Add R600 backend

A new backend supporting AMD GPUs: Radeon HD2XXX - HD7XXX

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@169915 91177308-0d34-0410-b5e6-96231b3b80d8

39 files changed, 636 insertions, 0 deletions

diff --git a/test/CodeGen/R600/add.v4i32.ll b/test/CodeGen/R600/add.v4i32.ll
new file mode 100644
index 0000000..ac4a874
--- /dev/null
+++ b/test/CodeGen/R600/add.v4i32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: ADD_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x i32> addrspace(1)* %in, i32 1
+  %a = load <4 x i32> addrspace(1) * %in
+  %b = load <4 x i32> addrspace(1) * %b_ptr
+  %result = add <4 x i32> %a, %b
+  store <4 x i32> %result, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/and.v4i32.ll b/test/CodeGen/R600/and.v4i32.ll
new file mode 100644
index 0000000..662085e
--- /dev/null
+++ b/test/CodeGen/R600/and.v4i32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: AND_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: AND_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: AND_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: AND_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x i32> addrspace(1)* %in, i32 1
+  %a = load <4 x i32> addrspace(1) * %in
+  %b = load <4 x i32> addrspace(1) * %b_ptr
+  %result = and <4 x i32> %a, %b
+  store <4 x i32> %result, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fabs.ll b/test/CodeGen/R600/fabs.ll
new file mode 100644
index 0000000..0407533
--- /dev/null
+++ b/test/CodeGen/R600/fabs.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: MOV T{{[0-9]+\.[XYZW], \|T[0-9]+\.[XYZW]\|}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @fabs( float %r0)
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @fabs(float ) readnone
diff --git a/test/CodeGen/R600/fadd.ll b/test/CodeGen/R600/fadd.ll
new file mode 100644
index 0000000..d7d1b65
--- /dev/null
+++ b/test/CodeGen/R600/fadd.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = fadd float %r0, %r1
+   call void @llvm.AMDGPU.store.output(float %r2, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
diff --git a/test/CodeGen/R600/fadd.v4f32.ll b/test/CodeGen/R600/fadd.v4f32.ll
new file mode 100644
index 0000000..85dbfd5
--- /dev/null
+++ b/test/CodeGen/R600/fadd.v4f32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
+  %a = load <4 x float> addrspace(1) * %in
+  %b = load <4 x float> addrspace(1) * %b_ptr
+  %result = fadd <4 x float> %a, %b
+  store <4 x float> %result, <4 x float> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fcmp-cnd.ll b/test/CodeGen/R600/fcmp-cnd.ll
new file mode 100644
index 0000000..a94cfb5
--- /dev/null
+++ b/test/CodeGen/R600/fcmp-cnd.ll
@@ -0,0 +1,14 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;Not checking arguments 2 and 3 to CNDE, because they may change between
+;registers and literal.x depending on what the optimizer does.
+;CHECK: CNDE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(i32 addrspace(1)* %out, float addrspace(1)* %in) {
+entry:
+  %0 = load float addrspace(1)* %in
+  %cmp = fcmp oeq float %0, 0.000000e+00
+  %value = select i1 %cmp, i32 2, i32 3 
+  store i32 %value, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fcmp-cnde-int-args.ll b/test/CodeGen/R600/fcmp-cnde-int-args.ll
new file mode 100644
index 0000000..5c981ef
--- /dev/null
+++ b/test/CodeGen/R600/fcmp-cnde-int-args.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; This test checks a bug in R600TargetLowering::LowerSELECT_CC where the
+; chance to optimize the fcmp + select instructions to CNDE was missed
+; due to the fact that the operands to fcmp and select had different types
+
+;CHECK: CNDE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], literal.x, 0.0, -1}}
+
+define void @test(i32 addrspace(1)* %out, float addrspace(1)* %in) {
+entry:
+  %0 = load float addrspace(1)* %in
+  %cmp = fcmp oeq float %0, 0.000000e+00
+  %value = select i1 %cmp, i32 -1, i32 0
+  store i32 %value, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fcmp.ll b/test/CodeGen/R600/fcmp.ll
new file mode 100644
index 0000000..1dcd07c
--- /dev/null
+++ b/test/CodeGen/R600/fcmp.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: SETE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MOV T{{[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
+;CHECK: FLT_TO_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(i32 addrspace(1)* %out, float addrspace(1)* %in) {
+entry:
+  %0 = load float addrspace(1)* %in
+  %arrayidx1 = getelementptr inbounds float addrspace(1)* %in, i32 1
+  %1 = load float addrspace(1)* %arrayidx1
+  %cmp = fcmp oeq float %0, %1
+  %sext = sext i1 %cmp to i32
+  store i32 %sext, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fdiv.v4f32.ll b/test/CodeGen/R600/fdiv.v4f32.ll
new file mode 100644
index 0000000..b013fd6
--- /dev/null
+++ b/test/CodeGen/R600/fdiv.v4f32.ll
@@ -0,0 +1,19 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: RECIP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL NON-IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: RECIP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL NON-IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: RECIP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL NON-IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: RECIP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL NON-IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
+  %a = load <4 x float> addrspace(1) * %in
+  %b = load <4 x float> addrspace(1) * %b_ptr
+  %result = fdiv <4 x float> %a, %b
+  store <4 x float> %result, <4 x float> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/floor.ll b/test/CodeGen/R600/floor.ll
new file mode 100644
index 0000000..845330f
--- /dev/null
+++ b/test/CodeGen/R600/floor.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: FLOOR T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @floor(float %r0)
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @floor(float) readonly
diff --git a/test/CodeGen/R600/fmax.ll b/test/CodeGen/R600/fmax.ll
new file mode 100644
index 0000000..3708f0b
--- /dev/null
+++ b/test/CodeGen/R600/fmax.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: MAX T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = fcmp uge float %r0, %r1
+   %r3 = select i1 %r2, float %r0, float %r1
+   call void @llvm.AMDGPU.store.output(float %r3, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
diff --git a/test/CodeGen/R600/fmin.ll b/test/CodeGen/R600/fmin.ll
new file mode 100644
index 0000000..19d59ab
--- /dev/null
+++ b/test/CodeGen/R600/fmin.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: MIN T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = fcmp uge float %r0, %r1
+   %r3 = select i1 %r2, float %r1, float %r0
+   call void @llvm.AMDGPU.store.output(float %r3, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
diff --git a/test/CodeGen/R600/fmul.ll b/test/CodeGen/R600/fmul.ll
new file mode 100644
index 0000000..eb1d523
--- /dev/null
+++ b/test/CodeGen/R600/fmul.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; CHECK: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = fmul float %r0, %r1
+   call void @llvm.AMDGPU.store.output(float %r2, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
diff --git a/test/CodeGen/R600/fmul.v4f32.ll b/test/CodeGen/R600/fmul.v4f32.ll
new file mode 100644
index 0000000..6d44a0c
--- /dev/null
+++ b/test/CodeGen/R600/fmul.v4f32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
+  %a = load <4 x float> addrspace(1) * %in
+  %b = load <4 x float> addrspace(1) * %b_ptr
+  %result = fmul <4 x float> %a, %b
+  store <4 x float> %result, <4 x float> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/fsub.ll b/test/CodeGen/R600/fsub.ll
new file mode 100644
index 0000000..0ec1c37
--- /dev/null
+++ b/test/CodeGen/R600/fsub.ll
@@ -0,0 +1,17 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; CHECK: MOV T{{[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
+; CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = fsub float %r0, %r1
+   call void @llvm.AMDGPU.store.output(float %r2, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
diff --git a/test/CodeGen/R600/fsub.v4f32.ll b/test/CodeGen/R600/fsub.v4f32.ll
new file mode 100644
index 0000000..612a57e
--- /dev/null
+++ b/test/CodeGen/R600/fsub.v4f32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: ADD T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
+  %a = load <4 x float> addrspace(1) * %in
+  %b = load <4 x float> addrspace(1) * %b_ptr
+  %result = fsub <4 x float> %a, %b
+  store <4 x float> %result, <4 x float> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/i8_to_double_to_float.ll b/test/CodeGen/R600/i8_to_double_to_float.ll
new file mode 100644
index 0000000..39f3322
--- /dev/null
+++ b/test/CodeGen/R600/i8_to_double_to_float.ll
@@ -0,0 +1,11 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: UINT_TO_FLT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(float addrspace(1)* %out, i8 addrspace(1)* %in) {
+  %1 = load i8 addrspace(1)* %in
+  %2 = uitofp i8 %1 to double
+  %3 = fptrunc double %2 to float
+  store float %3, float addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/icmp-select-sete-reverse-args.ll b/test/CodeGen/R600/icmp-select-sete-reverse-args.ll
new file mode 100644
index 0000000..aad44d9
--- /dev/null
+++ b/test/CodeGen/R600/icmp-select-sete-reverse-args.ll
@@ -0,0 +1,18 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;Test that a select with reversed True/False values is correctly lowered
+;to a SETNE_INT.  There should only be one SETNE_INT instruction.
+
+;CHECK: SETNE_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK_NOT: SETNE_INT
+
+define void @test(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
+entry:
+  %0 = load i32 addrspace(1)* %in
+  %arrayidx1 = getelementptr inbounds i32 addrspace(1)* %in, i32 1
+  %1 = load i32 addrspace(1)* %arrayidx1
+  %cmp = icmp eq i32 %0, %1
+  %value = select i1 %cmp, i32 0, i32 -1
+  store i32 %value, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/lit.local.cfg b/test/CodeGen/R600/lit.local.cfg
new file mode 100644
index 0000000..36ee493
--- /dev/null
+++ b/test/CodeGen/R600/lit.local.cfg
@@ -0,0 +1,13 @@
+config.suffixes = ['.ll', '.c', '.cpp']
+
+def getRoot(config):
+    if not config.parent:
+        return config
+    return getRoot(config.parent)
+
+root = getRoot(config)
+
+targets = set(root.targets_to_build.split())
+if not 'R600' in targets:
+    config.unsupported = True
+
diff --git a/test/CodeGen/R600/literals.ll b/test/CodeGen/R600/literals.ll
new file mode 100644
index 0000000..4c731b2
--- /dev/null
+++ b/test/CodeGen/R600/literals.ll
@@ -0,0 +1,30 @@
+; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; Test using an integer literal constant.
+; Generated ASM should be:
+; ADD_INT REG literal.x, 5
+; or
+; ADD_INT literal.x REG, 5
+
+; CHECK: ADD_INT {{[A-Z0-9,. ]*}}literal.x,{{[A-Z0-9,. ]*}} 5
+define void @i32_literal(i32 addrspace(1)* %out, i32 %in) {
+entry:
+  %0 = add i32 5, %in
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; Test using a float literal constant.
+; Generated ASM should be:
+; ADD REG literal.x, 5.0
+; or
+; ADD literal.x REG, 5.0
+
+; CHECK: ADD {{[A-Z0-9,. ]*}}literal.x,{{[A-Z0-9,. ]*}} {{[0-9]+}}(5.0
+define void @float_literal(float addrspace(1)* %out, float %in) {
+entry:
+  %0 = fadd float 5.0, %in
+  store float %0, float addrspace(1)* %out
+  ret void
+}
+
diff --git a/test/CodeGen/R600/llvm.AMDGPU.mul.ll b/test/CodeGen/R600/llvm.AMDGPU.mul.ll
new file mode 100644
index 0000000..693eb27
--- /dev/null
+++ b/test/CodeGen/R600/llvm.AMDGPU.mul.ll
@@ -0,0 +1,17 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: MUL NON-IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = call float @llvm.AMDGPU.mul( float %r0, float %r1)
+   call void @llvm.AMDGPU.store.output(float %r2, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @llvm.AMDGPU.mul(float ,float ) readnone
diff --git a/test/CodeGen/R600/llvm.AMDGPU.trunc.ll b/test/CodeGen/R600/llvm.AMDGPU.trunc.ll
new file mode 100644
index 0000000..fac957f
--- /dev/null
+++ b/test/CodeGen/R600/llvm.AMDGPU.trunc.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: TRUNC T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.AMDGPU.trunc( float %r0)
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @llvm.AMDGPU.trunc(float ) readnone
diff --git a/test/CodeGen/R600/llvm.cos.ll b/test/CodeGen/R600/llvm.cos.ll
new file mode 100644
index 0000000..dc120bf
--- /dev/null
+++ b/test/CodeGen/R600/llvm.cos.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: COS T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.cos.f32(float %r0)
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.cos.f32(float) readnone
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
diff --git a/test/CodeGen/R600/llvm.pow.ll b/test/CodeGen/R600/llvm.pow.ll
new file mode 100644
index 0000000..0ae9172
--- /dev/null
+++ b/test/CodeGen/R600/llvm.pow.ll
@@ -0,0 +1,19 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: LOG_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK-NEXT: MUL_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK-NEXT: EXP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.R600.load.input(i32 1)
+   %r2 = call float @llvm.pow.f32( float %r0, float %r1)
+   call void @llvm.AMDGPU.store.output(float %r2, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @llvm.pow.f32(float ,float ) readonly
diff --git a/test/CodeGen/R600/llvm.sin.ll b/test/CodeGen/R600/llvm.sin.ll
new file mode 100644
index 0000000..5cd6998
--- /dev/null
+++ b/test/CodeGen/R600/llvm.sin.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: SIN T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = call float @llvm.sin.f32( float %r0)
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.sin.f32(float) readnone
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
diff --git a/test/CodeGen/R600/load.constant_addrspace.f32.ll b/test/CodeGen/R600/load.constant_addrspace.f32.ll
new file mode 100644
index 0000000..9362728
--- /dev/null
+++ b/test/CodeGen/R600/load.constant_addrspace.f32.ll
@@ -0,0 +1,9 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: VTX_READ_32 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @test(float addrspace(1)* %out, float addrspace(2)* %in) {
+  %1 = load float addrspace(2)* %in
+  store float %1, float addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/load.i8.ll b/test/CodeGen/R600/load.i8.ll
new file mode 100644
index 0000000..b070dcd
--- /dev/null
+++ b/test/CodeGen/R600/load.i8.ll
@@ -0,0 +1,10 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: VTX_READ_8 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @test(i32 addrspace(1)* %out, i8 addrspace(1)* %in) {
+  %1 = load i8 addrspace(1)* %in
+  %2 = zext i8 %1 to i32
+  store i32 %2, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/reciprocal.ll b/test/CodeGen/R600/reciprocal.ll
new file mode 100644
index 0000000..6838c1a
--- /dev/null
+++ b/test/CodeGen/R600/reciprocal.ll
@@ -0,0 +1,16 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: RECIP_IEEE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test() {
+   %r0 = call float @llvm.R600.load.input(i32 0)
+   %r1 = fdiv float 1.0, %r0
+   call void @llvm.AMDGPU.store.output(float %r1, i32 0)
+   ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)
+
+declare float @llvm.AMDGPU.rcp(float ) readnone
diff --git a/test/CodeGen/R600/sdiv.ll b/test/CodeGen/R600/sdiv.ll
new file mode 100644
index 0000000..3556fac
--- /dev/null
+++ b/test/CodeGen/R600/sdiv.ll
@@ -0,0 +1,21 @@
+; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; The code generated by sdiv is long and complex and may frequently change.
+; The goal of this test is to make sure the ISel doesn't fail.
+;
+; This program was previously failing to compile when one of the selectcc
+; opcodes generated by the sdiv lowering was being legalized and optimized to:
+; selectcc Remainder -1, 0, -1, SETGT
+; This was fixed by adding an additional pattern in R600Instructions.td to
+; match this pattern with a CNDGE_INT.
+
+; CHECK: RETURN
+
+define void @test(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
+  %den_ptr = getelementptr i32 addrspace(1)* %in, i32 1
+  %num = load i32 addrspace(1) * %in
+  %den = load i32 addrspace(1) * %den_ptr
+  %result = sdiv i32 %num, %den
+  store i32 %result, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/selectcc-icmp-select-float.ll b/test/CodeGen/R600/selectcc-icmp-select-float.ll
new file mode 100644
index 0000000..f65a300
--- /dev/null
+++ b/test/CodeGen/R600/selectcc-icmp-select-float.ll
@@ -0,0 +1,15 @@
+; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; Note additional optimizations may cause this SGT to be replaced with a
+; CND* instruction.
+; CHECK: SGT_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], literal.x, -1}}
+; Test a selectcc with i32 LHS/RHS and float True/False
+
+define void @test(float addrspace(1)* %out, i32 addrspace(1)* %in) {
+entry:
+  %0 = load i32 addrspace(1)* %in
+  %1 = icmp sge i32 %0, 0
+  %2 = select i1 %1, float 1.0, float 0.0
+  store float %2, float addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/selectcc_cnde.ll b/test/CodeGen/R600/selectcc_cnde.ll
new file mode 100644
index 0000000..f0a0f51
--- /dev/null
+++ b/test/CodeGen/R600/selectcc_cnde.ll
@@ -0,0 +1,11 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK-NOT: SETE
+;CHECK: CNDE T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], 1.0, literal.x, [-0-9]+\(2.0}}
+define void @test(float addrspace(1)* %out, float addrspace(1)* %in) {
+  %1 = load float addrspace(1)* %in
+  %2 = fcmp oeq float %1, 0.0
+  %3 = select i1 %2, float 1.0, float 2.0
+  store float %3, float addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/selectcc_cnde_int.ll b/test/CodeGen/R600/selectcc_cnde_int.ll
new file mode 100644
index 0000000..b38078e
--- /dev/null
+++ b/test/CodeGen/R600/selectcc_cnde_int.ll
@@ -0,0 +1,11 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK-NOT: SETE_INT
+;CHECK: CNDE_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], 1, literal.x, 2}}
+define void @test(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
+  %1 = load i32 addrspace(1)* %in
+  %2 = icmp eq i32 %1, 0
+  %3 = select i1 %2, i32 1, i32 2
+  store i32 %3, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/setcc.v4i32.ll b/test/CodeGen/R600/setcc.v4i32.ll
new file mode 100644
index 0000000..0752f2e
--- /dev/null
+++ b/test/CodeGen/R600/setcc.v4i32.ll
@@ -0,0 +1,12 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+;CHECK: SETE_INT T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x i32> addrspace(1)* %in, i32 1
+  %a = load <4 x i32> addrspace(1) * %in
+  %b = load <4 x i32> addrspace(1) * %b_ptr
+  %result = icmp eq <4 x i32> %a, %b
+  %sext = sext <4 x i1> %result to <4 x i32>
+  store <4 x i32> %sext, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/short-args.ll b/test/CodeGen/R600/short-args.ll
new file mode 100644
index 0000000..1070250
--- /dev/null
+++ b/test/CodeGen/R600/short-args.ll
@@ -0,0 +1,37 @@
+; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+; CHECK: VTX_READ_8 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @i8_arg(i32 addrspace(1)* nocapture %out, i8 %in) nounwind {
+entry:
+  %0 = zext i8 %in to i32
+  store i32 %0, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
+; CHECK: VTX_READ_8 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @i8_zext_arg(i32 addrspace(1)* nocapture %out, i8 zeroext %in) nounwind {
+entry:
+  %0 = zext i8 %in to i32
+  store i32 %0, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
+; CHECK: VTX_READ_16 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @i16_arg(i32 addrspace(1)* nocapture %out, i16 %in) nounwind {
+entry:
+  %0 = zext i16 %in to i32
+  store i32 %0, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
+; CHECK: VTX_READ_16 T{{[0-9]+\.X, T[0-9]+\.X}}
+
+define void @i16_zext_arg(i32 addrspace(1)* nocapture %out, i16 zeroext %in) nounwind {
+entry:
+  %0 = zext i16 %in to i32
+  store i32 %0, i32 addrspace(1)* %out, align 4
+  ret void
+}
diff --git a/test/CodeGen/R600/store.v4f32.ll b/test/CodeGen/R600/store.v4f32.ll
new file mode 100644
index 0000000..8b0d244
--- /dev/null
+++ b/test/CodeGen/R600/store.v4f32.ll
@@ -0,0 +1,9 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: RAT_WRITE_CACHELESS_128 T{{[0-9]+\.XYZW, T[0-9]+\.X}}, 1
+
+define void @test(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
+  %1 = load <4 x float> addrspace(1) * %in
+  store <4 x float> %1, <4 x float> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/store.v4i32.ll b/test/CodeGen/R600/store.v4i32.ll
new file mode 100644
index 0000000..a659815
--- /dev/null
+++ b/test/CodeGen/R600/store.v4i32.ll
@@ -0,0 +1,9 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;CHECK: RAT_WRITE_CACHELESS_128 T{{[0-9]+\.XYZW, T[0-9]+\.X}}, 1
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %1 = load <4 x i32> addrspace(1) * %in
+  store <4 x i32> %1, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/udiv.v4i32.ll b/test/CodeGen/R600/udiv.v4i32.ll
new file mode 100644
index 0000000..47657a6
--- /dev/null
+++ b/test/CodeGen/R600/udiv.v4i32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;The code generated by udiv is long and complex and may frequently change.
+;The goal of this test is to make sure the ISel doesn't fail when it gets
+;a v4i32 udiv
+;CHECK: RETURN
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x i32> addrspace(1)* %in, i32 1
+  %a = load <4 x i32> addrspace(1) * %in
+  %b = load <4 x i32> addrspace(1) * %b_ptr
+  %result = udiv <4 x i32> %a, %b
+  store <4 x i32> %result, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/R600/urem.v4i32.ll b/test/CodeGen/R600/urem.v4i32.ll
new file mode 100644
index 0000000..2e7388c
--- /dev/null
+++ b/test/CodeGen/R600/urem.v4i32.ll
@@ -0,0 +1,15 @@
+;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+
+;The code generated by urem is long and complex and may frequently change.
+;The goal of this test is to make sure the ISel doesn't fail when it gets
+;a v4i32 urem
+;CHECK: RETURN
+
+define void @test(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
+  %b_ptr = getelementptr <4 x i32> addrspace(1)* %in, i32 1
+  %a = load <4 x i32> addrspace(1) * %in
+  %b = load <4 x i32> addrspace(1) * %b_ptr
+  %result = urem <4 x i32> %a, %b
+  store <4 x i32> %result, <4 x i32> addrspace(1)* %out
+  ret void
+}
diff --git a/test/CodeGen/SI/sanity.ll b/test/CodeGen/SI/sanity.ll
new file mode 100644
index 0000000..62cdcf5
--- /dev/null
+++ b/test/CodeGen/SI/sanity.ll
@@ -0,0 +1,37 @@
+;RUN: llc < %s -march=r600 -mcpu=SI | FileCheck %s
+
+; CHECK: S_ENDPGM
+
+define void @main() {
+main_body:
+  call void @llvm.AMDGPU.shader.type(i32 1)
+  %0 = load <4 x i32> addrspace(2)* addrspace(8)* inttoptr (i32 6 to <4 x i32> addrspace(2)* addrspace(8)*)
+  %1 = getelementptr <4 x i32> addrspace(2)* %0, i32 0
+  %2 = load <4 x i32> addrspace(2)* %1
+  %3 = call i32 @llvm.SI.vs.load.buffer.index()
+  %4 = call <4 x float> @llvm.SI.vs.load.input(<4 x i32> %2, i32 0, i32 %3)
+  %5 = extractelement <4 x float> %4, i32 0
+  %6 = extractelement <4 x float> %4, i32 1
+  %7 = extractelement <4 x float> %4, i32 2
+  %8 = extractelement <4 x float> %4, i32 3
+  %9 = load <4 x i32> addrspace(2)* addrspace(8)* inttoptr (i32 6 to <4 x i32> addrspace(2)* addrspace(8)*)
+  %10 = getelementptr <4 x i32> addrspace(2)* %9, i32 1
+  %11 = load <4 x i32> addrspace(2)* %10
+  %12 = call i32 @llvm.SI.vs.load.buffer.index()
+  %13 = call <4 x float> @llvm.SI.vs.load.input(<4 x i32> %11, i32 0, i32 %12)
+  %14 = extractelement <4 x float> %13, i32 0
+  %15 = extractelement <4 x float> %13, i32 1
+  %16 = extractelement <4 x float> %13, i32 2
+  %17 = extractelement <4 x float> %13, i32 3
+  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %14, float %15, float %16, float %17)
+  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %5, float %6, float %7, float %8)
+  ret void
+}
+
+declare void @llvm.AMDGPU.shader.type(i32)
+
+declare i32 @llvm.SI.vs.load.buffer.index() readnone
+
+declare <4 x float> @llvm.SI.vs.load.input(<4 x i32>, i32, i32)
+
+declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)