Information about the number of indices in memory accesses

Project Infrastructure LLVM Dev List Archives
#1

Hi all,

For loads and stores i want to extract information about the number of indices accessed. For instance:

struct S {int X, int *Y};

__global__ void kernel(int *A, int **B, struct S) {
int x = A[..][..]; // -> L: A[..][..]
int y = *B[2]; // -> L: B[0][2]
int z = S.y[..]; // -> L: S.1[..]

// etc..
}

I am performing some preprocessing on IR to:
1. Move constant inline GEPs into instructions
2. For loads and stores without a GEP operand, explicitly create a (trivial) GEP with index 0

So now the operand of every load and store is a GEP instruction.

For simple stuff i am getting the right answer but when the index expression becomes more complex multiple GEPs are introduced. For instance:

*(A+2*(blockDim.x*blockIdx.x+threadIdx.x+1)+2+3) = 5;

produces:

%6 = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
%7 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x()
%8 = mul i32 %6, %7,
%9 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
%10 = add i32 %8, %9,
%11 = add i32 %10, 1,
%12 = mul i32 2, %11,
%13 = zext i32 %12 to i64,
%14 = getelementptr inbounds i32, i32* %0, i64 %13
%15 = getelementptr inbounds i32, i32* %14, i64 2
%16 = getelementptr inbounds i32, i32* %15, i64 3
store i32 5, i32* %16, align 4,

So i guess relying on the number of GEPs to figure the number of indices is only a heuristic. Is there a more robust way to go on about it? Or some example i can look at?

PS: I'm only interested about CUDA kernels.

Ees

#2

Anyone? I'd really appreciate any hints to look up as i'm somewhat stuck with this.

Cheers.

#3

Hi Ees,

SCEV Delinearization is the closest I know. But it has its problems. Well for one your expression should be SCEVable.

But more importantly, SCEV Delinearization is trying to deduce something that is high-level (actually source-level) from a low-level IR in which a lot of this info has been lost. So, since there’s not a 1-1 mapping from high-level code to LLVM IR, going backwards will always be imperfect.

And so since you are too trying to deduce a high-level thing, I believe that any solution will be imperfect.

Best,
Stefanos

#4

As Stefanos mentioned, LLVM-IR is generally too low-level for this. C/C++ semantics also don’t make it easier due to possible aliasing.

The 3 cases are very different to analyze.

  1. A[…][…] is syntactically invalid. A is a single pointer.
  2. *B[2] Is not equivalent to B[0][2], but to B[2][0]. This jagged/indirect/pointer-chasing arrays accesses are expressed as chains of getelementptr and load instructions. However, they cannot be usefully optimized because pointers on each level can actually point to the same thing. Eg:

int data[] = { 41, 42, 43 };
B[0] = data;
B[1] = data;
B[2] = data;

therefore

B[0][1] == B[1][1] == B[2][1] == 42;

  1. S.y[…] can be one or multiple getelementptr instructions and is the most analyzable since all indices are static at compile-time. The number of subscripts and their values can directly be taken from the getelementptr instruction(s). The pointers-to-int can still alias with other ints in the program.

  2. These case of VLA or manually linearized array:

int C = alloca(sizeof(int) * m * n);
C[x + my];

This is intended to have two dimensions, but appear as just one in the LLVM-IR. The following (C99 VLA) is compiled to approximately the same LLVM-IR

int C[n][m];
C[y][x];

Delinearization as mentioned by Stefanos tries to recover the two subscripts x and y, but can only do so heuristically. Also keep in mind that
C[1][-1] appears as the same IR as C[0][m-1], so there is no unique way to delinerarize. In particular, one cannot just assume that if all indices are different, that the memory locations being accessed are different (again, a pointer aliasing problem)

Michael

#5

Michael makes a great point about aliasing here and different indexing that accesses the same element!

Another note: x = A[0][2] is fundamentally different depending on the type of A. If e.g. A was declared: int A[10][20], there’s only one load. A is a (and is treated as) a linear buffer,
and GEPs only pinpoint the specific position of A[0][2] in this buffer (i.e. 0*10 + 2). But if A was e.g. this: int **A, there two loads. One load to get the “pointer of the zeroth row” and another load to get the 2nd element off of that row.

So, you see, all these things make any deduction method very very imprecise.

Best,
Stefanos

Στις Σάβ, 3 Οκτ 2020 στις 5:13 π.μ., ο/η Michael Kruse <llvmdev@meinersbur.de> έγραψε:

#6

Stefanos, Michael, thank you for your responses. I’ve been toying with this for some days now and i kind of figured/feared this might not be possible, i just thought, that maybe there is something. I’m new to LLVM so i’m not aware of most things. In any case i’ve decided to take another direction with this.

May i ask one more thing? Say i want for each load/store to get a name for the base pointer. For instance the “A”, “B”, “S.1” in the example above.

In this example:

%14 = getelementptr inbounds i32, i32* %0, i64 %13
%15 = getelementptr inbounds i32, i32* %14, i64 2
%16 = getelementptr inbounds i32, i32* %15, i64 3
store i32 5, i32* %16, align 4,

I can follow the GEP chain until i hit an argument/global/alloca. But what if there are temporaries in between them? Is there a more generic/systematic way to go on about it?
Maybe some project you are aware of that does something similar to have a look at?

Cheers,
Ees

#7

Hi Ees,

You may want to have a look at GetUnderlyingObjects (https://llvm.org/doxygen/namespacellvm.html#ad3428471506e6c03e9395a697a897a83).
The page takes some time to load. Alternatively, you can see the implementation at ValueTracking.cpp (https://llvm.org/doxygen/ValueTracking_8cpp_source.html)
and use it by including llvm/Analysis/ValueTracking.h.

Note that you probably don’t want to call this frequently.

Best,
Stefanos

Στις Σάβ, 3 Οκτ 2020 στις 3:26 μ.μ., ο/η Ees <kayesg42@gmail.com> έγραψε: