cos.f64(double)
diff --git a/test/CodeGen/PTX/mad.ll b/test/CodeGen/PTX/mad.ll
new file mode 100644
index 000000000000..0c25f2c0030a
--- /dev/null
+++ b/test/CodeGen/PTX/mad.ll
@@ -0,0 +1,17 @@
+; RUN: llc < %s -march=ptx32 -mattr=+sm13 | FileCheck %s
+
+define ptx_device float @t1_f32(float %x, float %y, float %z) {
+; CHECK: mad.rn.f32 f0, f1, f2, f3;
+; CHECK-NEXT: ret;
+	%a = fmul float %x, %y
+  %b = fadd float %a, %z
+	ret float %b
+}
+
+define ptx_device double @t1_f64(double %x, double %y, double %z) {
+; CHECK: mad.rn.f64 fd0, fd1, fd2, fd3;
+; CHECK-NEXT: ret;
+	%a = fmul double %x, %y
+  %b = fadd double %a, %z
+	ret double %b
+}
diff --git a/test/CodeGen/PTX/mov.ll b/test/CodeGen/PTX/mov.ll
index c365e9beb897..120572a0e868 100644
--- a/test/CodeGen/PTX/mov.ll
+++ b/test/CodeGen/PTX/mov.ll
@@ -1,13 +1,62 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
-define ptx_device i32 @t1() {
-; CHECK: mov.s32 r0, 0;
+define ptx_device i16 @t1_u16() {
+; CHECK: mov.u16 rh0, 0;
+; CHECK: ret;
+	ret i16 0
+}
+
+define ptx_device i32 @t1_u32() {
+; CHECK: mov.u32 r0, 0;
 ; CHECK: ret;
 	ret i32 0
 }
 
-define ptx_device i32 @t2(i32 %x) {
-; CHECK: mov.s32 r0, r1;
+define ptx_device i64 @t1_u64() {
+; CHECK: mov.u64 rd0, 0;
+; CHECK: ret;
+	ret i64 0
+}
+
+define ptx_device float @t1_f32() {
+; CHECK: mov.f32 f0, 0F00000000;
+; CHECK: ret;
+	ret float 0.0
+}
+
+define ptx_device double @t1_f64() {
+; CHECK: mov.f64 fd0, 0D0000000000000000;
+; CHECK: ret;
+	ret double 0.0
+}
+
+define ptx_device i16 @t2_u16(i16 %x) {
+; CHECK: mov.u16 rh0, rh1;
+; CHECK: ret;
+	ret i16 %x
+}
+
+define ptx_device i32 @t2_u32(i32 %x) {
+; CHECK: mov.u32 r0, r1;
 ; CHECK: ret;
 	ret i32 %x
 }
+
+define ptx_device i64 @t2_u64(i64 %x) {
+; CHECK: mov.u64 rd0, rd1;
+; CHECK: ret;
+	ret i64 %x
+}
+
+define ptx_device float @t3_f32(float %x) {
+; CHECK: mov.f32 f0, f1;
+; CHECK-NEXT: ret;
+	ret float %x
+}
+
+define ptx_device double @t3_f64(double %x) {
+; CHECK: mov.f64 fd0, fd1;
+; CHECK-NEXT: ret;
+	ret double %x
+}
+
diff --git a/test/CodeGen/PTX/mul.ll b/test/CodeGen/PTX/mul.ll
new file mode 100644
index 000000000000..5ce042675dc8
--- /dev/null
+++ b/test/CodeGen/PTX/mul.ll
@@ -0,0 +1,39 @@
+; RUN: llc < %s -march=ptx32 | FileCheck %s
+
+;define ptx_device i32 @t1(i32 %x, i32 %y) {
+;	%z = mul i32 %x, %y
+;	ret i32 %z
+;}
+
+;define ptx_device i32 @t2(i32 %x) {
+;	%z = mul i32 %x, 1
+;	ret i32 %z
+;}
+
+define ptx_device float @t1_f32(float %x, float %y) {
+; CHECK: mul.f32 f0, f1, f2
+; CHECK-NEXT: ret;
+  %z = fmul float %x, %y
+  ret float %z
+}
+
+define ptx_device double @t1_f64(double %x, double %y) {
+; CHECK: mul.f64 fd0, fd1, fd2
+; CHECK-NEXT: ret;
+  %z = fmul double %x, %y
+  ret double %z
+}
+
+define ptx_device float @t2_f32(float %x) {
+; CHECK: mul.f32 f0, f1, 0F40A00000;
+; CHECK-NEXT: ret;
+  %z = fmul float %x, 5.0
+  ret float %z
+}
+
+define ptx_device double @t2_f64(double %x) {
+; CHECK: mul.f64 fd0, fd1, 0D4014000000000000;
+; CHECK-NEXT: ret;
+  %z = fmul double %x, 5.0
+  ret double %z
+}
diff --git a/test/CodeGen/PTX/options.ll b/test/CodeGen/PTX/options.ll
index a14d5c9c27ba..ac33fef0d6e3 100644
--- a/test/CodeGen/PTX/options.ll
+++ b/test/CodeGen/PTX/options.ll
@@ -1,5 +1,9 @@
-; RUN: llc < %s -march=ptx -ptx-version=2.0 | grep ".version 2.0"
-; RUN: llc < %s -march=ptx -ptx-target=sm_20 | grep ".target sm_20"
+; RUN: llc < %s -march=ptx32 -mattr=ptx20 | grep ".version 2.0"
+; RUN: llc < %s -march=ptx32 -mattr=ptx21 | grep ".version 2.1"
+; RUN: llc < %s -march=ptx32 -mattr=ptx22 | grep ".version 2.2"
+; RUN: llc < %s -march=ptx32 -mattr=sm10 | grep ".target sm_10"
+; RUN: llc < %s -march=ptx32 -mattr=sm13 | grep ".target sm_13"
+; RUN: llc < %s -march=ptx32 -mattr=sm20 | grep ".target sm_20"
 
 define ptx_device void @t1() {
 	ret void
diff --git a/test/CodeGen/PTX/parameter-order.ll b/test/CodeGen/PTX/parameter-order.ll
new file mode 100644
index 000000000000..95d4a328149c
--- /dev/null
+++ b/test/CodeGen/PTX/parameter-order.ll
@@ -0,0 +1,8 @@
+; RUN: llc < %s -march=ptx32 | FileCheck %s
+
+; CHECK: .func (.reg .u32 r0) test_parameter_order (.reg .f32 f1, .reg .u32 r1, .reg .u32 r2, .reg .f32 f2)
+define ptx_device i32 @test_parameter_order(float %a, i32 %b, i32 %c, float %d) {
+; CHECK: sub.u32 r0, r1, r2
+	%result = sub i32 %b, %c
+	ret i32 %result
+}
diff --git a/test/CodeGen/PTX/ret.ll b/test/CodeGen/PTX/ret.ll
index d5037f25fd36..ba0523f6424a 100644
--- a/test/CodeGen/PTX/ret.ll
+++ b/test/CodeGen/PTX/ret.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
 define ptx_device void @t1() {
 ; CHECK: ret;
diff --git a/test/CodeGen/PTX/setp.ll b/test/CodeGen/PTX/setp.ll
new file mode 100644
index 000000000000..5836122049e6
--- /dev/null
+++ b/test/CodeGen/PTX/setp.ll
@@ -0,0 +1,134 @@
+; RUN: llc < %s -march=ptx32 | FileCheck %s
+
+define ptx_device i32 @test_setp_eq_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.eq.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp eq i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_ne_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.ne.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ne i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_lt_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.lt.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ult i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_le_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.le.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ule i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_gt_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.gt.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ugt i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_ge_u32_rr(i32 %x, i32 %y) {
+; CHECK: setp.ge.u32 p0, r1, r2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp uge i32 %x, %y
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_eq_u32_ri(i32 %x) {
+; CHECK: setp.eq.u32 p0, r1, 1;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp eq i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_ne_u32_ri(i32 %x) {
+; CHECK: setp.ne.u32 p0, r1, 1;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ne i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_lt_u32_ri(i32 %x) {
+; CHECK: setp.eq.u32 p0, r1, 0;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ult i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_le_u32_ri(i32 %x) {
+; CHECK: setp.lt.u32 p0, r1, 2;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ule i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_gt_u32_ri(i32 %x) {
+; CHECK: setp.gt.u32 p0, r1, 1;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp ugt i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_ge_u32_ri(i32 %x) {
+; CHECK: setp.ne.u32 p0, r1, 0;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%p = icmp uge i32 %x, 1
+	%z = zext i1 %p to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_4_op_format_1(i32 %x, i32 %y, i32 %u, i32 %v) {
+; CHECK: setp.gt.u32 p0, r3, r4;
+; CHECK-NEXT: setp.eq.and.u32 p0, r1, r2, p0;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%c = icmp eq i32 %x, %y
+	%d = icmp ugt i32 %u, %v
+	%e = and i1 %c, %d
+	%z = zext i1 %e to i32
+	ret i32 %z
+}
+
+define ptx_device i32 @test_setp_4_op_format_2(i32 %x, i32 %y, i32 %w) {
+; CHECK: cvt.pred.u32 p0, r3;
+; CHECK-NEXT: setp.eq.and.u32 p0, r1, r2, !p0;
+; CHECK-NEXT: cvt.u32.pred r0, p0;
+; CHECK-NEXT: ret;
+	%c = trunc i32 %w to i1
+	%d = icmp eq i32 %x, %y
+	%e = xor i1 %c, 1
+	%f = and i1 %d, %e
+	%z = zext i1 %f to i32
+	ret i32 %z
+}
diff --git a/test/CodeGen/PTX/shl.ll b/test/CodeGen/PTX/shl.ll
index b564b43ab932..6e72c9221325 100644
--- a/test/CodeGen/PTX/shl.ll
+++ b/test/CodeGen/PTX/shl.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
 define ptx_device i32 @t1(i32 %x, i32 %y) {
 ; CHECK: shl.b32 r0, r1, r2
diff --git a/test/CodeGen/PTX/shr.ll b/test/CodeGen/PTX/shr.ll
index 3f8ade862b75..8693e0ecf49a 100644
--- a/test/CodeGen/PTX/shr.ll
+++ b/test/CodeGen/PTX/shr.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
 define ptx_device i32 @t1(i32 %x, i32 %y) {
 ; CHECK: shr.u32 r0, r1, r2
diff --git a/test/CodeGen/PTX/st.ll b/test/CodeGen/PTX/st.ll
index 2cbacb9ee59c..dee5c61abe66 100644
--- a/test/CodeGen/PTX/st.ll
+++ b/test/CodeGen/PTX/st.ll
@@ -1,71 +1,402 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
-;CHECK: .extern .global .s32 array[];
-@array = external global [10 x i32]
+;CHECK: .extern .global .b8 array_i16[20];
+@array_i16 = external global [10 x i16]
 
-;CHECK: .extern .const .s32 array_constant[];
-@array_constant = external addrspace(1) constant [10 x i32]
+;CHECK: .extern .const .b8 array_constant_i16[20];
+@array_constant_i16 = external addrspace(1) constant [10 x i16]
 
-;CHECK: .extern .local .s32 array_local[];
-@array_local = external addrspace(2) global [10 x i32]
+;CHECK: .extern .local .b8 array_local_i16[20];
+@array_local_i16 = external addrspace(2) global [10 x i16]
 
-;CHECK: .extern .shared .s32 array_shared[];
-@array_shared = external addrspace(4) global [10 x i32]
+;CHECK: .extern .shared .b8 array_shared_i16[20];
+@array_shared_i16 = external addrspace(4) global [10 x i16]
 
-define ptx_device void @t1(i32* %p, i32 %x) {
+;CHECK: .extern .global .b8 array_i32[40];
+@array_i32 = external global [10 x i32]
+
+;CHECK: .extern .const .b8 array_constant_i32[40];
+@array_constant_i32 = external addrspace(1) constant [10 x i32]
+
+;CHECK: .extern .local .b8 array_local_i32[40];
+@array_local_i32 = external addrspace(2) global [10 x i32]
+
+;CHECK: .extern .shared .b8 array_shared_i32[40];
+@array_shared_i32 = external addrspace(4) global [10 x i32]
+
+;CHECK: .extern .global .b8 array_i64[80];
+@array_i64 = external global [10 x i64]
+
+;CHECK: .extern .const .b8 array_constant_i64[80];
+@array_constant_i64 = external addrspace(1) constant [10 x i64]
+
+;CHECK: .extern .local .b8 array_local_i64[80];
+@array_local_i64 = external addrspace(2) global [10 x i64]
+
+;CHECK: .extern .shared .b8 array_shared_i64[80];
+@array_shared_i64 = external addrspace(4) global [10 x i64]
+
+;CHECK: .extern .global .b8 array_float[40];
+@array_float = external global [10 x float]
+
+;CHECK: .extern .const .b8 array_constant_float[40];
+@array_constant_float = external addrspace(1) constant [10 x float]
+
+;CHECK: .extern .local .b8 array_local_float[40];
+@array_local_float = external addrspace(2) global [10 x float]
+
+;CHECK: .extern .shared .b8 array_shared_float[40];
+@array_shared_float = external addrspace(4) global [10 x float]
+
+;CHECK: .extern .global .b8 array_double[80];
+@array_double = external global [10 x double]
+
+;CHECK: .extern .const .b8 array_constant_double[80];
+@array_constant_double = external addrspace(1) constant [10 x double]
+
+;CHECK: .extern .local .b8 array_local_double[80];
+@array_local_double = external addrspace(2) global [10 x double]
+
+;CHECK: .extern .shared .b8 array_shared_double[80];
+@array_shared_double = external addrspace(4) global [10 x double]
+
+
+define ptx_device void @t1_u16(i16* %p, i16 %x) {
 entry:
-;CHECK: st.global.s32 [r1], r2;
+;CHECK: st.global.u16 [r1], rh1;
+;CHECK-NEXT: ret;
+  store i16 %x, i16* %p
+  ret void
+}
+
+define ptx_device void @t1_u32(i32* %p, i32 %x) {
+entry:
+;CHECK: st.global.u32 [r1], r2;
+;CHECK-NEXT: ret;
   store i32 %x, i32* %p
   ret void
 }
 
-define ptx_device void @t2(i32* %p, i32 %x) {
+define ptx_device void @t1_u64(i64* %p, i64 %x) {
+entry:
+;CHECK: st.global.u64 [r1], rd1;
+;CHECK-NEXT: ret;
+  store i64 %x, i64* %p
+  ret void
+}
+
+define ptx_device void @t1_f32(float* %p, float %x) {
+entry:
+;CHECK: st.global.f32 [r1], f1;
+;CHECK-NEXT: ret;
+  store float %x, float* %p
+  ret void
+}
+
+define ptx_device void @t1_f64(double* %p, double %x) {
+entry:
+;CHECK: st.global.f64 [r1], fd1;
+;CHECK-NEXT: ret;
+  store double %x, double* %p
+  ret void
+}
+
+define ptx_device void @t2_u16(i16* %p, i16 %x) {
 entry:
-;CHECK: st.global.s32 [r1+4], r2;
+;CHECK: st.global.u16 [r1+2], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr i16* %p, i32 1
+  store i16 %x, i16* %i
+  ret void
+}
+
+define ptx_device void @t2_u32(i32* %p, i32 %x) {
+entry:
+;CHECK: st.global.u32 [r1+4], r2;
+;CHECK-NEXT: ret;
   %i = getelementptr i32* %p, i32 1
   store i32 %x, i32* %i
   ret void
 }
 
-define ptx_device void @t3(i32* %p, i32 %q, i32 %x) {
-;CHECK: .reg .s32 r0;
+define ptx_device void @t2_u64(i64* %p, i64 %x) {
+entry:
+;CHECK: st.global.u64 [r1+8], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr i64* %p, i32 1
+  store i64 %x, i64* %i
+  ret void
+}
+
+define ptx_device void @t2_f32(float* %p, float %x) {
+entry:
+;CHECK: st.global.f32 [r1+4], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr float* %p, i32 1
+  store float %x, float* %i
+  ret void
+}
+
+define ptx_device void @t2_f64(double* %p, double %x) {
+entry:
+;CHECK: st.global.f64 [r1+8], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr double* %p, i32 1
+  store double %x, double* %i
+  ret void
+}
+
+define ptx_device void @t3_u16(i16* %p, i32 %q, i16 %x) {
+entry:
+;CHECK: shl.b32 r0, r2, 1;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: st.global.u16 [r0], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr i16* %p, i32 %q
+  store i16 %x, i16* %i
+  ret void
+}
+
+define ptx_device void @t3_u32(i32* %p, i32 %q, i32 %x) {
 entry:
 ;CHECK: shl.b32 r0, r2, 2;
-;CHECK: add.s32 r0, r1, r0;
-;CHECK: st.global.s32 [r0], r3;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: st.global.u32 [r0], r3;
+;CHECK-NEXT: ret;
   %i = getelementptr i32* %p, i32 %q
   store i32 %x, i32* %i
   ret void
 }
 
-define ptx_device void @t4_global(i32 %x) {
+define ptx_device void @t3_u64(i64* %p, i32 %q, i64 %x) {
+entry:
+;CHECK: shl.b32 r0, r2, 3;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: st.global.u64 [r0], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr i64* %p, i32 %q
+  store i64 %x, i64* %i
+  ret void
+}
+
+define ptx_device void @t3_f32(float* %p, i32 %q, float %x) {
+entry:
+;CHECK: shl.b32 r0, r2, 2;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: st.global.f32 [r0], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr float* %p, i32 %q
+  store float %x, float* %i
+  ret void
+}
+
+define ptx_device void @t3_f64(double* %p, i32 %q, double %x) {
+entry:
+;CHECK: shl.b32 r0, r2, 3;
+;CHECK-NEXT: add.u32 r0, r1, r0;
+;CHECK-NEXT: st.global.f64 [r0], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr double* %p, i32 %q
+  store double %x, double* %i
+  ret void
+}
+
+define ptx_device void @t4_global_u16(i16 %x) {
+entry:
+;CHECK: mov.u32 r0, array_i16;
+;CHECK-NEXT: st.global.u16 [r0], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16]* @array_i16, i16 0, i16 0
+  store i16 %x, i16* %i
+  ret void
+}
+
+define ptx_device void @t4_global_u32(i32 %x) {
 entry:
-;CHECK: st.global.s32 [array], r1;
-  %i = getelementptr [10 x i32]* @array, i32 0, i32 0
+;CHECK: mov.u32 r0, array_i32;
+;CHECK-NEXT: st.global.u32 [r0], r1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32]* @array_i32, i32 0, i32 0
   store i32 %x, i32* %i
   ret void
 }
 
-define ptx_device void @t4_local(i32 %x) {
+define ptx_device void @t4_global_u64(i64 %x) {
+entry:
+;CHECK: mov.u32 r0, array_i64;
+;CHECK-NEXT: st.global.u64 [r0], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64]* @array_i64, i32 0, i32 0
+  store i64 %x, i64* %i
+  ret void
+}
+
+define ptx_device void @t4_global_f32(float %x) {
+entry:
+;CHECK: mov.u32 r0, array_float;
+;CHECK-NEXT: st.global.f32 [r0], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float]* @array_float, i32 0, i32 0
+  store float %x, float* %i
+  ret void
+}
+
+define ptx_device void @t4_global_f64(double %x) {
+entry:
+;CHECK: mov.u32 r0, array_double;
+;CHECK-NEXT: st.global.f64 [r0], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double]* @array_double, i32 0, i32 0
+  store double %x, double* %i
+  ret void
+}
+
+define ptx_device void @t4_local_u16(i16 %x) {
 entry:
-;CHECK: st.local.s32 [array_local], r1;
-  %i = getelementptr [10 x i32] addrspace(2)* @array_local, i32 0, i32 0
+;CHECK: mov.u32 r0, array_local_i16;
+;CHECK-NEXT: st.local.u16 [r0], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16] addrspace(2)* @array_local_i16, i32 0, i32 0
+  store i16 %x, i16 addrspace(2)* %i
+  ret void
+}
+
+define ptx_device void @t4_local_u32(i32 %x) {
+entry:
+;CHECK: mov.u32 r0, array_local_i32;
+;CHECK-NEXT: st.local.u32 [r0], r1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32] addrspace(2)* @array_local_i32, i32 0, i32 0
   store i32 %x, i32 addrspace(2)* %i
   ret void
 }
 
-define ptx_device void @t4_shared(i32 %x) {
+define ptx_device void @t4_local_u64(i64 %x) {
+entry:
+;CHECK: mov.u32 r0, array_local_i64;
+;CHECK-NEXT: st.local.u64 [r0], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64] addrspace(2)* @array_local_i64, i32 0, i32 0
+  store i64 %x, i64 addrspace(2)* %i
+  ret void
+}
+
+define ptx_device void @t4_local_f32(float %x) {
+entry:
+;CHECK: mov.u32 r0, array_local_float;
+;CHECK-NEXT: st.local.f32 [r0], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float] addrspace(2)* @array_local_float, i32 0, i32 0
+  store float %x, float addrspace(2)* %i
+  ret void
+}
+
+define ptx_device void @t4_local_f64(double %x) {
+entry:
+;CHECK: mov.u32 r0, array_local_double;
+;CHECK-NEXT: st.local.f64 [r0], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double] addrspace(2)* @array_local_double, i32 0, i32 0
+  store double %x, double addrspace(2)* %i
+  ret void
+}
+
+define ptx_device void @t4_shared_u16(i16 %x) {
 entry:
-;CHECK: st.shared.s32 [array_shared], r1;
-  %i = getelementptr [10 x i32] addrspace(4)* @array_shared, i32 0, i32 0
+;CHECK: mov.u32 r0, array_shared_i16;
+;CHECK-NEXT: st.shared.u16 [r0], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16] addrspace(4)* @array_shared_i16, i32 0, i32 0
+  store i16 %x, i16 addrspace(4)* %i
+  ret void
+}
+
+define ptx_device void @t4_shared_u32(i32 %x) {
+entry:
+;CHECK: mov.u32 r0, array_shared_i32;
+;CHECK-NEXT: st.shared.u32 [r0], r1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32] addrspace(4)* @array_shared_i32, i32 0, i32 0
   store i32 %x, i32 addrspace(4)* %i
   ret void
 }
 
-define ptx_device void @t5(i32 %x) {
+define ptx_device void @t4_shared_u64(i64 %x) {
+entry:
+;CHECK: mov.u32 r0, array_shared_i64;
+;CHECK-NEXT: st.shared.u64 [r0], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64] addrspace(4)* @array_shared_i64, i32 0, i32 0
+  store i64 %x, i64 addrspace(4)* %i
+  ret void
+}
+
+define ptx_device void @t4_shared_f32(float %x) {
+entry:
+;CHECK: mov.u32 r0, array_shared_float;
+;CHECK-NEXT: st.shared.f32 [r0], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float] addrspace(4)* @array_shared_float, i32 0, i32 0
+  store float %x, float addrspace(4)* %i
+  ret void
+}
+
+define ptx_device void @t4_shared_f64(double %x) {
+entry:
+;CHECK: mov.u32 r0, array_shared_double;
+;CHECK-NEXT: st.shared.f64 [r0], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double] addrspace(4)* @array_shared_double, i32 0, i32 0
+  store double %x, double addrspace(4)* %i
+  ret void
+}
+
+define ptx_device void @t5_u16(i16 %x) {
 entry:
-;CHECK: st.global.s32 [array+4], r1;
-  %i = getelementptr [10 x i32]* @array, i32 0, i32 1
+;CHECK: mov.u32 r0, array_i16;
+;CHECK-NEXT: st.global.u16 [r0+2], rh1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i16]* @array_i16, i32 0, i32 1
+  store i16 %x, i16* %i
+  ret void
+}
+
+define ptx_device void @t5_u32(i32 %x) {
+entry:
+;CHECK: mov.u32 r0, array_i32;
+;CHECK-NEXT: st.global.u32 [r0+4], r1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i32]* @array_i32, i32 0, i32 1
   store i32 %x, i32* %i
   ret void
 }
+
+define ptx_device void @t5_u64(i64 %x) {
+entry:
+;CHECK: mov.u32 r0, array_i64;
+;CHECK-NEXT: st.global.u64 [r0+8], rd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x i64]* @array_i64, i32 0, i32 1
+  store i64 %x, i64* %i
+  ret void
+}
+
+define ptx_device void @t5_f32(float %x) {
+entry:
+;CHECK: mov.u32 r0, array_float;
+;CHECK-NEXT: st.global.f32 [r0+4], f1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x float]* @array_float, i32 0, i32 1
+  store float %x, float* %i
+  ret void
+}
+
+define ptx_device void @t5_f64(double %x) {
+entry:
+;CHECK: mov.u32 r0, array_double;
+;CHECK-NEXT: st.global.f64 [r0+8], fd1;
+;CHECK-NEXT: ret;
+  %i = getelementptr [10 x double]* @array_double, i32 0, i32 1
+  store double %x, double* %i
+  ret void
+}
diff --git a/test/CodeGen/PTX/sub.ll b/test/CodeGen/PTX/sub.ll
index aab3fdadad13..7dd2c6f6ac79 100644
--- a/test/CodeGen/PTX/sub.ll
+++ b/test/CodeGen/PTX/sub.ll
@@ -1,15 +1,71 @@
-; RUN: llc < %s -march=ptx | FileCheck %s
+; RUN: llc < %s -march=ptx32 | FileCheck %s
 
-define ptx_device i32 @t1(i32 %x, i32 %y) {
-;CHECK: sub.s32 r0, r1, r2;
+define ptx_device i16 @t1_u16(i16 %x, i16 %y) {
+; CHECK: sub.u16 rh0, rh1, rh2;
+; CHECK-NEXT: ret;
+	%z = sub i16 %x, %y
+	ret i16 %z
+}
+
+define ptx_device i32 @t1_u32(i32 %x, i32 %y) {
+; CHECK: sub.u32 r0, r1, r2;
+; CHECK-NEXT: ret;
 	%z = sub i32 %x, %y
-;CHECK: ret;
 	ret i32 %z
 }
 
-define ptx_device i32 @t2(i32 %x) {
-;CHECK: add.s32 r0, r1, -1;
+define ptx_device i64 @t1_u64(i64 %x, i64 %y) {
+; CHECK: sub.u64 rd0, rd1, rd2;
+; CHECK-NEXT: ret;
+	%z = sub i64 %x, %y
+	ret i64 %z
+}
+
+define ptx_device float @t1_f32(float %x, float %y) {
+; CHECK: sub.f32 f0, f1, f2
+; CHECK-NEXT: ret;
+  %z = fsub float %x, %y
+  ret float %z
+}
+
+define ptx_device double @t1_f64(double %x, double %y) {
+; CHECK: sub.f64 fd0, fd1, fd2
+; CHECK-NEXT: ret;
+  %z = fsub double %x, %y
+  ret double %z
+}
+
+define ptx_device i16 @t2_u16(i16 %x) {
+; CHECK: add.u16 rh0, rh1, -1;
+; CHECK-NEXT: ret;
+	%z = sub i16 %x, 1
+	ret i16 %z
+}
+
+define ptx_device i32 @t2_u32(i32 %x) {
+; CHECK: add.u32 r0, r1, -1;
+; CHECK-NEXT: ret;
 	%z = sub i32 %x, 1
-;CHECK: ret;
 	ret i32 %z
 }
+
+define ptx_device i64 @t2_u64(i64 %x) {
+; CHECK: add.u64 rd0, rd1, -1;
+; CHECK-NEXT: ret;
+	%z = sub i64 %x, 1
+	ret i64 %z
+}
+
+define ptx_device float @t2_f32(float %x) {
+; CHECK: add.f32 f0, f1, 0FBF800000;
+; CHECK-NEXT: ret;
+  %z = fsub float %x, 1.0
+  ret float %z
+}
+
+define ptx_device double @t2_f64(double %x) {
+; CHECK: add.f64 fd0, fd1, 0DBFF0000000000000;
+; CHECK-NEXT: ret;
+  %z = fsub double %x, 1.0
+  ret double %z
+}