mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-22 18:54:02 +01:00
8980610845
For a long time, the InstCombine pass handled target specific intrinsics. Having target specific code in general passes was noted as an area for improvement for a long time. D81728 moves most target specific code out of the InstCombine pass. Applying the target specific combinations in an extra pass would probably result in inferior optimizations compared to the current fixed-point iteration, therefore the InstCombine pass resorts to newly introduced functions in the TargetTransformInfo when it encounters unknown intrinsics. The patch should not have any effect on generated code (under the assumption that code never uses intrinsics from a foreign target). This introduces three new functions: TargetTransformInfo::instCombineIntrinsic TargetTransformInfo::simplifyDemandedUseBitsIntrinsic TargetTransformInfo::simplifyDemandedVectorEltsIntrinsic A few target specific parts are left in the InstCombine folder, where it makes sense to share code. The largest left-over part in InstCombineCalls.cpp is the code shared between arm and aarch64. This allows to move about 3000 lines out from InstCombine to the targets. Differential Revision: https://reviews.llvm.org/D81728
26 lines
1.2 KiB
LLVM
26 lines
1.2 KiB
LLVM
; RUN: opt < %s -instcombine -mtriple=arm -S | FileCheck %s
|
|
|
|
; The alignment arguments for NEON load/store intrinsics can be increased
|
|
; by instcombine. Check for this.
|
|
|
|
; CHECK: vld4.v2i32.p0i8({{.*}}, i32 32)
|
|
; CHECK: vst4.p0i8.v2i32({{.*}}, i32 16)
|
|
|
|
@x = common global [8 x i32] zeroinitializer, align 32
|
|
@y = common global [8 x i32] zeroinitializer, align 16
|
|
|
|
%struct.__neon_int32x2x4_t = type { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }
|
|
|
|
define void @test() nounwind ssp {
|
|
%tmp1 = call %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32.p0i8(i8* bitcast ([8 x i32]* @x to i8*), i32 1)
|
|
%tmp2 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 0
|
|
%tmp3 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 1
|
|
%tmp4 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 2
|
|
%tmp5 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 3
|
|
call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* bitcast ([8 x i32]* @y to i8*), <2 x i32> %tmp2, <2 x i32> %tmp3, <2 x i32> %tmp4, <2 x i32> %tmp5, i32 1)
|
|
ret void
|
|
}
|
|
|
|
declare %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32.p0i8(i8*, i32) nounwind readonly
|
|
declare void @llvm.arm.neon.vst4.p0i8.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32) nounwind
|