maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp

900 lines
32 KiB
C++
Raw Normal View History

[PM] Switch the TargetMachine interface from accepting a pass manager base which it adds a single analysis pass to, to instead return the type erased TargetTransformInfo object constructed for that TargetMachine. This removes all of the pass variants for TTI. There is now a single TTI *pass* in the Analysis layer. All of the Analysis <-> Target communication is through the TTI's type erased interface itself. While the diff is large here, it is nothing more that code motion to make types available in a header file for use in a different source file within each target. I've tried to keep all the doxygen comments and file boilerplate in line with this move, but let me know if I missed anything. With this in place, the next step to making TTI work with the new pass manager is to introduce a really simple new-style analysis that produces a TTI object via a callback into this routine on the target machine. Once we have that, we'll have the building blocks necessary to accept a function argument as well. llvm-svn: 227685
2015-01-31 19:17:59 +08:00
//===-- PPCTargetTransformInfo.cpp - PPC specific TTI ---------------------===//
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
//
Update the file headers across all of the LLVM projects in the monorepo to reflect the new license. We understand that people may be surprised that we're moving the header entirely to discuss the new license. We checked this carefully with the Foundation's lawyer and we believe this is the correct approach. Essentially, all code in the project is now made available by the LLVM project under our new license, so you will see that the license headers include that license only. Some of our contributors have contributed code under our old license, and accordingly, we have retained a copy of our old license notice in the top-level files in each project and repository. llvm-svn: 351636
2019-01-19 16:50:56 +08:00
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
//
//===----------------------------------------------------------------------===//
[PM] Switch the TargetMachine interface from accepting a pass manager base which it adds a single analysis pass to, to instead return the type erased TargetTransformInfo object constructed for that TargetMachine. This removes all of the pass variants for TTI. There is now a single TTI *pass* in the Analysis layer. All of the Analysis <-> Target communication is through the TTI's type erased interface itself. While the diff is large here, it is nothing more that code motion to make types available in a header file for use in a different source file within each target. I've tried to keep all the doxygen comments and file boilerplate in line with this move, but let me know if I missed anything. With this in place, the next step to making TTI work with the new pass manager is to introduce a really simple new-style analysis that produces a TTI object via a callback into this routine on the target machine. Once we have that, we'll have the building blocks necessary to accept a function argument as well. llvm-svn: 227685
2015-01-31 19:17:59 +08:00
#include "PPCTargetTransformInfo.h"
[CodeGen] Generic Hardware Loop Support Patch which introduces a target-independent framework for generating hardware loops at the IR level. Most of the code has been taken from PowerPC CTRLoops and PowerPC has been ported over to use this generic pass. The target dependent parts have been moved into TargetTransformInfo, via isHardwareLoopProfitable, with HardwareLoopInfo introduced to transfer information from the backend. Three generic intrinsics have been introduced: - void @llvm.set_loop_iterations Takes as a single operand, the number of iterations to be executed. - i1 @llvm.loop_decrement(anyint) Takes the maximum number of elements processed in an iteration of the loop body and subtracts this from the total count. Returns false when the loop should exit. - anyint @llvm.loop_decrement_reg(anyint, anyint) Takes the number of elements remaining to be processed as well as the maximum numbe of elements processed in an iteration of the loop body. Returns the updated number of elements remaining. llvm-svn: 362774
2019-06-07 15:35:30 +08:00
#include "llvm/Analysis/CodeMetrics.h"
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
#include "llvm/Analysis/TargetTransformInfo.h"
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
#include "llvm/CodeGen/BasicTTIImpl.h"
Fix a bunch more layering of CodeGen headers that are in Target All these headers already depend on CodeGen headers so moving them into CodeGen fixes the layering (since CodeGen depends on Target, not the other way around). llvm-svn: 318490
2017-11-17 09:07:10 +08:00
#include "llvm/CodeGen/CostTable.h"
#include "llvm/CodeGen/TargetLowering.h"
[CodeGen] Generic Hardware Loop Support Patch which introduces a target-independent framework for generating hardware loops at the IR level. Most of the code has been taken from PowerPC CTRLoops and PowerPC has been ported over to use this generic pass. The target dependent parts have been moved into TargetTransformInfo, via isHardwareLoopProfitable, with HardwareLoopInfo introduced to transfer information from the backend. Three generic intrinsics have been introduced: - void @llvm.set_loop_iterations Takes as a single operand, the number of iterations to be executed. - i1 @llvm.loop_decrement(anyint) Takes the maximum number of elements processed in an iteration of the loop body and subtracts this from the total count. Returns false when the loop should exit. - anyint @llvm.loop_decrement_reg(anyint, anyint) Takes the number of elements remaining to be processed as well as the maximum numbe of elements processed in an iteration of the loop body. Returns the updated number of elements remaining. llvm-svn: 362774
2019-06-07 15:35:30 +08:00
#include "llvm/CodeGen/TargetSchedule.h"
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
#include "llvm/Support/CommandLine.h"
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
#include "llvm/Support/Debug.h"
using namespace llvm;
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE definition below all of the header #include lines, lib/Target/... edition. llvm-svn: 206842
2014-04-22 10:41:26 +08:00
#define DEBUG_TYPE "ppctti"
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
static cl::opt<bool> DisablePPCConstHoist("disable-ppc-constant-hoisting",
cl::desc("disable constant hoisting on PPC"), cl::init(false), cl::Hidden);
[TTI] Add getCacheLineSize Summary: And use it in PPCLoopDataPrefetch.cpp. @hfinkel, please let me know if your preference would be to preserve the ppc-loop-prefetch-cache-line option in order to be able to override the value of TTI::getCacheLineSize for PPC. Reviewers: hfinkel Subscribers: hulx2000, mcrosier, mssimpso, hfinkel, llvm-commits Differential Revision: http://reviews.llvm.org/D16306 llvm-svn: 258419
2016-01-22 02:28:36 +08:00
// This is currently only used for the data prefetch pass which is only enabled
// for BG/Q by default.
static cl::opt<unsigned>
CacheLineSize("ppc-loop-prefetch-cache-line", cl::Hidden, cl::init(64),
cl::desc("The loop prefetch cache line size"));
Re-commit : [PowerPC] Add handling for ColdCC calling convention and a pass to mark candidates with coldcc attribute. This recommits r322721 reverted due to sanitizer memory leak build bot failures. Original commit message: This patch adds support for the coldcc calling convention for Power. This changes the set of non-volatile registers. It includes a pass to stress test the implementation by marking all static directly called functions with the coldcc attribute through the option -enable-coldcc-stress-test. It also includes an option, -ppc-enable-coldcc, to add the coldcc attribute to functions which are cold at all call sites based on BlockFrequencyInfo when the containing function does not call any non cold functions. Differential Revision: https://reviews.llvm.org/D38413 llvm-svn: 323778
2018-01-31 00:17:22 +08:00
static cl::opt<bool>
EnablePPCColdCC("ppc-enable-coldcc", cl::Hidden, cl::init(false),
cl::desc("Enable using coldcc calling conv for cold "
"internal functions"));
[CodeGen] Generic Hardware Loop Support Patch which introduces a target-independent framework for generating hardware loops at the IR level. Most of the code has been taken from PowerPC CTRLoops and PowerPC has been ported over to use this generic pass. The target dependent parts have been moved into TargetTransformInfo, via isHardwareLoopProfitable, with HardwareLoopInfo introduced to transfer information from the backend. Three generic intrinsics have been introduced: - void @llvm.set_loop_iterations Takes as a single operand, the number of iterations to be executed. - i1 @llvm.loop_decrement(anyint) Takes the maximum number of elements processed in an iteration of the loop body and subtracts this from the total count. Returns false when the loop should exit. - anyint @llvm.loop_decrement_reg(anyint, anyint) Takes the number of elements remaining to be processed as well as the maximum numbe of elements processed in an iteration of the loop body. Returns the updated number of elements remaining. llvm-svn: 362774
2019-06-07 15:35:30 +08:00
// The latency of mtctr is only justified if there are more than 4
// comparisons that will be removed as a result.
static cl::opt<unsigned>
SmallCTRLoopThreshold("min-ctr-loop-threshold", cl::init(4), cl::Hidden,
cl::desc("Loops with a constant trip count smaller than "
"this value will not use the count register."));
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
//===----------------------------------------------------------------------===//
//
// PPC cost model.
//
//===----------------------------------------------------------------------===//
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
TargetTransformInfo::PopcntSupportKind
PPCTTIImpl::getPopcntSupport(unsigned TyWidth) {
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
assert(isPowerOf2_32(TyWidth) && "Ty width must be power of 2");
[PowerPC] Refactor popcnt[dw] target features Instead of using two feature bits, one to indicate the availability of the popcnt[dw] instructions, and another to indicate whether or not they're fast, use a single enum. This allows more consistent control via target attribute strings, and via Clang's command line. llvm-svn: 264690
2016-03-29 09:36:01 +08:00
if (ST->hasPOPCNTD() != PPCSubtarget::POPCNTD_Unavailable && TyWidth <= 64)
return ST->hasPOPCNTD() == PPCSubtarget::POPCNTD_Slow ?
TTI::PSK_SlowHardware : TTI::PSK_FastHardware;
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::PSK_Software;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getIntImmCost(const APInt &Imm, Type *Ty) {
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (DisablePPCConstHoist)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return BaseT::getIntImmCost(Imm, Ty);
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
assert(Ty->isIntegerTy());
unsigned BitSize = Ty->getPrimitiveSizeInBits();
if (BitSize == 0)
return ~0U;
if (Imm == 0)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (Imm.getBitWidth() <= 64) {
if (isInt<16>(Imm.getSExtValue()))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Basic;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (isInt<32>(Imm.getSExtValue())) {
// A constant that can be materialized using lis.
if ((Imm.getZExtValue() & 0xFFFF) == 0)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Basic;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return 2 * TTI::TCC_Basic;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
}
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return 4 * TTI::TCC_Basic;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getIntImmCost(Intrinsic::ID IID, unsigned Idx, const APInt &Imm,
Type *Ty) {
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (DisablePPCConstHoist)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return BaseT::getIntImmCost(IID, Idx, Imm, Ty);
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
assert(Ty->isIntegerTy());
unsigned BitSize = Ty->getPrimitiveSizeInBits();
if (BitSize == 0)
return ~0U;
switch (IID) {
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
default:
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Intrinsic::sadd_with_overflow:
case Intrinsic::uadd_with_overflow:
case Intrinsic::ssub_with_overflow:
case Intrinsic::usub_with_overflow:
if ((Idx == 1) && Imm.getBitWidth() <= 64 && isInt<16>(Imm.getSExtValue()))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
break;
Revert "r225811 - Revert "r225808 - [PowerPC] Add StackMap/PatchPoint support"" This re-applies r225808, fixed to avoid problems with SDAG dependencies along with the preceding fix to ScheduleDAGSDNodes::RegDefIter::InitNodeNumDefs. These problems caused the original regression tests to assert/segfault on many (but not all) systems. Original commit message: This commit does two things: 1. Refactors PPCFastISel to use more of the common infrastructure for call lowering (this lets us take advantage of this common code for lowering some common intrinsics, stackmap/patchpoint among them). 2. Adds support for stackmap/patchpoint lowering. For the most part, this is very similar to the support in the AArch64 target, with the obvious differences (different registers, NOP instructions, etc.). The test cases are adapted from the AArch64 test cases. One difference of note is that the patchpoint call sequence takes 24 bytes, so you can't use less than that (on AArch64 you can go down to 16). Also, as noted in the docs, we take the patchpoint address to be the actual code address (assuming the call is local in the TOC-sharing sense), which should yield higher performance than generating the full cross-DSO indirect-call sequence and is likely just as useful for JITed code (if not, we'll change it). StackMaps and Patchpoints are still marked as experimental, and so this support is doubly experimental. So go ahead and experiment! llvm-svn: 225909
2015-01-14 09:07:51 +08:00
case Intrinsic::experimental_stackmap:
if ((Idx < 2) || (Imm.getBitWidth() <= 64 && isInt<64>(Imm.getSExtValue())))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
Revert "r225811 - Revert "r225808 - [PowerPC] Add StackMap/PatchPoint support"" This re-applies r225808, fixed to avoid problems with SDAG dependencies along with the preceding fix to ScheduleDAGSDNodes::RegDefIter::InitNodeNumDefs. These problems caused the original regression tests to assert/segfault on many (but not all) systems. Original commit message: This commit does two things: 1. Refactors PPCFastISel to use more of the common infrastructure for call lowering (this lets us take advantage of this common code for lowering some common intrinsics, stackmap/patchpoint among them). 2. Adds support for stackmap/patchpoint lowering. For the most part, this is very similar to the support in the AArch64 target, with the obvious differences (different registers, NOP instructions, etc.). The test cases are adapted from the AArch64 test cases. One difference of note is that the patchpoint call sequence takes 24 bytes, so you can't use less than that (on AArch64 you can go down to 16). Also, as noted in the docs, we take the patchpoint address to be the actual code address (assuming the call is local in the TOC-sharing sense), which should yield higher performance than generating the full cross-DSO indirect-call sequence and is likely just as useful for JITed code (if not, we'll change it). StackMaps and Patchpoints are still marked as experimental, and so this support is doubly experimental. So go ahead and experiment! llvm-svn: 225909
2015-01-14 09:07:51 +08:00
break;
case Intrinsic::experimental_patchpoint_void:
case Intrinsic::experimental_patchpoint_i64:
if ((Idx < 4) || (Imm.getBitWidth() <= 64 && isInt<64>(Imm.getSExtValue())))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
Revert "r225811 - Revert "r225808 - [PowerPC] Add StackMap/PatchPoint support"" This re-applies r225808, fixed to avoid problems with SDAG dependencies along with the preceding fix to ScheduleDAGSDNodes::RegDefIter::InitNodeNumDefs. These problems caused the original regression tests to assert/segfault on many (but not all) systems. Original commit message: This commit does two things: 1. Refactors PPCFastISel to use more of the common infrastructure for call lowering (this lets us take advantage of this common code for lowering some common intrinsics, stackmap/patchpoint among them). 2. Adds support for stackmap/patchpoint lowering. For the most part, this is very similar to the support in the AArch64 target, with the obvious differences (different registers, NOP instructions, etc.). The test cases are adapted from the AArch64 test cases. One difference of note is that the patchpoint call sequence takes 24 bytes, so you can't use less than that (on AArch64 you can go down to 16). Also, as noted in the docs, we take the patchpoint address to be the actual code address (assuming the call is local in the TOC-sharing sense), which should yield higher performance than generating the full cross-DSO indirect-call sequence and is likely just as useful for JITed code (if not, we'll change it). StackMaps and Patchpoints are still marked as experimental, and so this support is doubly experimental. So go ahead and experiment! llvm-svn: 225909
2015-01-14 09:07:51 +08:00
break;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return PPCTTIImpl::getIntImmCost(Imm, Ty);
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm,
Type *Ty) {
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (DisablePPCConstHoist)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return BaseT::getIntImmCost(Opcode, Idx, Imm, Ty);
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
assert(Ty->isIntegerTy());
unsigned BitSize = Ty->getPrimitiveSizeInBits();
if (BitSize == 0)
return ~0U;
unsigned ImmIdx = ~0U;
bool ShiftedFree = false, RunFree = false, UnsignedFree = false,
ZeroFree = false;
switch (Opcode) {
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
default:
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Instruction::GetElementPtr:
// Always hoist the base address of a GetElementPtr. This prevents the
// creation of new constants for every base constant that gets constant
// folded with the offset.
if (Idx == 0)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return 2 * TTI::TCC_Basic;
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Instruction::And:
RunFree = true; // (for the rotate-and-mask instructions)
Replace "fallthrough" comments with LLVM_FALLTHROUGH This is a mechanical change of comments in switches like fallthrough, fall-through, or fall-thru to use the LLVM_FALLTHROUGH macro instead. llvm-svn: 278902
2016-08-17 13:10:15 +08:00
LLVM_FALLTHROUGH;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Instruction::Add:
case Instruction::Or:
case Instruction::Xor:
ShiftedFree = true;
Replace "fallthrough" comments with LLVM_FALLTHROUGH This is a mechanical change of comments in switches like fallthrough, fall-through, or fall-thru to use the LLVM_FALLTHROUGH macro instead. llvm-svn: 278902
2016-08-17 13:10:15 +08:00
LLVM_FALLTHROUGH;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Instruction::Sub:
case Instruction::Mul:
case Instruction::Shl:
case Instruction::LShr:
case Instruction::AShr:
ImmIdx = 1;
break;
case Instruction::ICmp:
UnsignedFree = true;
ImmIdx = 1;
Replace "fallthrough" comments with LLVM_FALLTHROUGH This is a mechanical change of comments in switches like fallthrough, fall-through, or fall-thru to use the LLVM_FALLTHROUGH macro instead. llvm-svn: 278902
2016-08-17 13:10:15 +08:00
// Zero comparisons can use record-form instructions.
LLVM_FALLTHROUGH;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
case Instruction::Select:
ZeroFree = true;
break;
case Instruction::PHI:
case Instruction::Call:
case Instruction::Ret:
case Instruction::Load:
case Instruction::Store:
break;
}
if (ZeroFree && Imm == 0)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (Idx == ImmIdx && Imm.getBitWidth() <= 64) {
if (isInt<16>(Imm.getSExtValue()))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (RunFree) {
if (Imm.getBitWidth() <= 32 &&
(isShiftedMask_32(Imm.getZExtValue()) ||
isShiftedMask_32(~Imm.getZExtValue())))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (ST->isPPC64() &&
(isShiftedMask_64(Imm.getZExtValue()) ||
isShiftedMask_64(~Imm.getZExtValue())))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
if (UnsignedFree && isUInt<16>(Imm.getZExtValue()))
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
if (ShiftedFree && (Imm.getZExtValue() & 0xFFFF) == 0)
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return TTI::TCC_Free;
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
return PPCTTIImpl::getIntImmCost(Imm, Ty);
[PowerPC] [Constant Hoisting] Enable constant hoisting on PPC Implements the various TTI functions to enable constant hoisting on PPC. The only significant test-suite change is this: MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup (which essentially reverses the slowdown from r206120). llvm-svn: 206141
2014-04-14 07:02:40 +08:00
}
The cost of splitting a large vector instruction is not being taken into account by the getUserCost function. This was leading to some loops being over unrolled. The cost of a vector instruction is now being multiplied by the cost of the type legalization. This will return a more accurate cost. Committing on behalf on Brad Nemanich (brad.nemanich@ibm.com) Differential Revision: https://reviews.llvm.org/D38961 llvm-svn: 316174
2017-10-20 02:16:31 +08:00
unsigned PPCTTIImpl::getUserCost(const User *U,
ArrayRef<const Value *> Operands) {
if (U->getType()->isVectorTy()) {
// Instructions that need to be split should cost more.
std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, U->getType());
return LT.first * BaseT::getUserCost(U, Operands);
}
Remove trailing space sed -Ei 's/[[:space:]]+$//' include/**/*.{def,h,td} lib/**/*.{cpp,h} llvm-svn: 338293
2018-07-31 03:41:25 +08:00
The cost of splitting a large vector instruction is not being taken into account by the getUserCost function. This was leading to some loops being over unrolled. The cost of a vector instruction is now being multiplied by the cost of the type legalization. This will return a more accurate cost. Committing on behalf on Brad Nemanich (brad.nemanich@ibm.com) Differential Revision: https://reviews.llvm.org/D38961 llvm-svn: 316174
2017-10-20 02:16:31 +08:00
return BaseT::getUserCost(U, Operands);
}
[CodeGen] Generic Hardware Loop Support Patch which introduces a target-independent framework for generating hardware loops at the IR level. Most of the code has been taken from PowerPC CTRLoops and PowerPC has been ported over to use this generic pass. The target dependent parts have been moved into TargetTransformInfo, via isHardwareLoopProfitable, with HardwareLoopInfo introduced to transfer information from the backend. Three generic intrinsics have been introduced: - void @llvm.set_loop_iterations Takes as a single operand, the number of iterations to be executed. - i1 @llvm.loop_decrement(anyint) Takes the maximum number of elements processed in an iteration of the loop body and subtracts this from the total count. Returns false when the loop should exit. - anyint @llvm.loop_decrement_reg(anyint, anyint) Takes the number of elements remaining to be processed as well as the maximum numbe of elements processed in an iteration of the loop body. Returns the updated number of elements remaining. llvm-svn: 362774
2019-06-07 15:35:30 +08:00
bool PPCTTIImpl::mightUseCTR(BasicBlock *BB,
TargetLibraryInfo *LibInfo) {
const PPCTargetMachine &TM = ST->getTargetMachine();
// Loop through the inline asm constraints and look for something that
// clobbers ctr.
auto asmClobbersCTR = [](InlineAsm *IA) {
InlineAsm::ConstraintInfoVector CIV = IA->ParseConstraints();
for (unsigned i = 0, ie = CIV.size(); i < ie; ++i) {
InlineAsm::ConstraintInfo &C = CIV[i];
if (C.Type != InlineAsm::isInput)
for (unsigned j = 0, je = C.Codes.size(); j < je; ++j)
if (StringRef(C.Codes[j]).equals_lower("{ctr}"))
return true;
}
return false;
};
// Determining the address of a TLS variable results in a function call in
// certain TLS models.
std::function<bool(const Value*)> memAddrUsesCTR =
[&memAddrUsesCTR, &TM](const Value *MemAddr) -> bool {
const auto *GV = dyn_cast<GlobalValue>(MemAddr);
if (!GV) {
// Recurse to check for constants that refer to TLS global variables.
if (const auto *CV = dyn_cast<Constant>(MemAddr))
for (const auto &CO : CV->operands())
if (memAddrUsesCTR(CO))
return true;
return false;
}
if (!GV->isThreadLocal())
return false;
TLSModel::Model Model = TM.getTLSModel(GV);
return Model == TLSModel::GeneralDynamic ||
Model == TLSModel::LocalDynamic;
};
auto isLargeIntegerTy = [](bool Is32Bit, Type *Ty) {
if (IntegerType *ITy = dyn_cast<IntegerType>(Ty))
return ITy->getBitWidth() > (Is32Bit ? 32U : 64U);
return false;
};
for (BasicBlock::iterator J = BB->begin(), JE = BB->end();
J != JE; ++J) {
if (CallInst *CI = dyn_cast<CallInst>(J)) {
// Inline ASM is okay, unless it clobbers the ctr register.
if (InlineAsm *IA = dyn_cast<InlineAsm>(CI->getCalledValue())) {
if (asmClobbersCTR(IA))
return true;
continue;
}
if (Function *F = CI->getCalledFunction()) {
// Most intrinsics don't become function calls, but some might.
// sin, cos, exp and log are always calls.
unsigned Opcode = 0;
if (F->getIntrinsicID() != Intrinsic::not_intrinsic) {
switch (F->getIntrinsicID()) {
default: continue;
// If we have a call to ppc_is_decremented_ctr_nonzero, or ppc_mtctr
// we're definitely using CTR.
case Intrinsic::set_loop_iterations:
case Intrinsic::loop_decrement:
return true;
// VisualStudio defines setjmp as _setjmp
#if defined(_MSC_VER) && defined(setjmp) && \
!defined(setjmp_undefined_for_msvc)
# pragma push_macro("setjmp")
# undef setjmp
# define setjmp_undefined_for_msvc
#endif
case Intrinsic::setjmp:
#if defined(_MSC_VER) && defined(setjmp_undefined_for_msvc)
// let's return it to _setjmp state
# pragma pop_macro("setjmp")
# undef setjmp_undefined_for_msvc
#endif
case Intrinsic::longjmp:
// Exclude eh_sjlj_setjmp; we don't need to exclude eh_sjlj_longjmp
// because, although it does clobber the counter register, the
// control can't then return to inside the loop unless there is also
// an eh_sjlj_setjmp.
case Intrinsic::eh_sjlj_setjmp:
case Intrinsic::memcpy:
case Intrinsic::memmove:
case Intrinsic::memset:
case Intrinsic::powi:
case Intrinsic::log:
case Intrinsic::log2:
case Intrinsic::log10:
case Intrinsic::exp:
case Intrinsic::exp2:
case Intrinsic::pow:
case Intrinsic::sin:
case Intrinsic::cos:
return true;
case Intrinsic::copysign:
if (CI->getArgOperand(0)->getType()->getScalarType()->
isPPC_FP128Ty())
return true;
else
continue; // ISD::FCOPYSIGN is never a library call.
case Intrinsic::sqrt: Opcode = ISD::FSQRT; break;
case Intrinsic::floor: Opcode = ISD::FFLOOR; break;
case Intrinsic::ceil: Opcode = ISD::FCEIL; break;
case Intrinsic::trunc: Opcode = ISD::FTRUNC; break;
case Intrinsic::rint: Opcode = ISD::FRINT; break;
case Intrinsic::nearbyint: Opcode = ISD::FNEARBYINT; break;
case Intrinsic::round: Opcode = ISD::FROUND; break;
case Intrinsic::minnum: Opcode = ISD::FMINNUM; break;
case Intrinsic::maxnum: Opcode = ISD::FMAXNUM; break;
case Intrinsic::umul_with_overflow: Opcode = ISD::UMULO; break;
case Intrinsic::smul_with_overflow: Opcode = ISD::SMULO; break;
}
}
// PowerPC does not use [US]DIVREM or other library calls for
// operations on regular types which are not otherwise library calls
// (i.e. soft float or atomics). If adapting for targets that do,
// additional care is required here.
LibFunc Func;
if (!F->hasLocalLinkage() && F->hasName() && LibInfo &&
LibInfo->getLibFunc(F->getName(), Func) &&
LibInfo->hasOptimizedCodeGen(Func)) {
// Non-read-only functions are never treated as intrinsics.
if (!CI->onlyReadsMemory())
return true;
// Conversion happens only for FP calls.
if (!CI->getArgOperand(0)->getType()->isFloatingPointTy())
return true;
switch (Func) {
default: return true;
case LibFunc_copysign:
case LibFunc_copysignf:
continue; // ISD::FCOPYSIGN is never a library call.
case LibFunc_copysignl:
return true;
case LibFunc_fabs:
case LibFunc_fabsf:
case LibFunc_fabsl:
continue; // ISD::FABS is never a library call.
case LibFunc_sqrt:
case LibFunc_sqrtf:
case LibFunc_sqrtl:
Opcode = ISD::FSQRT; break;
case LibFunc_floor:
case LibFunc_floorf:
case LibFunc_floorl:
Opcode = ISD::FFLOOR; break;
case LibFunc_nearbyint:
case LibFunc_nearbyintf:
case LibFunc_nearbyintl:
Opcode = ISD::FNEARBYINT; break;
case LibFunc_ceil:
case LibFunc_ceilf:
case LibFunc_ceill:
Opcode = ISD::FCEIL; break;
case LibFunc_rint:
case LibFunc_rintf:
case LibFunc_rintl:
Opcode = ISD::FRINT; break;
case LibFunc_round:
case LibFunc_roundf:
case LibFunc_roundl:
Opcode = ISD::FROUND; break;
case LibFunc_trunc:
case LibFunc_truncf:
case LibFunc_truncl:
Opcode = ISD::FTRUNC; break;
case LibFunc_fmin:
case LibFunc_fminf:
case LibFunc_fminl:
Opcode = ISD::FMINNUM; break;
case LibFunc_fmax:
case LibFunc_fmaxf:
case LibFunc_fmaxl:
Opcode = ISD::FMAXNUM; break;
}
}
if (Opcode) {
EVT EVTy =
TLI->getValueType(DL, CI->getArgOperand(0)->getType(), true);
if (EVTy == MVT::Other)
return true;
if (TLI->isOperationLegalOrCustom(Opcode, EVTy))
continue;
else if (EVTy.isVector() &&
TLI->isOperationLegalOrCustom(Opcode, EVTy.getScalarType()))
continue;
return true;
}
}
return true;
} else if (isa<BinaryOperator>(J) &&
J->getType()->getScalarType()->isPPC_FP128Ty()) {
// Most operations on ppc_f128 values become calls.
return true;
} else if (isa<UIToFPInst>(J) || isa<SIToFPInst>(J) ||
isa<FPToUIInst>(J) || isa<FPToSIInst>(J)) {
CastInst *CI = cast<CastInst>(J);
if (CI->getSrcTy()->getScalarType()->isPPC_FP128Ty() ||
CI->getDestTy()->getScalarType()->isPPC_FP128Ty() ||
isLargeIntegerTy(!TM.isPPC64(), CI->getSrcTy()->getScalarType()) ||
isLargeIntegerTy(!TM.isPPC64(), CI->getDestTy()->getScalarType()))
return true;
} else if (isLargeIntegerTy(!TM.isPPC64(),
J->getType()->getScalarType()) &&
(J->getOpcode() == Instruction::UDiv ||
J->getOpcode() == Instruction::SDiv ||
J->getOpcode() == Instruction::URem ||
J->getOpcode() == Instruction::SRem)) {
return true;
} else if (!TM.isPPC64() &&
isLargeIntegerTy(false, J->getType()->getScalarType()) &&
(J->getOpcode() == Instruction::Shl ||
J->getOpcode() == Instruction::AShr ||
J->getOpcode() == Instruction::LShr)) {
// Only on PPC32, for 128-bit integers (specifically not 64-bit
// integers), these might be runtime calls.
return true;
} else if (isa<IndirectBrInst>(J) || isa<InvokeInst>(J)) {
// On PowerPC, indirect jumps use the counter register.
return true;
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(J)) {
if (SI->getNumCases() + 1 >= (unsigned)TLI->getMinimumJumpTableEntries())
return true;
}
// FREM is always a call.
if (J->getOpcode() == Instruction::FRem)
return true;
if (ST->useSoftFloat()) {
switch(J->getOpcode()) {
case Instruction::FAdd:
case Instruction::FSub:
case Instruction::FMul:
case Instruction::FDiv:
case Instruction::FPTrunc:
case Instruction::FPExt:
case Instruction::FPToUI:
case Instruction::FPToSI:
case Instruction::UIToFP:
case Instruction::SIToFP:
case Instruction::FCmp:
return true;
}
}
for (Value *Operand : J->operands())
if (memAddrUsesCTR(Operand))
return true;
}
return false;
}
bool PPCTTIImpl::isHardwareLoopProfitable(Loop *L, ScalarEvolution &SE,
AssumptionCache &AC,
TargetLibraryInfo *LibInfo,
[NFC] move some hardware loop checking code to a common place for other using. Differential Revision: https://reviews.llvm.org/D63478 llvm-svn: 363758
2019-06-19 09:26:31 +08:00
HardwareLoopInfo &HWLoopInfo) {
[CodeGen] Generic Hardware Loop Support Patch which introduces a target-independent framework for generating hardware loops at the IR level. Most of the code has been taken from PowerPC CTRLoops and PowerPC has been ported over to use this generic pass. The target dependent parts have been moved into TargetTransformInfo, via isHardwareLoopProfitable, with HardwareLoopInfo introduced to transfer information from the backend. Three generic intrinsics have been introduced: - void @llvm.set_loop_iterations Takes as a single operand, the number of iterations to be executed. - i1 @llvm.loop_decrement(anyint) Takes the maximum number of elements processed in an iteration of the loop body and subtracts this from the total count. Returns false when the loop should exit. - anyint @llvm.loop_decrement_reg(anyint, anyint) Takes the number of elements remaining to be processed as well as the maximum numbe of elements processed in an iteration of the loop body. Returns the updated number of elements remaining. llvm-svn: 362774
2019-06-07 15:35:30 +08:00
const PPCTargetMachine &TM = ST->getTargetMachine();
TargetSchedModel SchedModel;
SchedModel.init(ST);
// Do not convert small short loops to CTR loop.
unsigned ConstTripCount = SE.getSmallConstantTripCount(L);
if (ConstTripCount && ConstTripCount < SmallCTRLoopThreshold) {
SmallPtrSet<const Value *, 32> EphValues;
CodeMetrics::collectEphemeralValues(L, &AC, EphValues);
CodeMetrics Metrics;
for (BasicBlock *BB : L->blocks())
Metrics.analyzeBasicBlock(BB, *this, EphValues);
// 6 is an approximate latency for the mtctr instruction.
if (Metrics.NumInsts <= (6 * SchedModel.getIssueWidth()))
return false;
}
// We don't want to spill/restore the counter register, and so we don't
// want to use the counter register if the loop contains calls.
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I)
if (mightUseCTR(*I, LibInfo))
return false;
SmallVector<BasicBlock*, 4> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
// If there is an exit edge known to be frequently taken,
// we should not transform this loop.
for (auto &BB : ExitingBlocks) {
Instruction *TI = BB->getTerminator();
if (!TI) continue;
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
uint64_t TrueWeight = 0, FalseWeight = 0;
if (!BI->isConditional() ||
!BI->extractProfMetadata(TrueWeight, FalseWeight))
continue;
// If the exit path is more frequent than the loop path,
// we return here without further analysis for this loop.
bool TrueIsExit = !L->contains(BI->getSuccessor(0));
if (( TrueIsExit && FalseWeight < TrueWeight) ||
(!TrueIsExit && FalseWeight > TrueWeight))
return false;
}
}
LLVMContext &C = L->getHeader()->getContext();
HWLoopInfo.CountType = TM.isPPC64() ?
Type::getInt64Ty(C) : Type::getInt32Ty(C);
HWLoopInfo.LoopDecrement = ConstantInt::get(HWLoopInfo.CountType, 1);
return true;
}
[LoopUnroll] Pass SCEV to getUnrollingPreferences hook. NFCI. Reviewers: sanjoy, anna, reames, apilipenko, igor-laevsky, mkuper Subscribers: jholewinski, arsenm, mzolotukhin, nemanjai, nhaehnle, javed.absar, mcrosier, llvm-commits Differential Revision: https://reviews.llvm.org/D34531 llvm-svn: 306554
2017-06-28 23:53:17 +08:00
void PPCTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
TTI::UnrollingPreferences &UP) {
[multiversion] Switch the TTI queries from TargetMachine to Subtarget now that we have a correct and cached subtarget specific to the function. Also, finish providing a cached per-function subtarget in the core LLVMTargetMachine -- that layer hadn't switched over yet. The only use of the TargetMachine was to re-lookup a subtarget for a particular function to work around the fact that TTI was immutable. Now that it is per-function and we haved a cached subtarget, use it. This still leaves a few interfaces with real warts on them where we were passing Function objects through the TTI interface. I'll remove these and clean their usage up in subsequent commits now that this isn't necessary. llvm-svn: 227738
2015-02-01 22:22:17 +08:00
if (ST->getDarwinDirective() == PPC::DIR_A2) {
Implement TTI getUnrollingPreferences for PowerPC The PowerPC A2 core greatly benefits from aggressive concatenation unrolling; use the new getUnrollingPreferences to enable this by default when targeting the PPC A2 core. llvm-svn: 190549
2013-09-12 05:20:40 +08:00
// The A2 is in-order with a deep pipeline, and concatenation unrolling
// helps expose latency-hiding opportunities to the instruction scheduler.
UP.Partial = UP.Runtime = true;
[PPC/LoopUnrollRuntime] Don't avoid high-cost trip count computation on the PPC/A2 On X86 (and similar OOO cores) unrolling is very limited, and even if the runtime unrolling is otherwise profitable, the expense of a division to compute the trip count could greatly outweigh the benefits. On the A2, we unroll a lot, and the benefits of unrolling are more significant (seeing a 5x or 6x speedup is not uncommon), so we're more able to tolerate the expense, on average, of a division to compute the trip count. llvm-svn: 237947
2015-05-22 04:30:23 +08:00
// We unroll a lot on the A2 (hundreds of instructions), and the benefits
// often outweigh the cost of a division to compute the trip count.
UP.AllowExpensiveTripCount = true;
Implement TTI getUnrollingPreferences for PowerPC The PowerPC A2 core greatly benefits from aggressive concatenation unrolling; use the new getUnrollingPreferences to enable this by default when targeting the PPC A2 core. llvm-svn: 190549
2013-09-12 05:20:40 +08:00
}
[PowerPC] Enable late partial unrolling on the POWER7 The P7 benefits from not have really-small loops so that we either have multiple dispatch groups in the loop and/or the ability to form more-full dispatch groups during scheduling. Setting the partial unrolling threshold to 44 seems good, empirically, for the P7. Compared to using no late partial unrolling, this yields the following test-suite speedups: SingleSource/Benchmarks/Adobe-C++/simple_types_constant_folding -66.3253% +/- 24.1975% SingleSource/Benchmarks/Misc-C++/oopack_v1p8 -44.0169% +/- 29.4881% SingleSource/Benchmarks/Misc/pi -27.8351% +/- 12.2712% SingleSource/Benchmarks/Stanford/Bubblesort -30.9898% +/- 22.4647% I've speculatively added a similar setting for the P8. Also, I've noticed that the unroller does not quite calculate the unrolling factor correctly for really tiny loops because it neglects to account for the fact that not every loop body replicant contains an ending branch and counter increment. I'll fix that later. llvm-svn: 225522
2015-01-09 23:51:16 +08:00
[LoopUnroll] Pass SCEV to getUnrollingPreferences hook. NFCI. Reviewers: sanjoy, anna, reames, apilipenko, igor-laevsky, mkuper Subscribers: jholewinski, arsenm, mzolotukhin, nemanjai, nhaehnle, javed.absar, mcrosier, llvm-commits Differential Revision: https://reviews.llvm.org/D34531 llvm-svn: 306554
2017-06-28 23:53:17 +08:00
BaseT::getUnrollingPreferences(L, SE, UP);
Implement TTI getUnrollingPreferences for PowerPC The PowerPC A2 core greatly benefits from aggressive concatenation unrolling; use the new getUnrollingPreferences to enable this by default when targeting the PPC A2 core. llvm-svn: 190549
2013-09-12 05:20:40 +08:00
}
Re-commit : [PowerPC] Add handling for ColdCC calling convention and a pass to mark candidates with coldcc attribute. This recommits r322721 reverted due to sanitizer memory leak build bot failures. Original commit message: This patch adds support for the coldcc calling convention for Power. This changes the set of non-volatile registers. It includes a pass to stress test the implementation by marking all static directly called functions with the coldcc attribute through the option -enable-coldcc-stress-test. It also includes an option, -ppc-enable-coldcc, to add the coldcc attribute to functions which are cold at all call sites based on BlockFrequencyInfo when the containing function does not call any non cold functions. Differential Revision: https://reviews.llvm.org/D38413 llvm-svn: 323778
2018-01-31 00:17:22 +08:00
// This function returns true to allow using coldcc calling convention.
// Returning true results in coldcc being used for functions which are cold at
// all call sites when the callers of the functions are not calling any other
// non coldcc functions.
bool PPCTTIImpl::useColdCCForColdCall(Function &F) {
return EnablePPCColdCC;
}
Do not restrict interleaved unrolling to small loops, depending on the target. llvm-svn: 231528
2015-03-07 07:12:04 +08:00
bool PPCTTIImpl::enableAggressiveInterleaving(bool LoopHasReductions) {
[PowerPC] Always use aggressive interleaving on the A2 On the A2, with an eye toward QPX unaligned-load merging, we should always use aggressive interleaving. It is generally superior to only using concatenation unrolling. llvm-svn: 246819
2015-09-04 07:23:00 +08:00
// On the A2, always unroll aggressively. For QPX unaligned loads, we depend
// on combining the loads generated for consecutive accesses, and failure to
// do so is particularly expensive. This makes it much more likely (compared
// to only using concatenation unrolling).
if (ST->getDarwinDirective() == PPC::DIR_A2)
return true;
Do not restrict interleaved unrolling to small loops, depending on the target. llvm-svn: 231528
2015-03-07 07:12:04 +08:00
return LoopHasReductions;
}
[ExpandMemCmp] Move all options to TargetTransformInfo. Split off from D60318. llvm-svn: 364281
2019-06-25 16:04:13 +08:00
PPCTTIImpl::TTI::MemCmpExpansionOptions
PPCTTIImpl::enableMemCmpExpansion(bool OptSize, bool IsZeroCmp) const {
TTI::MemCmpExpansionOptions Options;
Options.LoadSizes = {8, 4, 2, 1};
Options.MaxNumLoads = TLI->getMaxExpandSizeMemcmp(OptSize);
return Options;
[PPC] Inline expansion of memcmp This patch does an inline expansion of memcmp. It changes the memcmp library call into an inline expansion when the size is known at compile time and is under a target specified threshold. This expansion is implemented in CodeGenPrepare and expands into straight line code. The target specifies a maximum load size and the expansion works by using this size to load the two sources, compare, and exit early if a difference is found. It also has a special case when the memcmp result is used in a compare to zero equality. Differential Revision: https://reviews.llvm.org/D28637 llvm-svn: 304313
2017-06-01 01:12:38 +08:00
}
[PowerPC] Enable interleaved-access vectorization This adds a basic cost model for interleaved-access vectorization (and a better default for shuffles), and enables interleaved-access vectorization by default. The relevant difference from the default cost model for interleaved-access vectorization, is that on PPC, the shuffles that end up being used are *much* cheaper than modeling the process with insert/extract pairs (which are quite expensive, especially on older cores). llvm-svn: 246824
2015-09-04 08:10:41 +08:00
bool PPCTTIImpl::enableInterleavedAccessVectorization() {
return true;
}
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
unsigned PPCTTIImpl::getNumberOfRegisters(bool Vector) {
[PowerPC] Add support for the QPX vector instruction set This adds support for the QPX vector instruction set, which is used by the enhanced A2 cores on the IBM BG/Q supercomputers. QPX vectors are 256 bytes wide, holding 4 double-precision floating-point values. Boolean values, modeled here as <4 x i1> are actually also represented as floating-point values (essentially { -1, 1 } for { false, true }). QPX shares many features with Altivec and VSX, but is distinct from both of them. One major difference is that, instead of adding completely-separate vector registers, QPX vector registers are extensions of the scalar floating-point registers (lane 0 is the corresponding scalar floating-point value). The operations supported on QPX vectors mirrors that supported on the scalar floating-point values (with some additional ones for permutations and logical/comparison operations). I've been maintaining this support out-of-tree, as part of the bgclang project, for several years. This is not the entire bgclang patch set, but is most of the subset that can be cleanly integrated into LLVM proper at this time. Adding this to the LLVM backend is part of my efforts to rebase bgclang to the current LLVM trunk, but is independently useful (especially for codes that use LLVM as a JIT in library form). The assembler/disassembler test coverage is complete. The CodeGen test coverage is not, but I've included some tests, and more will be added as follow-up work. llvm-svn: 230413
2015-02-25 09:06:45 +08:00
if (Vector && !ST->hasAltivec() && !ST->hasQPX())
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
return 0;
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
return ST->hasVSX() ? 64 : 32;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
Const correctness for TTI::getRegisterBitWidth Summary: The method TargetTransformInfo::getRegisterBitWidth() is declared const, but the type erasing implementation classes (TargetTransformInfo::Concept & TargetTransformInfo::Model) that were introduced by Chandler in https://reviews.llvm.org/D7293 do not have the method declared const. This is an NFC to tidy up the const consistency between TTI and its implementation. Reviewers: chandlerc, rnk, reames Reviewed By: reames Subscribers: reames, jfb, arsenm, dschuff, nemanjai, nhaehnle, javed.absar, sbc100, jgravelle-google, llvm-commits Differential Revision: https://reviews.llvm.org/D33903 llvm-svn: 305189
2017-06-12 22:22:21 +08:00
unsigned PPCTTIImpl::getRegisterBitWidth(bool Vector) const {
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
if (Vector) {
[PowerPC] Add support for the QPX vector instruction set This adds support for the QPX vector instruction set, which is used by the enhanced A2 cores on the IBM BG/Q supercomputers. QPX vectors are 256 bytes wide, holding 4 double-precision floating-point values. Boolean values, modeled here as <4 x i1> are actually also represented as floating-point values (essentially { -1, 1 } for { false, true }). QPX shares many features with Altivec and VSX, but is distinct from both of them. One major difference is that, instead of adding completely-separate vector registers, QPX vector registers are extensions of the scalar floating-point registers (lane 0 is the corresponding scalar floating-point value). The operations supported on QPX vectors mirrors that supported on the scalar floating-point values (with some additional ones for permutations and logical/comparison operations). I've been maintaining this support out-of-tree, as part of the bgclang project, for several years. This is not the entire bgclang patch set, but is most of the subset that can be cleanly integrated into LLVM proper at this time. Adding this to the LLVM backend is part of my efforts to rebase bgclang to the current LLVM trunk, but is independently useful (especially for codes that use LLVM as a JIT in library form). The assembler/disassembler test coverage is complete. The CodeGen test coverage is not, but I've included some tests, and more will be added as follow-up work. llvm-svn: 230413
2015-02-25 09:06:45 +08:00
if (ST->hasQPX()) return 256;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
if (ST->hasAltivec()) return 128;
return 0;
}
if (ST->isPPC64())
return 64;
return 32;
}
Revert "[System Model] [TTI] Update cache and prefetch TTI interfaces" This broke some PPC prefetching tests. This reverts commit 9fdfb045ae8bb643ab0d0455dcf9ecaea3b1eb3c. llvm-svn: 365680
2019-07-11 02:25:58 +08:00
unsigned PPCTTIImpl::getCacheLineSize() {
[PowerPC] Correctly specify the cache line size for Power 7, 8 and 9. Fixes PPCTTIImpl::getCacheLineSize() returning the wrong cache line size for newer ppc processors. Commiting on behalf of Stefan Pintilie. Differential Revision: https://reviews.llvm.org/D33656 llvm-svn: 304317
2017-06-01 02:20:17 +08:00
// Check first if the user specified a custom line size.
if (CacheLineSize.getNumOccurrences() > 0)
return CacheLineSize;
// On P7, P8 or P9 we have a cache line size of 128.
unsigned Directive = ST->getDarwinDirective();
if (Directive == PPC::DIR_PWR7 || Directive == PPC::DIR_PWR8 ||
Directive == PPC::DIR_PWR9)
return 128;
// On other processors return a default of 64 bytes.
return 64;
[TTI] Add getCacheLineSize Summary: And use it in PPCLoopDataPrefetch.cpp. @hfinkel, please let me know if your preference would be to preserve the ppc-loop-prefetch-cache-line option in order to be able to override the value of TTI::getCacheLineSize for PPC. Reviewers: hfinkel Subscribers: hulx2000, mcrosier, mssimpso, hfinkel, llvm-commits Differential Revision: http://reviews.llvm.org/D16306 llvm-svn: 258419
2016-01-22 02:28:36 +08:00
}
Revert "[System Model] [TTI] Update cache and prefetch TTI interfaces" This broke some PPC prefetching tests. This reverts commit 9fdfb045ae8bb643ab0d0455dcf9ecaea3b1eb3c. llvm-svn: 365680
2019-07-11 02:25:58 +08:00
unsigned PPCTTIImpl::getPrefetchDistance() {
[PPC] Remove -ppc-loop-prefetch-distance in favor of -prefetch-distance After the previous change, this can now be overridden centrally in the pass. llvm-svn: 264807
2016-03-30 07:45:56 +08:00
// This seems like a reasonable default for the BG/Q (this pass is enabled, by
// default, only on the BG/Q).
return 300;
}
[TTI] Add getPrefetchDistance from PPCLoopDataPrefetch, NFC This patch is part of the work to make PPCLoopDataPrefetch target-independent (http://thread.gmane.org/gmane.comp.compilers.llvm.devel/92758). As it was discussed in the above thread, getPrefetchDistance is currently using instruction count which may change in the future. llvm-svn: 258995
2016-01-28 06:21:25 +08:00
[X86] Disable loop unrolling in loop vectorization pass when VF is 1. The patch disabled unrolling in loop vectorization pass when VF==1 on x86 architecture, by setting MaxInterleaveFactor to 1. Unrolling in loop vectorization pass may introduce the cost of overflow check, memory boundary check and extra prologue/epilogue code when regular unroller will unroll the loop another time. Disable it when VF==1 remove the unnecessary cost on x86. The same can be done for other platforms after verifying interleaving/memory bound checking to be not perf critical on those platforms. Differential Revision: http://reviews.llvm.org/D9515 llvm-svn: 236613
2015-05-07 01:12:25 +08:00
unsigned PPCTTIImpl::getMaxInterleaveFactor(unsigned VF) {
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
unsigned Directive = ST->getDarwinDirective();
// The 440 has no SIMD support, but floating-point instructions
// have a 5-cycle latency, so unroll by 5x for latency hiding.
if (Directive == PPC::DIR_440)
return 5;
// The A2 has no SIMD support, but floating-point instructions
// have a 6-cycle latency, so unroll by 6x for latency hiding.
if (Directive == PPC::DIR_A2)
return 6;
// FIXME: For lack of any better information, do no harm...
if (Directive == PPC::DIR_E500mc || Directive == PPC::DIR_E5500)
return 1;
Change max interleave factor to 12 for POWER7 and POWER8. llvm-svn: 228973
2015-02-13 06:57:58 +08:00
// For P7 and P8, floating-point instructions have a 6-cycle latency and
// there are two execution units, so unroll by 12x for latency hiding.
[Power9] Add support for -mcpu=pwr9 in the back end This patch corresponds to review: http://reviews.llvm.org/D19683 Simply adds the bits for being able to specify -mcpu=pwr9 to the back end. llvm-svn: 268950
2016-05-10 02:54:58 +08:00
// FIXME: the same for P9 as previous gen until POWER9 scheduling is ready
if (Directive == PPC::DIR_PWR7 || Directive == PPC::DIR_PWR8 ||
Directive == PPC::DIR_PWR9)
Change max interleave factor to 12 for POWER7 and POWER8. llvm-svn: 228973
2015-02-13 06:57:58 +08:00
return 12;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
// For most things, modern systems have two execution units (and
// out-of-order execution).
return 2;
}
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
// Adjust the cost of vector instructions on targets which there is overlap
// between the vector and scalar units, thereby reducing the overall throughput
// of vector code wrt. scalar code.
int PPCTTIImpl::vectorCostAdjustment(int Cost, unsigned Opcode, Type *Ty1,
Type *Ty2) {
if (!ST->vectorsUseTwoUnits() || !Ty1->isVectorTy())
return Cost;
std::pair<int, MVT> LT1 = TLI->getTypeLegalizationCost(DL, Ty1);
// If type legalization involves splitting the vector, we don't want to
// double the cost at every step - only the last step.
if (LT1.first != 1 || !LT1.second.isVector())
return Cost;
test commit (add blank line) NFC llvm-svn: 352897
2019-02-02 02:55:43 +08:00
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
int ISD = TLI->InstructionOpcodeToISD(Opcode);
if (TLI->isOperationExpand(ISD, LT1.second))
return Cost;
if (Ty2) {
std::pair<int, MVT> LT2 = TLI->getTypeLegalizationCost(DL, Ty2);
if (LT2.first != 1 || !LT2.second.isVector())
return Cost;
}
return Cost * 2;
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getArithmeticInstrCost(
[PM] Change the core design of the TTI analysis to use a polymorphic type erased interface and a single analysis pass rather than an extremely complex analysis group. The end result is that the TTI analysis can contain a type erased implementation that supports the polymorphic TTI interface. We can build one from a target-specific implementation or from a dummy one in the IR. I've also factored all of the code into "mix-in"-able base classes, including CRTP base classes to facilitate calling back up to the most specialized form when delegating horizontally across the surface. These aren't as clean as I would like and I'm planning to work on cleaning some of this up, but I wanted to start by putting into the right form. There are a number of reasons for this change, and this particular design. The first and foremost reason is that an analysis group is complete overkill, and the chaining delegation strategy was so opaque, confusing, and high overhead that TTI was suffering greatly for it. Several of the TTI functions had failed to be implemented in all places because of the chaining-based delegation making there be no checking of this. A few other functions were implemented with incorrect delegation. The message to me was very clear working on this -- the delegation and analysis group structure was too confusing to be useful here. The other reason of course is that this is *much* more natural fit for the new pass manager. This will lay the ground work for a type-erased per-function info object that can look up the correct subtarget and even cache it. Yet another benefit is that this will significantly simplify the interaction of the pass managers and the TargetMachine. See the future work below. The downside of this change is that it is very, very verbose. I'm going to work to improve that, but it is somewhat an implementation necessity in C++ to do type erasure. =/ I discussed this design really extensively with Eric and Hal prior to going down this path, and afterward showed them the result. No one was really thrilled with it, but there doesn't seem to be a substantially better alternative. Using a base class and virtual method dispatch would make the code much shorter, but as discussed in the update to the programmer's manual and elsewhere, a polymorphic interface feels like the more principled approach even if this is perhaps the least compelling example of it. ;] Ultimately, there is still a lot more to be done here, but this was the huge chunk that I couldn't really split things out of because this was the interface change to TTI. I've tried to minimize all the other parts of this. The follow up work should include at least: 1) Improving the TargetMachine interface by having it directly return a TTI object. Because we have a non-pass object with value semantics and an internal type erasure mechanism, we can narrow the interface of the TargetMachine to *just* do what we need: build and return a TTI object that we can then insert into the pass pipeline. 2) Make the TTI object be fully specialized for a particular function. This will include splitting off a minimal form of it which is sufficient for the inliner and the old pass manager. 3) Add a new pass manager analysis which produces TTI objects from the target machine for each function. This may actually be done as part of #2 in order to use the new analysis to implement #2. 4) Work on narrowing the API between TTI and the targets so that it is easier to understand and less verbose to type erase. 5) Work on narrowing the API between TTI and its clients so that it is easier to understand and less verbose to forward. 6) Try to improve the CRTP-based delegation. I feel like this code is just a bit messy and exacerbating the complexity of implementing the TTI in each target. Many thanks to Eric and Hal for their help here. I ended up blocked on this somewhat more abruptly than I expected, and so I appreciate getting it sorted out very quickly. Differential Revision: http://reviews.llvm.org/D7293 llvm-svn: 227669
2015-01-31 11:43:40 +08:00
unsigned Opcode, Type *Ty, TTI::OperandValueKind Op1Info,
TTI::OperandValueKind Op2Info, TTI::OperandValueProperties Opd1PropInfo,
[X86] updating TTI costs for arithmetic instructions on X86\SLM arch. updated instructions: pmulld, pmullw, pmulhw, mulsd, mulps, mulpd, divss, divps, divsd, divpd, addpd and subpd. special optimization case which replaces pmulld with pmullw\pmulhw\pshuf seq. In case if the real operands bitwidth <= 16. Differential Revision: https://reviews.llvm.org/D28104 llvm-svn: 291657
2017-01-11 16:23:37 +08:00
TTI::OperandValueProperties Opd2PropInfo, ArrayRef<const Value *> Args) {
Remove unused variables, silences -Wunused-variable llvm-svn: 173526
2013-01-26 07:17:21 +08:00
assert(TLI->InstructionOpcodeToISD(Opcode) && "Invalid opcode");
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
// Fallback to the default implementation.
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
int Cost = BaseT::getArithmeticInstrCost(Opcode, Ty, Op1Info, Op2Info,
Opd1PropInfo, Opd2PropInfo);
return vectorCostAdjustment(Cost, Opcode, Ty, nullptr);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *Tp, int Index,
Type *SubTp) {
[PowerPC] Enable interleaved-access vectorization This adds a basic cost model for interleaved-access vectorization (and a better default for shuffles), and enables interleaved-access vectorization by default. The relevant difference from the default cost model for interleaved-access vectorization, is that on PPC, the shuffles that end up being used are *much* cheaper than modeling the process with insert/extract pairs (which are quite expensive, especially on older cores). llvm-svn: 246824
2015-09-04 08:10:41 +08:00
// Legalize the type.
std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, Tp);
// PPC, for both Altivec/VSX and QPX, support cheap arbitrary permutations
// (at least in the sense that there need only be one non-loop-invariant
// instruction). We need one such shuffle instruction for each actual
// register (this is not true for arbitrary shuffles, but is true for the
// structured types of shuffles covered by TTI::ShuffleKind).
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
return vectorCostAdjustment(LT.first, Instruction::ShuffleVector, Tp,
nullptr);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[SystemZ] TargetTransformInfo cost functions implemented. getArithmeticInstrCost(), getShuffleCost(), getCastInstrCost(), getCmpSelInstrCost(), getVectorInstrCost(), getMemoryOpCost(), getInterleavedMemoryOpCost() implemented. Interleaved access vectorization enabled. BasicTTIImpl::getCastInstrCost() improved to check for legal extending loads, in which case the cost of the z/sext instruction becomes 0. Review: Ulrich Weigand, Renato Golin. https://reviews.llvm.org/D29631 llvm-svn: 300052
2017-04-12 19:49:08 +08:00
int PPCTTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src,
const Instruction *I) {
Remove unused variables, silences -Wunused-variable llvm-svn: 173526
2013-01-26 07:17:21 +08:00
assert(TLI->InstructionOpcodeToISD(Opcode) && "Invalid opcode");
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
int Cost = BaseT::getCastInstrCost(Opcode, Dst, Src);
return vectorCostAdjustment(Cost, Opcode, Dst, Src);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[SystemZ] TargetTransformInfo cost functions implemented. getArithmeticInstrCost(), getShuffleCost(), getCastInstrCost(), getCmpSelInstrCost(), getVectorInstrCost(), getMemoryOpCost(), getInterleavedMemoryOpCost() implemented. Interleaved access vectorization enabled. BasicTTIImpl::getCastInstrCost() improved to check for legal extending loads, in which case the cost of the z/sext instruction becomes 0. Review: Ulrich Weigand, Renato Golin. https://reviews.llvm.org/D29631 llvm-svn: 300052
2017-04-12 19:49:08 +08:00
int PPCTTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy, Type *CondTy,
const Instruction *I) {
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
int Cost = BaseT::getCmpSelInstrCost(Opcode, ValTy, CondTy, I);
return vectorCostAdjustment(Cost, Opcode, ValTy, nullptr);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getVectorInstrCost(unsigned Opcode, Type *Val, unsigned Index) {
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
assert(Val->isVectorTy() && "This must be a vector type");
Refine fix to bug 15041. Thanks to help from Nadav and Hal, I have a more reasonable (and even correct!) approach. This specifically penalizes the insertelement and extractelement operations for the performance hit that will occur on PowerPC processors. llvm-svn: 174725
2013-02-09 02:19:17 +08:00
int ISD = TLI->InstructionOpcodeToISD(Opcode);
assert(ISD && "Invalid opcode");
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
int Cost = BaseT::getVectorInstrCost(Opcode, Val, Index);
Cost = vectorCostAdjustment(Cost, Opcode, Val, nullptr);
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
if (ST->hasVSX() && Val->getScalarType()->isDoubleTy()) {
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
// Double-precision scalars are already located in index #0 (or #1 if LE).
Fix operator precedence warning. NFCI. Reported in https://www.viva64.com/en/b/0629/ llvm-svn: 359469
2019-04-30 01:04:14 +08:00
if (ISD == ISD::EXTRACT_VECTOR_ELT &&
Index == (ST->isLittleEndian() ? 1 : 0))
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
return 0;
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
return Cost;
[PowerPC] Add support for the QPX vector instruction set This adds support for the QPX vector instruction set, which is used by the enhanced A2 cores on the IBM BG/Q supercomputers. QPX vectors are 256 bytes wide, holding 4 double-precision floating-point values. Boolean values, modeled here as <4 x i1> are actually also represented as floating-point values (essentially { -1, 1 } for { false, true }). QPX shares many features with Altivec and VSX, but is distinct from both of them. One major difference is that, instead of adding completely-separate vector registers, QPX vector registers are extensions of the scalar floating-point registers (lane 0 is the corresponding scalar floating-point value). The operations supported on QPX vectors mirrors that supported on the scalar floating-point values (with some additional ones for permutations and logical/comparison operations). I've been maintaining this support out-of-tree, as part of the bgclang project, for several years. This is not the entire bgclang patch set, but is most of the subset that can be cleanly integrated into LLVM proper at this time. Adding this to the LLVM backend is part of my efforts to rebase bgclang to the current LLVM trunk, but is independently useful (especially for codes that use LLVM as a JIT in library form). The assembler/disassembler test coverage is complete. The CodeGen test coverage is not, but I've included some tests, and more will be added as follow-up work. llvm-svn: 230413
2015-02-25 09:06:45 +08:00
} else if (ST->hasQPX() && Val->getScalarType()->isFloatingPointTy()) {
// Floating point scalars are already located in index #0.
if (Index == 0)
return 0;
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
return Cost;
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
}
Refine fix to bug 15041. Thanks to help from Nadav and Hal, I have a more reasonable (and even correct!) approach. This specifically penalizes the insertelement and extractelement operations for the performance hit that will occur on PowerPC processors. llvm-svn: 174725
2013-02-09 02:19:17 +08:00
// Estimated cost of a load-hit-store delay. This was obtained
// experimentally as a minimum needed to prevent unprofitable
// vectorization for the paq8p benchmark. It may need to be
// raised further if other unprofitable cases remain.
[PowerPC] Adjust load/store costs in PPCTTI This provides more realistic costs for the insert/extractelement instructions (which are load/store pairs), accounts for the cheap unaligned Altivec load sequence, and for unaligned VSX load/stores. Bad news: MultiSource/Applications/sgefa/sgefa - 35% slowdown (this will require more investigation) SingleSource/Benchmarks/McGill/queens - 20% slowdown (we no longer vectorize this, but it was a constant store that was scalarized) MultiSource/Benchmarks/FreeBench/pcompress2/pcompress2 - 2% slowdown Good news: SingleSource/Benchmarks/Shootout/ary3 - 54% speedup SingleSource/Benchmarks/Shootout-C++/ary - 40% speedup MultiSource/Benchmarks/Ptrdist/ks/ks - 35% speedup MultiSource/Benchmarks/FreeBench/neural/neural - 30% speedup MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt - 20% speedup Unfortunately, estimating the costs of the stack-based scalarization sequences is hard, and adjusting these costs is like a game of whac-a-mole :( I'll revisit this again after we have better codegen for vector extloads and truncstores and unaligned load/stores. llvm-svn: 205658
2014-04-05 07:51:18 +08:00
unsigned LHSPenalty = 2;
if (ISD == ISD::INSERT_VECTOR_ELT)
LHSPenalty += 7;
Refine fix to bug 15041. Thanks to help from Nadav and Hal, I have a more reasonable (and even correct!) approach. This specifically penalizes the insertelement and extractelement operations for the performance hit that will occur on PowerPC processors. llvm-svn: 174725
2013-02-09 02:19:17 +08:00
// Vector element insert/extract with Altivec is very expensive,
// because they require store and reload with the attendant
// processor stall for load-hit-store. Until VSX is available,
// these need to be estimated as very costly.
if (ISD == ISD::EXTRACT_VECTOR_ELT ||
ISD == ISD::INSERT_VECTOR_ELT)
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
return LHSPenalty + Cost;
Constrain PowerPC autovectorization to fix bug 15041. Certain vector operations don't vectorize well with the current PowerPC implementation. Element insert/extract performs poorly without VSX support because Altivec requires going through memory. SREM, UREM, and VSELECT all produce bad scalar code. There's a lot of work to do for the cost model before autovectorization will be tuned well, and this is not an attempt to address the larger problem. llvm-svn: 174660
2013-02-08 04:33:57 +08:00
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
return Cost;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
}
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int PPCTTIImpl::getMemoryOpCost(unsigned Opcode, Type *Src, unsigned Alignment,
[SystemZ] TargetTransformInfo cost functions implemented. getArithmeticInstrCost(), getShuffleCost(), getCastInstrCost(), getCmpSelInstrCost(), getVectorInstrCost(), getMemoryOpCost(), getInterleavedMemoryOpCost() implemented. Interleaved access vectorization enabled. BasicTTIImpl::getCastInstrCost() improved to check for legal extending loads, in which case the cost of the z/sext instruction becomes 0. Review: Ulrich Weigand, Renato Golin. https://reviews.llvm.org/D29631 llvm-svn: 300052
2017-04-12 19:49:08 +08:00
unsigned AddressSpace, const Instruction *I) {
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
// Legalize the type.
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, Src);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
assert((Opcode == Instruction::Load || Opcode == Instruction::Store) &&
"Invalid Opcode");
[TTI] Make the cost APIs in TargetTransformInfo consistently use 'int' rather than 'unsigned' for their costs. For something like costs in particular there is a natural "negative" value, that of savings or saved cost. As a consequence, there is a lot of code that subtracts or creates negative values based on cost, all of which is prone to awkwardness or bugs when dealing with an unsigned type. Similarly, we *never* want these values to wrap, as that would cause Very Bad code generation (likely percieved as an infinite loop as we try to emit over 2^32 instructions or some such insanity). All around 'int' seems a much better fit for these basic metrics. I've added asserts to ensure that at least the TTI interface never returns negative numbers here. If we ever have a use case for negative numbers, we can remove this, but this way a bug where someone used '-1' to produce a 'very large' cost will be caught by the assert. This passes all tests, and is also UBSan clean. No functional change intended. Differential Revision: http://reviews.llvm.org/D11741 llvm-svn: 244080
2015-08-06 02:08:10 +08:00
int Cost = BaseT::getMemoryOpCost(Opcode, Src, Alignment, AddressSpace);
[PowerPC] Update Vector Costs for P9 For the power9 CPU, vector operations consume a pair of execution units rather than one execution unit like a scalar operation. Update the target transform cost functions to reflect the higher cost of vector operations when targeting Power9. Patch by RolandF. Differential revision: https://reviews.llvm.org/D55461 llvm-svn: 352261
2019-01-26 09:18:48 +08:00
Cost = vectorCostAdjustment(Cost, Opcode, Src, nullptr);
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
[PowerPC] Cleanup cost model for unaligned vector loads/stores I'm adding a regression test to better cover code generation for unaligned vector loads and stores, but there's no functional change to the code generation here. There is an improvement to the cost model for unaligned vector loads and stores, mostly for QPX (for which we were not previously accounting for the permutation-based loads), and the cost model implementation is cleaner. llvm-svn: 246712
2015-09-03 05:03:28 +08:00
bool IsAltivecType = ST->hasAltivec() &&
(LT.second == MVT::v16i8 || LT.second == MVT::v8i16 ||
LT.second == MVT::v4i32 || LT.second == MVT::v4f32);
bool IsVSXType = ST->hasVSX() &&
(LT.second == MVT::v2f64 || LT.second == MVT::v2i64);
bool IsQPXType = ST->hasQPX() &&
(LT.second == MVT::v4f64 || LT.second == MVT::v4f32);
[ppc] Correctly compute the cost of loading 32/64 bit memory into VSR VSX has instructions lxsiwax/lxsdx that can load 32/64 bit value into VSX register cheaply. That patch makes it known to memory cost model, so the vectorization of the test case in pr30990 is beneficial. Differential Revision: https://reviews.llvm.org/D26713 llvm-svn: 288560
2016-12-03 08:41:43 +08:00
// VSX has 32b/64b load instructions. Legalization can handle loading of
// 32b/64b to VSR correctly and cheaply. But BaseT::getMemoryOpCost and
// PPCTargetLowering can't compute the cost appropriately. So here we
// explicitly check this case.
unsigned MemBytes = Src->getPrimitiveSizeInBits();
if (Opcode == Instruction::Load && ST->hasVSX() && IsAltivecType &&
(MemBytes == 64 || (ST->hasP8Vector() && MemBytes == 32)))
return 1;
// Aligned loads and stores are easy.
unsigned SrcBytes = LT.second.getStoreSize();
if (!SrcBytes || !Alignment || Alignment >= SrcBytes)
return Cost;
[PowerPC] Include the permutation cost for unaligned vector loads Pre-P8, when we generate code for unaligned vector loads (for Altivec and QPX types), even when accounting for the combining that takes place for multiple consecutive such loads, there is at least one load instructions and one permutation for each load. Make sure the cost reported reflects the cost of the permutes as well. llvm-svn: 246807
2015-09-04 05:23:18 +08:00
// If we can use the permutation-based load sequence, then this is also
// relatively cheap (not counting loop-invariant instructions): one load plus
// one permute (the last load in a series has extra cost, but we're
[PowerPC] Clarify a comment in PPCTTI about vector loads This should say that we could do unaligned vector loads on the P7 using VSX instructions, not that we should. llvm-svn: 264683
2016-03-29 06:39:35 +08:00
// neglecting that here). Note that on the P7, we could do unaligned loads
[PowerPC] Include the permutation cost for unaligned vector loads Pre-P8, when we generate code for unaligned vector loads (for Altivec and QPX types), even when accounting for the combining that takes place for multiple consecutive such loads, there is at least one load instructions and one permutation for each load. Make sure the cost reported reflects the cost of the permutes as well. llvm-svn: 246807
2015-09-04 05:23:18 +08:00
// for Altivec types using the VSX instructions, but that's more expensive
// than using the permutation-based load sequence. On the P8, that's no
// longer true.
if (Opcode == Instruction::Load &&
((!ST->hasP8Vector() && IsAltivecType) || IsQPXType) &&
Alignment >= LT.second.getScalarType().getStoreSize())
return Cost + LT.first; // Add the cost of the permutations.
[PowerPC] Cleanup cost model for unaligned vector loads/stores I'm adding a regression test to better cover code generation for unaligned vector loads and stores, but there's no functional change to the code generation here. There is an improvement to the cost model for unaligned vector loads and stores, mostly for QPX (for which we were not previously accounting for the permutation-based loads), and the cost model implementation is cleaner. llvm-svn: 246712
2015-09-03 05:03:28 +08:00
// For VSX, we can do unaligned loads and stores on Altivec/VSX types. On the
// P7, unaligned vector loads are more expensive than the permutation-based
// load sequence, so that might be used instead, but regardless, the net cost
// is about the same (not counting loop-invariant instructions).
if (IsVSXType || (ST->hasVSX() && IsAltivecType))
return Cost;
[PPC] Give unaligned memory access lower cost on processor that supports it Newer ppc supports unaligned memory access, it reduces the cost of unaligned memory access significantly. This patch handles this case in PPCTTIImpl::getMemoryOpCost. This patch fixes pr31492. Differential Revision: https://reviews.llvm.org/D28630 This is resubmit of r292680, which was reverted by r293092. The internal application failures were actually caused by a source code bug. llvm-svn: 295506
2017-02-18 06:29:39 +08:00
// Newer PPC supports unaligned memory access.
if (TLI->allowsMisalignedMemoryAccesses(LT.second, 0))
return Cost;
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
// PPC in general does not support unaligned loads and stores. They'll need
// to be decomposed based on the alignment factor.
[PowerPC] Cleanup cost model for unaligned vector loads/stores I'm adding a regression test to better cover code generation for unaligned vector loads and stores, but there's no functional change to the code generation here. There is an improvement to the cost model for unaligned vector loads and stores, mostly for QPX (for which we were not previously accounting for the permutation-based loads), and the cost model implementation is cleaner. llvm-svn: 246712
2015-09-03 05:03:28 +08:00
// Add the cost of each scalar load or store.
Cost += LT.first*(SrcBytes/Alignment-1);
// For a vector type, there is also scalarization overhead (only for
// stores, loads are expanded using the vector-load + permutation sequence,
// which is much less expensive).
if (Src->isVectorTy() && Opcode == Instruction::Store)
for (int i = 0, e = Src->getVectorNumElements(); i < e; ++i)
Cost += getVectorInstrCost(Instruction::ExtractElement, Src, i);
[PowerPC] Adjust load/store costs in PPCTTI This provides more realistic costs for the insert/extractelement instructions (which are load/store pairs), accounts for the cheap unaligned Altivec load sequence, and for unaligned VSX load/stores. Bad news: MultiSource/Applications/sgefa/sgefa - 35% slowdown (this will require more investigation) SingleSource/Benchmarks/McGill/queens - 20% slowdown (we no longer vectorize this, but it was a constant store that was scalarized) MultiSource/Benchmarks/FreeBench/pcompress2/pcompress2 - 2% slowdown Good news: SingleSource/Benchmarks/Shootout/ary3 - 54% speedup SingleSource/Benchmarks/Shootout-C++/ary - 40% speedup MultiSource/Benchmarks/Ptrdist/ks/ks - 35% speedup MultiSource/Benchmarks/FreeBench/neural/neural - 30% speedup MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt - 20% speedup Unfortunately, estimating the costs of the stack-based scalarization sequences is hard, and adjusting these costs is like a game of whac-a-mole :( I'll revisit this again after we have better codegen for vector extloads and truncstores and unaligned load/stores. llvm-svn: 205658
2014-04-05 07:51:18 +08:00
Initial implementation of PPCTargetTransformInfo This provides a place to add customized operation cost information and control some other target-specific IR-level transformations. The only non-trivial logic in this checkin assigns a higher cost to unaligned loads and stores (covered by the included test case). llvm-svn: 173520
2013-01-26 07:05:59 +08:00
return Cost;
}
[PowerPC] Enable interleaved-access vectorization This adds a basic cost model for interleaved-access vectorization (and a better default for shuffles), and enables interleaved-access vectorization by default. The relevant difference from the default cost model for interleaved-access vectorization, is that on PPC, the shuffles that end up being used are *much* cheaper than modeling the process with insert/extract pairs (which are quite expensive, especially on older cores). llvm-svn: 246824
2015-09-04 08:10:41 +08:00
int PPCTTIImpl::getInterleavedMemoryOpCost(unsigned Opcode, Type *VecTy,
unsigned Factor,
ArrayRef<unsigned> Indices,
unsigned Alignment,
recommit 344472 after fixing build failure on ARM and PPC. llvm-svn: 344475
2018-10-14 16:50:06 +08:00
unsigned AddressSpace,
[LV] Support vectorization of interleave-groups that require an epilog under optsize using masked wide loads Under Opt for Size, the vectorizer does not vectorize interleave-groups that have gaps at the end of the group (such as a loop that reads only the even elements: a[2*i]) because that implies that we'll require a scalar epilogue (which is not allowed under Opt for Size). This patch extends the support for masked-interleave-groups (introduced by D53011 for conditional accesses) to also cover the case of gaps in a group of loads; Targets that enable the masked-interleave-group feature don't have to invalidate interleave-groups of loads with gaps; they could now use masked wide-loads and shuffles (if that's what the cost model selects). Reviewers: Ayal, hsaito, dcaballe, fhahn Reviewed By: Ayal Differential Revision: https://reviews.llvm.org/D53668 llvm-svn: 345705
2018-10-31 17:57:56 +08:00
bool UseMaskForCond,
bool UseMaskForGaps) {
if (UseMaskForCond || UseMaskForGaps)
recommit 344472 after fixing build failure on ARM and PPC. llvm-svn: 344475
2018-10-14 16:50:06 +08:00
return BaseT::getInterleavedMemoryOpCost(Opcode, VecTy, Factor, Indices,
[LV] Support vectorization of interleave-groups that require an epilog under optsize using masked wide loads Under Opt for Size, the vectorizer does not vectorize interleave-groups that have gaps at the end of the group (such as a loop that reads only the even elements: a[2*i]) because that implies that we'll require a scalar epilogue (which is not allowed under Opt for Size). This patch extends the support for masked-interleave-groups (introduced by D53011 for conditional accesses) to also cover the case of gaps in a group of loads; Targets that enable the masked-interleave-group feature don't have to invalidate interleave-groups of loads with gaps; they could now use masked wide-loads and shuffles (if that's what the cost model selects). Reviewers: Ayal, hsaito, dcaballe, fhahn Reviewed By: Ayal Differential Revision: https://reviews.llvm.org/D53668 llvm-svn: 345705
2018-10-31 17:57:56 +08:00
Alignment, AddressSpace,
UseMaskForCond, UseMaskForGaps);
recommit 344472 after fixing build failure on ARM and PPC. llvm-svn: 344475
2018-10-14 16:50:06 +08:00
[PowerPC] Enable interleaved-access vectorization This adds a basic cost model for interleaved-access vectorization (and a better default for shuffles), and enables interleaved-access vectorization by default. The relevant difference from the default cost model for interleaved-access vectorization, is that on PPC, the shuffles that end up being used are *much* cheaper than modeling the process with insert/extract pairs (which are quite expensive, especially on older cores). llvm-svn: 246824
2015-09-04 08:10:41 +08:00
assert(isa<VectorType>(VecTy) &&
"Expect a vector type for interleaved memory op");
// Legalize the type.
std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, VecTy);
// Firstly, the cost of load/store operation.
int Cost = getMemoryOpCost(Opcode, VecTy, Alignment, AddressSpace);
// PPC, for both Altivec/VSX and QPX, support cheap arbitrary permutations
// (at least in the sense that there need only be one non-loop-invariant
// instruction). For each result vector, we need one shuffle per incoming
// vector (except that the first shuffle can take two incoming vectors
// because it does not need to take itself).
Cost += Factor*(LT.first-1);
return Cost;
}
[PowerPC] exclude ICmpZero in LSR if icmp can be replaced in later hardware loop. Differential Revision: https://reviews.llvm.org/D63477 llvm-svn: 364993
2019-07-03 09:49:03 +08:00
bool PPCTTIImpl::canSaveCmp(Loop *L, BranchInst **BI, ScalarEvolution *SE,
LoopInfo *LI, DominatorTree *DT,
AssumptionCache *AC, TargetLibraryInfo *LibInfo) {
// Process nested loops first.
for (Loop::iterator I = L->begin(), E = L->end(); I != E; ++I)
if (canSaveCmp(*I, BI, SE, LI, DT, AC, LibInfo))
return false; // Stop search.
HardwareLoopInfo HWLoopInfo(L);
if (!HWLoopInfo.canAnalyze(*LI))
return false;
if (!isHardwareLoopProfitable(L, *SE, *AC, LibInfo, HWLoopInfo))
return false;
if (!HWLoopInfo.isHardwareLoopCandidate(*SE, *LI, *DT))
return false;
*BI = HWLoopInfo.ExitBranch;
return true;
}