I am following this toy example to lower mlir to llvm ir and compile the result. Following the instructions, I can compile the project. However, the generated llvm ir (print.ll file) is different from what is in the github. For instance, I have ptr instead of i8* in the print.ll (example: declare ptr @malloc(i64) ). As a result lli cannot compile and gives error. I am using Ubuntu LLVM version 14.0.0.
Error:
lli: lli: ./print2.ll:9:9: error: expected type declare ptr @malloc(i64)
File print.ll:
; ModuleID = 'LLVMDialectModule'
source_filename = "LLVMDialectModule"
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
@nl = internal constant [2 x i8] c"\0A\00"
@frmt_spec = internal constant [4 x i8] c"%f \00"
**declare ptr @malloc(i64)**
declare void @free(ptr)
declare i32 @printf(ptr, ...)
define void @main() {
%1 = call ptr @malloc(i64 ptrtoint (ptr getelementptr (double, ptr null, i64 6) to i64))
%2 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } undef, ptr %1, 0
%3 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %2, ptr %1, 1
%4 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %3, i64 0, 2
%5 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %4, i64 2, 3, 0
%6 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %5, i64 3, 3, 1
%7 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %6, i64 3, 4, 0
%8 = insertvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %7, i64 1, 4, 1
%9 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%10 = getelementptr double, ptr %9, i64 0
store double 1.000000e+00, ptr %10, align 8
%11 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%12 = getelementptr double, ptr %11, i64 1
store double 2.000000e+00, ptr %12, align 8
%13 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%14 = getelementptr double, ptr %13, i64 2
store double 3.000000e+00, ptr %14, align 8
%15 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%16 = getelementptr double, ptr %15, i64 3
store double 4.000000e+00, ptr %16, align 8
%17 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%18 = getelementptr double, ptr %17, i64 4
store double 5.000000e+00, ptr %18, align 8
%19 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%20 = getelementptr double, ptr %19, i64 5
store double 6.000000e+00, ptr %20, align 8
br label %21
21: ; preds = %36, %0
%22 = phi i64 [ 0, %0 ], [ %38, %36 ]
%23 = icmp slt i64 %22, 2
br i1 %23, label %24, label %39
24: ; preds = %21
br label %25
25: ; preds = %28, %24
%26 = phi i64 [ 0, %24 ], [ %35, %28 ]
%27 = icmp slt i64 %26, 3
br i1 %27, label %28, label %36
28: ; preds = %25
%29 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 1
%30 = mul i64 %22, 3
%31 = add i64 %30, %26
%32 = getelementptr double, ptr %29, i64 %31
%33 = load double, ptr %32, align 8
%34 = call i32 (ptr, ...) @printf(ptr @frmt_spec, double %33)
%35 = add i64 %26, 1
br label %25
36: ; preds = %25
%37 = call i32 (ptr, ...) @printf(ptr @nl)
%38 = add i64 %22, 1
br label %21
39: ; preds = %21
%40 = extractvalue { ptr, ptr, i64, [2 x i64], [2 x i64] } %8, 0
call void @free(ptr %40)
ret void
}
!llvm.module.flags = !{!0}
!0 = !{i32 2, !"Debug Info Version", i32 3}
You need to use matching LLVM versions: one used by MLIR and another one locally. In your particular case, you'd need to upgrade your 14.0.0 to something newer – the one that does support opaque pointers.
You can install the latest LLVM using instructions from https://apt.llvm.org/