llvm-6502/test/Transforms/InstCombine/bitcast-alias-function.ll

240 lines
8.6 KiB
LLVM
Raw Normal View History

; RUN: opt -S -instcombine -o - %s | FileCheck %s
target datalayout = "e-p:32:32:32-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v64:64:64-v128:128:128-a0:0:64"
; Cases that should be bitcast
; Test cast between scalars with same bit sizes
@alias_i32_to_f32 = alias bitcast (i32 (i32)* @func_i32 to float (float)*)
; Test cast between vectors with same number of elements and bit sizes
@alias_v2i32_to_v2f32 = alias bitcast (<2 x i32> (<2 x i32>)* @func_v2i32 to <2 x float> (<2 x float>)*)
; Test cast from vector to scalar with same number of bits
@alias_v2f32_to_i64 = alias bitcast (i64 (i64)* @func_i64 to <2 x float> (<2 x float>)*)
; Test cast from scalar to vector with same number of bits
@alias_i64_to_v2f32 = alias bitcast (<2 x float> (<2 x float>)* @func_v2f32 to i64 (i64)*)
; Test cast between vectors of pointers
@alias_v2i32p_to_v2i64p = alias bitcast (<2 x i32*> (<2 x i32*>)* @func_v2i32p to <2 x i64*> (<2 x i64*>)*)
; Cases that should be invalid and unchanged
; Test cast between scalars with different bit sizes
@alias_i64_to_f32 = alias bitcast (i64 (i64)* @func_i64 to float (float)*)
; Test cast between vectors with different bit sizes but the
; same number of elements
@alias_v2i64_to_v2f32 = alias bitcast (<2 x i64> (<2 x i64>)* @func_v2i64 to <2 x float> (<2 x float>)*)
; Test cast between vectors with same number of bits and different
; numbers of elements
@alias_v2i32_to_v4f32 = alias bitcast (<2 x i32> (<2 x i32>)* @func_v2i32 to <4 x float> (<4 x float>)*)
; Test cast between scalar and vector with different number of bits
@alias_i64_to_v4f32 = alias bitcast (<4 x float> (<4 x float>)* @func_v4f32 to i64 (i64)*)
; Test cast between vector and scalar with different number of bits
@alias_v4f32_to_i64 = alias bitcast (i64 (i64)* @func_i64 to <4 x float> (<4 x float>)*)
; Test cast from scalar to vector of pointers with same number of bits
; We don't know the pointer size at this point, so this can't be done
@alias_i64_to_v2i32p = alias bitcast (<2 x i32*> (<2 x i32*>)* @func_v2i32p to i64 (i64)*)
; Test cast between vector of pointers and scalar with different number of bits
@alias_v4i32p_to_i64 = alias bitcast (i64 (i64)* @func_i64 to <4 x i32*> (<4 x i32*>)*)
define internal <2 x i32> @func_v2i32(<2 x i32> %v) noinline nounwind {
entry:
ret <2 x i32> %v
}
define internal <2 x float> @func_v2f32(<2 x float> %v) noinline nounwind {
entry:
ret <2 x float> %v
}
define internal <4 x float> @func_v4f32(<4 x float> %v) noinline nounwind {
entry:
ret <4 x float> %v
}
define internal i32 @func_i32(i32 %v) noinline nounwind {
entry:
ret i32 %v
}
define internal i64 @func_i64(i64 %v) noinline nounwind {
entry:
ret i64 %v
}
define internal <2 x i64> @func_v2i64(<2 x i64> %v) noinline nounwind {
entry:
ret <2 x i64> %v
}
define internal <2 x i32*> @func_v2i32p(<2 x i32*> %v) noinline nounwind {
entry:
ret <2 x i32*> %v
}
; Valid cases, only bitcast for argument / return type and call underlying function
; Sizes match, should only bitcast
define void @bitcast_alias_scalar(float* noalias %source, float* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_scalar
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load ...)) and (load (cast ...)): canonicalize toward the former. Historically, we've tried to load using the type of the *pointer*, and tried to match that type as closely as possible removing as many pointer casts as we could and trading them for bitcasts of the loaded value. This is deeply and fundamentally wrong. Repeat after me: memory does not have a type! This was a hard lesson for me to learn working on SROA. There is only one thing that should actually drive the type used for a pointer, and that is the type which we need to use to load from that pointer. Matching up pointer types to the loaded value types is very useful because it minimizes the physical size of the IR required for no-op casts. Similarly, the only thing that should drive the type used for a loaded value is *how that value is used*! Again, this minimizes casts. And in fact, the *only* thing motivating types in any part of LLVM's IR are the types used by the operations in the IR. We should match them as closely as possible. I've ended up removing some tests here as they were testing bugs or behavior that is no longer present. Mostly though, this is just cleanup to let the tests continue to function as intended. The only fallout I've found so far from this change was SROA and I have fixed it to not be impeded by the different type of load. If you find more places where this change causes optimizations not to fire, those too are likely bugs where we are assuming that the type of pointers is "significant" for optimization purposes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-18 06:36:22 +00:00
; CHECK: bitcast float* %source to i32*
; CHECK: load i32, i32*
; CHECK-NOT: fptoui
; CHECK-NOT: uitofp
; CHECK: bitcast float* %dest to i32*
; CHECK: store i32
%tmp = load float, float* %source, align 8
%call = call float @alias_i32_to_f32(float %tmp) nounwind
store float %call, float* %dest, align 8
ret void
}
; Sizes match, should only bitcast
define void @bitcast_alias_vector(<2 x float>* noalias %source, <2 x float>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load ...)) and (load (cast ...)): canonicalize toward the former. Historically, we've tried to load using the type of the *pointer*, and tried to match that type as closely as possible removing as many pointer casts as we could and trading them for bitcasts of the loaded value. This is deeply and fundamentally wrong. Repeat after me: memory does not have a type! This was a hard lesson for me to learn working on SROA. There is only one thing that should actually drive the type used for a pointer, and that is the type which we need to use to load from that pointer. Matching up pointer types to the loaded value types is very useful because it minimizes the physical size of the IR required for no-op casts. Similarly, the only thing that should drive the type used for a loaded value is *how that value is used*! Again, this minimizes casts. And in fact, the *only* thing motivating types in any part of LLVM's IR are the types used by the operations in the IR. We should match them as closely as possible. I've ended up removing some tests here as they were testing bugs or behavior that is no longer present. Mostly though, this is just cleanup to let the tests continue to function as intended. The only fallout I've found so far from this change was SROA and I have fixed it to not be impeded by the different type of load. If you find more places where this change causes optimizations not to fire, those too are likely bugs where we are assuming that the type of pointers is "significant" for optimization purposes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-18 06:36:22 +00:00
; CHECK: bitcast <2 x float>* %source to <2 x i32>*
; CHECK: load <2 x i32>, <2 x i32>*
; CHECK-NOT: fptoui
; CHECK-NOT: uitofp
; CHECK: bitcast <2 x float>* %dest to <2 x i32>*
; CHECK: store <2 x i32>
%tmp = load <2 x float>, <2 x float>* %source, align 8
%call = call <2 x float> @alias_v2i32_to_v2f32(<2 x float> %tmp) nounwind
store <2 x float> %call, <2 x float>* %dest, align 8
ret void
}
; Sizes match, should only bitcast
define void @bitcast_alias_vector_scalar_same_size(<2 x float>* noalias %source, <2 x float>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector_scalar_same_size
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load ...)) and (load (cast ...)): canonicalize toward the former. Historically, we've tried to load using the type of the *pointer*, and tried to match that type as closely as possible removing as many pointer casts as we could and trading them for bitcasts of the loaded value. This is deeply and fundamentally wrong. Repeat after me: memory does not have a type! This was a hard lesson for me to learn working on SROA. There is only one thing that should actually drive the type used for a pointer, and that is the type which we need to use to load from that pointer. Matching up pointer types to the loaded value types is very useful because it minimizes the physical size of the IR required for no-op casts. Similarly, the only thing that should drive the type used for a loaded value is *how that value is used*! Again, this minimizes casts. And in fact, the *only* thing motivating types in any part of LLVM's IR are the types used by the operations in the IR. We should match them as closely as possible. I've ended up removing some tests here as they were testing bugs or behavior that is no longer present. Mostly though, this is just cleanup to let the tests continue to function as intended. The only fallout I've found so far from this change was SROA and I have fixed it to not be impeded by the different type of load. If you find more places where this change causes optimizations not to fire, those too are likely bugs where we are assuming that the type of pointers is "significant" for optimization purposes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-18 06:36:22 +00:00
; CHECK: bitcast <2 x float>* %source to i64*
; CHECK: load i64, i64*
; CHECK: %call = call i64 @func_i64
; CHECK: bitcast <2 x float>* %dest to i64*
; CHECK: store i64
%tmp = load <2 x float>, <2 x float>* %source, align 8
%call = call <2 x float> @alias_v2f32_to_i64(<2 x float> %tmp) nounwind
store <2 x float> %call, <2 x float>* %dest, align 8
ret void
}
define void @bitcast_alias_scalar_vector_same_size(i64* noalias %source, i64* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_scalar_vector_same_size
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load ...)) and (load (cast ...)): canonicalize toward the former. Historically, we've tried to load using the type of the *pointer*, and tried to match that type as closely as possible removing as many pointer casts as we could and trading them for bitcasts of the loaded value. This is deeply and fundamentally wrong. Repeat after me: memory does not have a type! This was a hard lesson for me to learn working on SROA. There is only one thing that should actually drive the type used for a pointer, and that is the type which we need to use to load from that pointer. Matching up pointer types to the loaded value types is very useful because it minimizes the physical size of the IR required for no-op casts. Similarly, the only thing that should drive the type used for a loaded value is *how that value is used*! Again, this minimizes casts. And in fact, the *only* thing motivating types in any part of LLVM's IR are the types used by the operations in the IR. We should match them as closely as possible. I've ended up removing some tests here as they were testing bugs or behavior that is no longer present. Mostly though, this is just cleanup to let the tests continue to function as intended. The only fallout I've found so far from this change was SROA and I have fixed it to not be impeded by the different type of load. If you find more places where this change causes optimizations not to fire, those too are likely bugs where we are assuming that the type of pointers is "significant" for optimization purposes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-18 06:36:22 +00:00
; CHECK: bitcast i64* %source to <2 x float>*
; CHECK: load <2 x float>, <2 x float>*
; CHECK: call <2 x float> @func_v2f32
; CHECK: bitcast i64* %dest to <2 x float>*
; CHECK: store <2 x float>
%tmp = load i64, i64* %source, align 8
%call = call i64 @alias_i64_to_v2f32(i64 %tmp) nounwind
store i64 %call, i64* %dest, align 8
ret void
}
define void @bitcast_alias_vector_ptrs_same_size(<2 x i64*>* noalias %source, <2 x i64*>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector_ptrs_same_size
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load ...)) and (load (cast ...)): canonicalize toward the former. Historically, we've tried to load using the type of the *pointer*, and tried to match that type as closely as possible removing as many pointer casts as we could and trading them for bitcasts of the loaded value. This is deeply and fundamentally wrong. Repeat after me: memory does not have a type! This was a hard lesson for me to learn working on SROA. There is only one thing that should actually drive the type used for a pointer, and that is the type which we need to use to load from that pointer. Matching up pointer types to the loaded value types is very useful because it minimizes the physical size of the IR required for no-op casts. Similarly, the only thing that should drive the type used for a loaded value is *how that value is used*! Again, this minimizes casts. And in fact, the *only* thing motivating types in any part of LLVM's IR are the types used by the operations in the IR. We should match them as closely as possible. I've ended up removing some tests here as they were testing bugs or behavior that is no longer present. Mostly though, this is just cleanup to let the tests continue to function as intended. The only fallout I've found so far from this change was SROA and I have fixed it to not be impeded by the different type of load. If you find more places where this change causes optimizations not to fire, those too are likely bugs where we are assuming that the type of pointers is "significant" for optimization purposes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-18 06:36:22 +00:00
; CHECK: bitcast <2 x i64*>* %source to <2 x i32*>*
; CHECK: load <2 x i32*>, <2 x i32*>*
; CHECK: call <2 x i32*> @func_v2i32p
; CHECK: bitcast <2 x i64*>* %dest to <2 x i32*>*
; CHECK: store <2 x i32*>
%tmp = load <2 x i64*>, <2 x i64*>* %source, align 8
%call = call <2 x i64*> @alias_v2i32p_to_v2i64p(<2 x i64*> %tmp) nounwind
store <2 x i64*> %call, <2 x i64*>* %dest, align 8
ret void
}
; Invalid cases:
define void @bitcast_alias_mismatch_scalar_size(float* noalias %source, float* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_mismatch_scalar_size
; CHECK-NOT: fptoui
; CHECK: @alias_i64_to_f32
; CHECK-NOT: uitofp
%tmp = load float, float* %source, align 8
%call = call float @alias_i64_to_f32(float %tmp) nounwind
store float %call, float* %dest, align 8
ret void
}
define void @bitcast_alias_mismatch_vector_element_and_bit_size(<2 x float>* noalias %source, <2 x float>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_mismatch_vector_element_and_bit_size
; CHECK-NOT: fptoui <2 x float> %tmp to <2 x i64>
; CHECK: @alias_v2i64_to_v2f32
; CHECK-NOT: uitofp <2 x i64> %call to <2 x float>
%tmp = load <2 x float>, <2 x float>* %source, align 8
%call = call <2 x float> @alias_v2i64_to_v2f32(<2 x float> %tmp) nounwind
store <2 x float> %call, <2 x float>* %dest, align 8
ret void
}
define void @bitcast_alias_vector_mismatched_number_elements(<4 x float>* noalias %source, <4 x float>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector_mismatched_number_elements
; CHECK: %call = call <4 x float> @alias_v2i32_to_v4f32
%tmp = load <4 x float>, <4 x float>* %source, align 8
%call = call <4 x float> @alias_v2i32_to_v4f32(<4 x float> %tmp) nounwind
store <4 x float> %call, <4 x float>* %dest, align 8
ret void
}
define void @bitcast_alias_vector_scalar_mismatched_bit_size(<4 x float>* noalias %source, <4 x float>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector_scalar_mismatched_bit_size
; CHECK: %call = call <4 x float> @alias_v4f32_to_i64
%tmp = load <4 x float>, <4 x float>* %source, align 8
%call = call <4 x float> @alias_v4f32_to_i64(<4 x float> %tmp) nounwind
store <4 x float> %call, <4 x float>* %dest, align 8
ret void
}
define void @bitcast_alias_vector_ptrs_scalar_mismatched_bit_size(<4 x i32*>* noalias %source, <4 x i32*>* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_vector_ptrs_scalar_mismatched_bit_size
; CHECK: @alias_v4i32p_to_i64
%tmp = load <4 x i32*>, <4 x i32*>* %source, align 8
%call = call <4 x i32*> @alias_v4i32p_to_i64(<4 x i32*> %tmp) nounwind
store <4 x i32*> %call, <4 x i32*>* %dest, align 8
ret void
}
define void @bitcast_alias_scalar_vector_ptrs_same_size(i64* noalias %source, i64* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_scalar_vector_ptrs_same_size
; CHECK: @alias_i64_to_v2i32p
%tmp = load i64, i64* %source, align 8
%call = call i64 @alias_i64_to_v2i32p(i64 %tmp) nounwind
store i64 %call, i64* %dest, align 8
ret void
}
define void @bitcast_alias_scalar_vector_mismatched_bit_size(i64* noalias %source, i64* noalias %dest) nounwind {
entry:
; CHECK-LABEL: @bitcast_alias_scalar_vector_mismatched_bit_size
; CHECK: call i64 @alias_i64_to_v4f32
%tmp = load i64, i64* %source, align 8
%call = call i64 @alias_i64_to_v4f32(i64 %tmp) nounwind
store i64 %call, i64* %dest, align 8
ret void
}