shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/clang/test/CodeGen/builtin-align.c
Alex Richardson 8c387cbea7 Add builtins for aligning and checking alignment of pointers and integers 
This change introduces three new builtins (which work on both pointers
and integers) that can be used instead of common bitwise arithmetic:
__builtin_align_up(x, alignment), __builtin_align_down(x, alignment) and
__builtin_is_aligned(x, alignment).

I originally added these builtins to the CHERI fork of LLVM a few years ago
to handle the slightly different C semantics that we use for CHERI [1].
Until recently these builtins (or sequences of other builtins) were
required to generate correct code. I have since made changes to the default
C semantics so that they are no longer strictly necessary (but using them
does generate slightly more efficient code). However, based on our experience
using them in various projects over the past few years, I believe that adding
these builtins to clang would be useful.

These builtins have the following benefit over bit-manipulation and casts
via uintptr_t:

- The named builtins clearly convey the semantics of the operation. While
  checking alignment using __builtin_is_aligned(x, 16) versus
  ((x & 15) == 0) is probably not a huge win in readably, I personally find
  __builtin_align_up(x, N) a lot easier to read than (x+(N-1))&~(N-1).
- They preserve the type of the argument (including const qualifiers). When
  using casts via uintptr_t, it is easy to cast to the wrong type or strip
  qualifiers such as const.
- If the alignment argument is a constant value, clang can check that it is
  a power-of-two and within the range of the type. Since the semantics of
  these builtins is well defined compared to arbitrary bit-manipulation,
  it is possible to add a UBSAN checker that the run-time value is a valid
  power-of-two. I intend to add this as a follow-up to this change.
- The builtins avoids int-to-pointer casts both in C and LLVM IR.
  In the future (i.e. once most optimizations handle it), we could use the new
  llvm.ptrmask intrinsic to avoid the ptrtoint instruction that would normally
  be generated.
- They can be used to round up/down to the next aligned value for both
  integers and pointers without requiring two separate macros.
- In many projects the alignment operations are already wrapped in macros (e.g.
  roundup2 and rounddown2 in FreeBSD), so by replacing the macro implementation
  with a builtin call, we get improved diagnostics for many call-sites while
  only having to change a few lines.
- Finally, the builtins also emit assume_aligned metadata when used on pointers.
  This can improve code generation compared to the uintptr_t casts.

[1] In our CHERI compiler we have compilation mode where all pointers are
implemented as capabilities (essentially unforgeable 128-bit fat pointers).
In our original model, casts from uintptr_t (which is a 128-bit capability)
to an integer value returned the "offset" of the capability (i.e. the
difference between the virtual address and the base of the allocation).
This causes problems for cases such as checking the alignment: for example, the
expression `if ((uintptr_t)ptr & 63) == 0` is generally used to check if the
pointer is aligned to a multiple of 64 bytes. The problem with offsets is that
any pointer to the beginning of an allocation will have an offset of zero, so
this check always succeeds in that case (even if the address is not correctly
aligned). The same issues also exist when aligning up or down. Using the
alignment builtins ensures that the address is used instead of the offset. While
I have since changed the default C semantics to return the address instead of
the offset when casting, this offset compilation mode can still be used by
passing a command-line flag.

Reviewers: rsmith, aaron.ballman, theraven, fhahn, lebedev.ri, nlopes, aqjune
Reviewed By: aaron.ballman, lebedev.ri
Differential Revision: https://reviews.llvm.org/D71499
2020-01-09 21:48:29 +00:00

128 lines
6.7 KiB
C
Raw Blame History

/// Check the code generation for the alignment builtins
/// To make the test case easier to read, run SROA after generating IR to remove the alloca instructions.
// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -DTEST_VOID_PTR \
// RUN: -o - -emit-llvm %s -disable-O0-optnone | opt -S -sroa | \
// RUN: FileCheck %s -check-prefixes CHECK,POINTER,ALIGNMENT_EXT \
// RUN: -enable-var-scope '-D$PTRTYPE=i8'
// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -DTEST_FLOAT_PTR \
// RUN: -o - -emit-llvm %s -disable-O0-optnone | opt -S -sroa | \
// RUN: FileCheck %s -check-prefixes CHECK,POINTER,NON_I8_POINTER,ALIGNMENT_EXT \
// RUN: -enable-var-scope '-D$PTRTYPE=f32'
// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -DTEST_LONG \
// RUN: -o - -emit-llvm %s -disable-O0-optnone | opt -S -sroa | \
// RUN: FileCheck %s -check-prefixes CHECK,INTEGER,ALIGNMENT_EXT -enable-var-scope
/// Check that we can handle the case where the alignment parameter is wider
/// than the source type (generate a trunc on alignment instead of zext)
// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -DTEST_USHORT \
// RUN: -o - -emit-llvm %s -disable-O0-optnone | opt -S -sroa | \
// RUN: FileCheck %s -check-prefixes CHECK,INTEGER,ALIGNMENT_TRUNC -enable-var-scope
#ifdef TEST_VOID_PTR
#define TYPE void *
#elif defined(TEST_FLOAT_PTR)
#define TYPE float *
#elif defined(TEST_LONG)
#define TYPE long
#elif defined(TEST_CAP)
#define TYPE void *__capability
#elif defined(TEST_USHORT)
#define TYPE unsigned short
#else
#error MISSING TYPE
#endif
/// Check that constant initializers work and are correct
_Bool aligned_true = __builtin_is_aligned(1024, 512);
// CHECK: @aligned_true = global i8 1, align 1
_Bool aligned_false = __builtin_is_aligned(123, 512);
// CHECK: @aligned_false = global i8 0, align 1
int down_1 = __builtin_align_down(1023, 32);
// CHECK: @down_1 = global i32 992, align 4
int down_2 = __builtin_align_down(256, 32);
// CHECK: @down_2 = global i32 256, align 4
int up_1 = __builtin_align_up(1023, 32);
// CHECK: @up_1 = global i32 1024, align 4
int up_2 = __builtin_align_up(256, 32);
// CHECK: @up_2 = global i32 256, align 4
/// Capture the IR type here to use in the remaining FileCheck captures:
// CHECK: define {{[^@]+}}@get_type() #0
// CHECK-NEXT: entry:
// POINTER-NEXT: ret [[$TYPE:.+]] null
// INTEGER-NEXT: ret [[$TYPE:.+]] 0
//
TYPE get_type(void) {
return (TYPE)0;
}
// CHECK-LABEL: define {{[^@]+}}@is_aligned
// CHECK-SAME: ([[$TYPE]] {{[^%]*}}[[PTR:%.*]], i32 [[ALIGN:%.*]]) #0
// CHECK-NEXT: entry:
// ALIGNMENT_EXT-NEXT: [[ALIGNMENT:%.*]] = zext i32 [[ALIGN]] to [[ALIGN_TYPE:i64]]
// ALIGNMENT_TRUNC-NEXT: [[ALIGNMENT:%.*]] = trunc i32 [[ALIGN]] to [[ALIGN_TYPE:i16]]
// CHECK-NEXT: [[MASK:%.*]] = sub [[ALIGN_TYPE]] [[ALIGNMENT]], 1
// POINTER-NEXT: [[PTR:%.*]] = ptrtoint [[$TYPE]] %ptr to i64
// CHECK-NEXT: [[SET_BITS:%.*]] = and [[ALIGN_TYPE]] [[PTR]], [[MASK]]
// CHECK-NEXT: [[IS_ALIGNED:%.*]] = icmp eq [[ALIGN_TYPE]] [[SET_BITS]], 0
// CHECK-NEXT: ret i1 [[IS_ALIGNED]]
//
_Bool is_aligned(TYPE ptr, unsigned align) {
return __builtin_is_aligned(ptr, align);
}
// CHECK-LABEL: define {{[^@]+}}@align_up
// CHECK-SAME: ([[$TYPE]] {{[^%]*}}[[PTR:%.*]], i32 [[ALIGN:%.*]]) #0
// CHECK-NEXT: entry:
// ALIGNMENT_EXT-NEXT: [[ALIGNMENT:%.*]] = zext i32 [[ALIGN]] to [[ALIGN_TYPE:i64]]
// ALIGNMENT_TRUNC-NEXT: [[ALIGNMENT:%.*]] = trunc i32 [[ALIGN]] to [[ALIGN_TYPE:i16]]
// CHECK-NEXT: [[MASK:%.*]] = sub [[ALIGN_TYPE]] [[ALIGNMENT]], 1
// INTEGER-NEXT: [[OVER_BOUNDARY:%.*]] = add [[$TYPE]] [[PTR]], [[MASK]]
// NOTYET-POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = call [[$TYPE]] @llvm.ptrmask.p0[[$PTRTYPE]].p0i8.i64(i8* [[OVER_BOUNDARY]], [[ALIGN_TYPE]] [[INVERTED_MASK]])
// POINTER-NEXT: [[INTPTR:%.*]] = ptrtoint [[$TYPE]] [[PTR]] to [[ALIGN_TYPE]]
// POINTER-NEXT: [[OVER_BOUNDARY:%.*]] = add [[ALIGN_TYPE]] [[INTPTR]], [[MASK]]
// CHECK-NEXT: [[INVERTED_MASK:%.*]] = xor [[ALIGN_TYPE]] [[MASK]], -1
// CHECK-NEXT: [[ALIGNED_RESULT:%.*]] = and [[ALIGN_TYPE]] [[OVER_BOUNDARY]], [[INVERTED_MASK]]
// POINTER-NEXT: [[DIFF:%.*]] = sub i64 [[ALIGNED_RESULT]], [[INTPTR]]
// NON_I8_POINTER-NEXT: [[PTR:%.*]] = bitcast [[$TYPE]] {{%.*}} to i8*
// POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i64 [[DIFF]]
// NON_I8_POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = bitcast i8* {{%.*}} to [[$TYPE]]
// POINTER-NEXT: [[ASSUME_MASK:%.*]] = sub i64 %alignment, 1
// POINTER-NEXT: [[ASSUME_INTPTR:%.*]]= ptrtoint [[$TYPE]] [[ALIGNED_RESULT]] to i64
// POINTER-NEXT: [[MASKEDPTR:%.*]] = and i64 %ptrint, [[ASSUME_MASK]]
// POINTER-NEXT: [[MASKEDCOND:%.*]] = icmp eq i64 [[MASKEDPTR]], 0
// POINTER-NEXT: call void @llvm.assume(i1 [[MASKEDCOND]])
// CHECK-NEXT: ret [[$TYPE]] [[ALIGNED_RESULT]]
//
TYPE align_up(TYPE ptr, unsigned align) {
return __builtin_align_up(ptr, align);
}
// CHECK-LABEL: define {{[^@]+}}@align_down
// CHECK-SAME: ([[$TYPE]] {{[^%]*}}[[PTR:%.*]], i32 [[ALIGN:%.*]]) #0
// CHECK-NEXT: entry:
// ALIGNMENT_EXT-NEXT: [[ALIGNMENT:%.*]] = zext i32 [[ALIGN]] to [[ALIGN_TYPE:i64]]
// ALIGNMENT_TRUNC-NEXT: [[ALIGNMENT:%.*]] = trunc i32 [[ALIGN]] to [[ALIGN_TYPE:i16]]
// CHECK-NEXT: [[MASK:%.*]] = sub [[ALIGN_TYPE]] [[ALIGNMENT]], 1
// NOTYET-POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = call [[$TYPE]] @llvm.ptrmask.p0[[$PTRTYPE]].p0[[$PTRTYPE]].i64([[$TYPE]] [[PTR]], [[ALIGN_TYPE]] [[INVERTED_MASK]])
// POINTER-NEXT: [[INTPTR:%.*]] = ptrtoint [[$TYPE]] [[PTR]] to [[ALIGN_TYPE]]
// CHECK-NEXT: [[INVERTED_MASK:%.*]] = xor [[ALIGN_TYPE]] [[MASK]], -1
// POINTER-NEXT: [[ALIGNED_INTPTR:%.*]] = and [[ALIGN_TYPE]] [[INTPTR]], [[INVERTED_MASK]]
// POINTER-NEXT: [[DIFF:%.*]] = sub i64 [[ALIGNED_INTPTR]], [[INTPTR]]
// NON_I8_POINTER-NEXT: [[PTR:%.*]] = bitcast [[$TYPE]] {{%.*}} to i8*
// POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i64 [[DIFF]]
// NON_I8_POINTER-NEXT: [[ALIGNED_RESULT:%.*]] = bitcast i8* {{%.*}} to [[$TYPE]]
// INTEGER-NEXT: [[ALIGNED_RESULT:%.*]] = and [[ALIGN_TYPE]] [[PTR]], [[INVERTED_MASK]]
// POINTER-NEXT: [[ASSUME_MASK:%.*]] = sub i64 %alignment, 1
// POINTER-NEXT: [[ASSUME_INTPTR:%.*]]= ptrtoint [[$TYPE]] [[ALIGNED_RESULT]] to i64
// POINTER-NEXT: [[MASKEDPTR:%.*]] = and i64 %ptrint, [[ASSUME_MASK]]
// POINTER-NEXT: [[MASKEDCOND:%.*]] = icmp eq i64 [[MASKEDPTR]], 0
// POINTER-NEXT: call void @llvm.assume(i1 [[MASKEDCOND]])
// CHECK-NEXT: ret [[$TYPE]] [[ALIGNED_RESULT]]
//
TYPE align_down(TYPE ptr, unsigned align) {
return __builtin_align_down(ptr, align);
}
Reference in New Issue View Git Blame Copy Permalink