R600/SI: Add support for work item and work group intrinsics

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183138 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 211 insertions, 0 deletions

diff --git a/test/CodeGen/R600/work-item-intrinsics.ll b/test/CodeGen/R600/work-item-intrinsics.ll
new file mode 100644
index 00000000000..46e3e54b9a2
--- /dev/null
+++ b/test/CodeGen/R600/work-item-intrinsics.ll
@@ -0,0 +1,211 @@
+; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=R600-CHECK %s
+; RUN: llc < %s -march=r600 -mcpu=SI | FileCheck --check-prefix=SI-CHECK %s
+
+; R600-CHECK: @ngroups_x
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 0
+; SI-CHECK: @ngroups_x
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 0
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @ngroups_x (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.ngroups.x() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @ngroups_y
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 4
+; SI-CHECK: @ngroups_y
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 1
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @ngroups_y (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.ngroups.y() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @ngroups_z
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 8
+; SI-CHECK: @ngroups_z
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 2
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @ngroups_z (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.ngroups.z() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @global_size_x
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 12
+; SI-CHECK: @global_size_x
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 3
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @global_size_x (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.global.size.x() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @global_size_y
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 16
+; SI-CHECK: @global_size_y
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 4
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @global_size_y (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.global.size.y() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @global_size_z
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 20
+; SI-CHECK: @global_size_z
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 5
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @global_size_z (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.global.size.z() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @local_size_x
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 24
+; SI-CHECK: @local_size_x
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 6
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @local_size_x (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.local.size.x() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @local_size_y
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 28
+; SI-CHECK: @local_size_y
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 7
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @local_size_y (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.local.size.y() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; R600-CHECK: @local_size_z
+; R600-CHECK: RAT_WRITE_CACHELESS_32_eg [[VAL:T[0-9]+\.X]]
+; R600-CHECK: VTX_READ_32 [[VAL]], [[VAL]], 32
+; SI-CHECK: @local_size_z
+; SI-CHECK: S_LOAD_DWORD [[VAL:SGPR[0-9]+]], SGPR0_SGPR1, 8
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], [[VAL]]
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @local_size_z (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.local.size.z() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; The tgid values are stored in SGPRs offset by the number of user SGPRs.
+; Currently we always use exactly 2 user SGPRs for the pointer to the
+; kernel arguments, but this may change in the future.
+
+; SI-CHECK: @tgid_x
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], SGPR2
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @tgid_x (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tgid.x() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; SI-CHECK: @tgid_y
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], SGPR3
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @tgid_y (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tgid.y() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; SI-CHECK: @tgid_z
+; SI-CHECK: V_MOV_B32_e32 [[VVAL:VGPR[0-9]+]], SGPR4
+; SI-CHECK: BUFFER_STORE_DWORD [[VVAL]]
+define void @tgid_z (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tgid.z() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; SI-CHECK: @tidig_x
+; SI-CHECK: BUFFER_STORE_DWORD VGPR0
+define void @tidig_x (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tidig.x() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; SI-CHECK: @tidig_y
+; SI-CHECK: BUFFER_STORE_DWORD VGPR1
+define void @tidig_y (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tidig.y() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+; SI-CHECK: @tidig_z
+; SI-CHECK: BUFFER_STORE_DWORD VGPR2
+define void @tidig_z (i32 addrspace(1)* %out) {
+entry:
+  %0 = call i32 @llvm.r600.read.tidig.z() #0
+  store i32 %0, i32 addrspace(1)* %out
+  ret void
+}
+
+declare i32 @llvm.r600.read.ngroups.x() #0
+declare i32 @llvm.r600.read.ngroups.y() #0
+declare i32 @llvm.r600.read.ngroups.z() #0
+
+declare i32 @llvm.r600.read.global.size.x() #0
+declare i32 @llvm.r600.read.global.size.y() #0
+declare i32 @llvm.r600.read.global.size.z() #0
+
+declare i32 @llvm.r600.read.local.size.x() #0
+declare i32 @llvm.r600.read.local.size.y() #0
+declare i32 @llvm.r600.read.local.size.z() #0
+
+declare i32 @llvm.r600.read.tgid.x() #0
+declare i32 @llvm.r600.read.tgid.y() #0
+declare i32 @llvm.r600.read.tgid.z() #0
+
+declare i32 @llvm.r600.read.tidig.x() #0
+declare i32 @llvm.r600.read.tidig.y() #0
+declare i32 @llvm.r600.read.tidig.z() #0
+
+attributes #0 = { readnone }