[llvm-branch-commits] [AMDGPU] Baseline test for ptrtoaddr code generation (PR #143813)

[Matt Arsenault via llvm-branch-commits]

================
@@ -0,0 +1,118 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -global-isel -verify-machineinstrs < %s | FileCheck %s
+;; Check that we can lower ptrtoaddr differently from ptrtoint.
+;; Includes an ignored argument so the registers actually need to be written
+;; Also all functions are zeroext to show that non-address bits are masked off
+;; instead of filling them with arbitrary data
+
+define zeroext i128 @ptrtoint(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoint:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    v_mov_b32_e32 v2, v6
+; CHECK-NEXT:    v_mov_b32_e32 v3, v7
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoint ptr addrspace(8) %ptr to i128
+  ret i128 %ret
+}
+
+define zeroext i48 @ptrtoaddr(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoaddr:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_and_b32_e32 v1, 0xffff, v5
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoaddr ptr addrspace(8) %ptr to i48
+  ret i48 %ret
+}
+
+define <2 x i128> @ptrtoint_vec(ptr addrspace(8) %ignored, <2 x ptr addrspace(8)> %ptr) {
+; CHECK-LABEL: ptrtoint_vec:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    v_mov_b32_e32 v2, v6
+; CHECK-NEXT:    v_mov_b32_e32 v3, v7
+; CHECK-NEXT:    v_mov_b32_e32 v4, v8
+; CHECK-NEXT:    v_mov_b32_e32 v5, v9
+; CHECK-NEXT:    v_mov_b32_e32 v6, v10
+; CHECK-NEXT:    v_mov_b32_e32 v7, v11
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoint <2 x ptr addrspace(8)> %ptr to <2 x i128>
+  ret <2 x i128> %ret
+}
+
+;; Note: needs to be 2 x i64 instead of 2 x i48 since the latter is not supported.
+define <2 x i64> @ptrtoaddr_vec(ptr addrspace(8) %ignored, <2 x ptr addrspace(8)> %ptr) {
+; CHECK-LABEL: ptrtoaddr_vec:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    v_mov_b32_e32 v2, v8
+; CHECK-NEXT:    v_mov_b32_e32 v3, v9
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoaddr <2 x ptr addrspace(8)> %ptr to <2 x i64>
+  ret <2 x i64> %ret
+}
+
+define zeroext i256 @ptrtoint_ext(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoint_ext:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    v_mov_b32_e32 v2, v6
+; CHECK-NEXT:    v_mov_b32_e32 v3, v7
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    v_mov_b32_e32 v5, 0
+; CHECK-NEXT:    v_mov_b32_e32 v6, 0
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoint ptr addrspace(8) %ptr to i256
+  ret i256 %ret
+}
+
+;; Check that we extend the offset to i256 instead of reinterpreting all bits.
+;; FIXME: this is wrong, we are removing the trunc to i48:
+define zeroext i256 @ptrtoaddr_ext(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoaddr_ext:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    v_mov_b32_e32 v2, v6
+; CHECK-NEXT:    v_mov_b32_e32 v3, v7
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    v_mov_b32_e32 v5, 0
+; CHECK-NEXT:    v_mov_b32_e32 v6, 0
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoaddr ptr addrspace(8) %ptr to i256
+  ret i256 %ret
+}
+
+define zeroext i64 @ptrtoint_trunc(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoint_trunc:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v1, v5
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoint ptr addrspace(8) %ptr to i64
+  ret i64 %ret
+}
+
+define zeroext i16 @ptrtoaddr_trunc(ptr addrspace(8) %ignored, ptr addrspace(8) %ptr) {
+; CHECK-LABEL: ptrtoaddr_trunc:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_and_b32_e32 v0, 0xffff, v4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ret = ptrtoaddr ptr addrspace(8) %ptr to i16
+  ret i16 %ret
+}
----------------
arsenm wrote:

Should have SGPR variants too 

https://github.com/llvm/llvm-project/pull/143813