llvm-project/clang/lib/CodeGen
Simon Pilgrim e4074432d5 [X86] Remove avx512f integer and/or/xor/min/max reduction intrinsics and use generic equivalents
None of these have any reordering issues, and they still emit the same reduction intrinsics without any change in the existing test coverage:

llvm-project\clang\test\CodeGen\X86\avx512-reduceIntrin.c
llvm-project\clang\test\CodeGen\X86\avx512-reduceMinMaxIntrin.c

Differential Revision: https://reviews.llvm.org/D117881
2022-01-24 11:57:53 +00:00
..
ABIInfo.h Introduce _BitInt, deprecate _ExtInt 2021-12-06 12:52:01 -05:00
Address.h [CodeGen] Avoid some pointer element type accesses 2021-12-21 14:16:28 +01:00
BackendUtil.cpp [time-trace] Add optimizer and codegen regions to NPM 2022-01-21 19:17:57 -08:00
CGAtomic.cpp Use a sorted array instead of a map to store AttrBuilder string attributes 2022-01-10 14:49:53 +01:00
CGBlocks.cpp [CodeGen] Treat ObjC `__unsafe_unretained` and class types as trivial 2022-01-11 11:18:24 -08:00
CGBlocks.h [CodeGen] Treat ObjC `__unsafe_unretained` and class types as trivial 2022-01-11 11:18:24 -08:00
CGBuilder.h [CGBuilder] Add CreateGEP() overload that accepts an Address 2021-12-23 14:53:42 +01:00
CGBuiltin.cpp [X86] Remove avx512f integer and/or/xor/min/max reduction intrinsics and use generic equivalents 2022-01-24 11:57:53 +00:00
CGCUDANV.cpp [HIP] Fix -fgpu-rdc for Windows 2021-12-06 16:42:23 -05:00
CGCUDARuntime.cpp
CGCUDARuntime.h [HIP] Emit kernel symbol 2021-03-01 16:31:40 -05:00
CGCXX.cpp [OpaquePtr] Remove uses of CGF.Builder.CreateConstInBoundsGEP1_64() without type 2021-07-17 17:07:46 +02:00
CGCXXABI.cpp [MS] Implement on-demand TLS initialization for Microsoft CXX ABI 2022-01-13 21:23:23 -08:00
CGCXXABI.h [MS] Implement on-demand TLS initialization for Microsoft CXX ABI 2022-01-13 21:23:23 -08:00
CGCall.cpp [IRGen] Do not overwrite existing attributes in CGCall. 2022-01-20 13:45:19 +00:00
CGCall.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CGClass.cpp [CodeGen] Avoid more pointer element type accesses 2021-12-27 12:00:22 +01:00
CGCleanup.cpp Revert "Skip exception cleanups when the innermost scope is EHTerminateScope." 2022-01-14 18:59:24 -05:00
CGCleanup.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CGCoroutine.cpp [Coroutines] Set presplit attribute in Clang and mlir 2022-01-05 10:25:02 +08:00
CGDebugInfo.cpp [ADT] Add an in-place version of toHex() 2022-01-11 11:51:04 +01:00
CGDebugInfo.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CGDecl.cpp [CodeGen] Make element type in emitArrayDestroy() predictable 2022-01-11 09:25:29 +01:00
CGDeclCXX.cpp [CodeGen] Make element type in emitArrayDestroy() predictable 2022-01-11 09:25:29 +01:00
CGException.cpp [CodeGen] Avoid some pointer element type accesses 2021-12-21 14:16:28 +01:00
CGExpr.cpp [CodeGen] Avoid some pointer element type accesses 2022-01-10 15:02:55 +01:00
CGExprAgg.cpp [CodeGen] Avoid more pointer element type accesses 2021-12-27 12:00:22 +01:00
CGExprCXX.cpp [CodeGen] Use CreateConstInBoundsGEP() in one more place 2021-12-23 14:58:47 +01:00
CGExprComplex.cpp [Matrix] Implement C-style explicit type conversions for matrix types. 2021-04-10 11:48:41 +01:00
CGExprConstant.cpp [Clang] Add __builtin_function_start 2021-12-20 12:55:33 -08:00
CGExprScalar.cpp [CodeGen] Avoid more pointer element type accesses 2021-12-16 15:26:21 +01:00
CGGPUBuiltin.cpp [OpenMP] Lower printf to __llvm_omp_vprintf 2021-11-10 15:30:56 +00:00
CGLoopInfo.cpp [Clang] Ensure vector predication loop metadata is always emitted when pragma is specified. 2021-02-13 17:35:54 -06:00
CGLoopInfo.h [SVE] Add support to vectorize_width loop pragma for scalable vectors 2021-01-08 11:37:27 +00:00
CGNonTrivialStruct.cpp [CodeGen] Prefer CreateElementBitCast() where possible 2021-12-15 11:48:39 +01:00
CGObjC.cpp [clang] Use true/false instead of 1/0 (NFC) 2022-01-09 00:19:47 -08:00
CGObjCGNU.cpp [CodeGen] Avoid deprecated ConstantAddress constructor 2021-12-15 10:42:41 +01:00
CGObjCMac.cpp [CodeGen] Treat ObjC `__unsafe_unretained` and class types as trivial 2022-01-11 11:18:24 -08:00
CGObjCRuntime.cpp [clang] Add range accessor for ObjCAtTryStmt catch_stmts and use it 2021-10-27 08:57:05 -04:00
CGObjCRuntime.h Fix a variety of bugs with nil-receiver checks when targeting 2021-10-08 05:44:06 -04:00
CGOpenCLRuntime.cpp
CGOpenCLRuntime.h
CGOpenMPRuntime.cpp [clang] Fix bugprone argument comments (NFC) 2022-01-09 00:19:49 -08:00
CGOpenMPRuntime.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CGOpenMPRuntimeGPU.cpp [OpenMP][FIX] Change globalization alignment to 16 2021-12-27 16:58:25 -05:00
CGOpenMPRuntimeGPU.h [OpenMP][FIX] Pass the num_threads value directly to parallel_51 2021-12-09 16:30:29 -05:00
CGRecordLayout.h [clang] Remove redundant member initialization (NFC) 2022-01-02 10:20:23 -08:00
CGRecordLayoutBuilder.cpp [clang] Fix bugprone argument comments (NFC) 2022-01-09 00:19:49 -08:00
CGStmt.cpp [clang] number labels in asm goto strings after tied inputs 2022-01-11 12:09:24 -08:00
CGStmtOpenMP.cpp [OMPIRBuilder] Add support for simd (loop) directive. 2022-01-19 11:32:17 -05:00
CGVTT.cpp [AMDGPU] Set the default globals address space to 1 2020-11-20 15:46:53 +00:00
CGVTables.cpp [clang] Use true/false instead of 1/0 (NFC) 2022-01-09 00:19:47 -08:00
CGVTables.h
CGValue.h [CodeGen] Store element type in RValue 2021-12-17 09:05:59 +01:00
CMakeLists.txt [clang][openmp][NFC] Remove arch-specific CGOpenMPRuntimeGPU files 2021-11-09 15:11:05 -05:00
CodeGenABITypes.cpp
CodeGenAction.cpp Remove redundant return and continue statements (NFC) 2021-12-24 23:17:54 -08:00
CodeGenFunction.cpp [clang-cl] Support the /HOTPATCH flag 2022-01-20 12:57:19 -05:00
CodeGenFunction.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CodeGenModule.cpp [X86] Enable ibt-seal optimization when LTO is used in Kernel 2022-01-21 10:55:34 +08:00
CodeGenModule.h [CPU-Dispatch] Make sure Dispatch names get updated if previously mangled 2022-01-14 10:45:55 -08:00
CodeGenPGO.cpp [clang] Remove redundant member initialization (NFC) 2022-01-02 10:20:23 -08:00
CodeGenPGO.h [PGO] Don't reference functions unless value profiling is enabled 2021-05-20 11:09:24 -07:00
CodeGenTBAA.cpp Introduce _BitInt, deprecate _ExtInt 2021-12-06 12:52:01 -05:00
CodeGenTBAA.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
CodeGenTypeCache.h Fix __attribute__((annotate("")) with non-zero globals AS 2021-08-26 10:09:40 +01:00
CodeGenTypes.cpp [clang] Fix function pointer address space 2022-01-13 08:06:19 -08:00
CodeGenTypes.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
ConstantEmitter.h
ConstantInitBuilder.cpp
CoverageMappingGen.cpp [clang] Use isa instead of dyn_cast (NFC) 2021-11-14 09:32:40 -08:00
CoverageMappingGen.h [Driver] Rename -fprofile-{prefix-map,compilation-dir} to -fcoverage-{prefix-map,compilation-dir} 2021-02-25 21:40:12 -08:00
EHScopeStack.h [Windows SEH]: HARDWARE EXCEPTION HANDLING (MSVC -EHa) - Part 1 2021-05-17 22:42:17 -07:00
ItaniumCXXABI.cpp [MS] Implement on-demand TLS initialization for Microsoft CXX ABI 2022-01-13 21:23:23 -08:00
MacroPPCallbacks.cpp
MacroPPCallbacks.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
MicrosoftCXXABI.cpp [MS] Implement on-demand TLS initialization for Microsoft CXX ABI 2022-01-13 21:23:23 -08:00
ModuleBuilder.cpp [clang-repl] Allow Interpreter::getSymbolAddress to take a mangled name. 2021-11-10 12:52:05 +00:00
ObjectFilePCHContainerOperations.cpp Emit the C++ dialect in -gmodules .pcm files. 2022-01-10 16:13:40 -08:00
PatternInit.cpp
PatternInit.h
README.txt Revert "This is a test commit" 2020-12-23 13:04:37 -06:00
SanitizerMetadata.cpp Revert "[asan] Add support for disable_sanitizer_instrumentation attribute" 2021-12-10 14:33:38 -08:00
SanitizerMetadata.h
SwiftCallingConv.cpp Teach the swift calling convention about _Atomic types 2020-08-31 07:07:25 -07:00
TargetInfo.cpp [clang][AVR] Implement '__flashN' for variables on different flash banks 2022-01-19 11:24:01 +00:00
TargetInfo.h [clang] Remove unused forward declarations (NFC) 2022-01-08 11:56:40 -08:00
VarBypassDetector.cpp [clang,NFC] Fix typos in file headers 2021-02-25 12:47:02 -08:00
VarBypassDetector.h Use {DenseSet,SmallPtrSet}::contains (NFC) 2021-10-29 20:26:07 -07:00

README.txt

IRgen optimization opportunities.

//===---------------------------------------------------------------------===//

The common pattern of
--
short x; // or char, etc
(x == 10)
--
generates an zext/sext of x which can easily be avoided.

//===---------------------------------------------------------------------===//

Bitfields accesses can be shifted to simplify masking and sign
extension. For example, if the bitfield width is 8 and it is
appropriately aligned then is is a lot shorter to just load the char
directly.

//===---------------------------------------------------------------------===//

It may be worth avoiding creation of alloca's for formal arguments
for the common situation where the argument is never written to or has
its address taken. The idea would be to begin generating code by using
the argument directly and if its address is taken or it is stored to
then generate the alloca and patch up the existing code.

In theory, the same optimization could be a win for block local
variables as long as the declaration dominates all statements in the
block.

NOTE: The main case we care about this for is for -O0 -g compile time
performance, and in that scenario we will need to emit the alloca
anyway currently to emit proper debug info. So this is blocked by
being able to emit debug information which refers to an LLVM
temporary, not an alloca.

//===---------------------------------------------------------------------===//

We should try and avoid generating basic blocks which only contain
jumps. At -O0, this penalizes us all the way from IRgen (malloc &
instruction overhead), all the way down through code generation and
assembly time.

On 176.gcc:expr.ll, it looks like over 12% of basic blocks are just
direct branches!

//===---------------------------------------------------------------------===//