Files
clang-p2996/llvm/test/CodeGen/SPIRV/instructions/atomic-ptr.ll
Vyacheslav Levytskyy 67d3ef74b3 [SPIR-V] Rework usage of virtual registers' types and classes (#104104)
This PR continues https://github.com/llvm/llvm-project/pull/101732
changes in virtual register processing aimed to improve correctness of
emitted MIR between passes from the perspective of MachineVerifier.
Namely, the following changes are introduced:
* register classes (lib/Target/SPIRV/SPIRVRegisterInfo.td) and
instruction patterns (lib/Target/SPIRV/SPIRVInstrInfo.td) are corrected
and simplified (by removing unnecessary sophisticated options) -- e.g.,
this PR gets rid of duplicating 32/64 bits patterns, removes ANYID
register class and simplifies definition of the rest of register
classes,
* hardcoded LLT scalar types in passes before instruction selection are
corrected -- the goal is to have correct bit width before instruction
selection, and use 64 bits registers for pattern matching in the
instruction selection pass; 32-bit registers remain where they are
described in such terms by SPIR-V specification (like, for example,
creation of virtual registers for scope/mem semantics operands),
* rework virtual register type/class assignment for calls/builtins
lowering,
* a series of minor changes to fix validity of emitted code between
passes:
  - ensure that that bitcast changes the type,
  - fix the pattern for instruction selection for OpExtInst,
  - simplify inline asm operands usage,
  - account for arbitrary integer sizes / update legalizer rules;
* add '-verify-machineinstrs' to existed test cases.

See also https://github.com/llvm/llvm-project/issues/88129 that this PR
may resolve.

This PR fixes a great number of issues reported by MachineVerifier and,
as a result, reduces a number of failed test cases for the mode with
expensive checks set on from ~200 to ~57.
2024-08-22 09:40:27 +02:00

39 lines
1.7 KiB
LLVM

; The goal of the test case is to ensure that the Backend doesn't crash on the stage
; of type inference. Result SPIR-V is not expected to be valid from the perspective
; of spirv-val in this case, because there is a difference of accepted return types
; between atomicrmw and OpAtomicExchange.
; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s
; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
; CHECK-DAG: %[[#LongTy:]] = OpTypeInt 64 0
; CHECK-DAG: %[[#PtrLongTy:]] = OpTypePointer CrossWorkgroup %[[#LongTy]]
; CHECK-DAG: %[[#IntTy:]] = OpTypeInt 32 0
; CHECK-DAG: %[[#Scope:]] = OpConstant %[[#IntTy]] 1
; CHECK-DAG: %[[#MemSem:]] = OpConstant %[[#IntTy]] 8
; CHECK-DAG: %[[#PtrPtrLongTy:]] = OpTypePointer CrossWorkgroup %[[#PtrLongTy]]
; CHECK: OpFunction
; CHECK: %[[#Arg1:]] = OpFunctionParameter %[[#PtrPtrLongTy]]
; CHECK: %[[#Arg2:]] = OpFunctionParameter %[[#PtrLongTy]]
; CHECK: OpAtomicExchange %[[#PtrLongTy]] %[[#Arg1]] %[[#Scope]] %[[#MemSem]] %[[#Arg2]]
; CHECK: OpFunctionEnd
define dso_local spir_func void @test1(ptr addrspace(1) %arg1, ptr addrspace(1) byval(i64) %arg_ptr) {
entry:
%r = atomicrmw xchg ptr addrspace(1) %arg1, ptr addrspace(1) %arg_ptr acq_rel
ret void
}
; CHECK: OpFunction
; CHECK: %[[#Arg3:]] = OpFunctionParameter %[[#PtrLongTy]]
; CHECK: %[[#Arg4:]] = OpFunctionParameter %[[#LongTy]]
; CHECK: OpAtomicExchange %[[#LongTy]] %[[#Arg3]] %[[#Scope]] %[[#MemSem]] %[[#Arg4]]
; CHECK: OpFunctionEnd
define dso_local spir_func void @test2(ptr addrspace(1) %arg1, i64 %arg_ptr) {
entry:
%r = atomicrmw xchg ptr addrspace(1) %arg1, i64 %arg_ptr acq_rel
ret void
}