llvm-project/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll

179 lines
6.2 KiB
LLVM
Raw Normal View History

; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -loop-vectorize -verify-loop-info -simplifycfg < %s | FileCheck %s --check-prefix=UNROLL
; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -loop-vectorize -verify-loop-info < %s | FileCheck %s --check-prefix=UNROLL-NOSIMPLIFY
; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=2 -force-vector-interleave=1 -loop-vectorize -verify-loop-info -simplifycfg < %s | FileCheck %s --check-prefix=VEC
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
; Test predication of stores.
define i32 @test(i32* nocapture %f) #0 {
entry:
br label %for.body
; VEC-LABEL: test
[LV] Unify vector and scalar maps This patch unifies the data structures we use for mapping instructions from the original loop to their corresponding instructions in the new loop. Previously, we maintained two distinct maps for this purpose: WidenMap and ScalarIVMap. WidenMap maintained the vector values each instruction from the old loop was represented with, and ScalarIVMap maintained the scalar values each scalarized induction variable was represented with. With this patch, all values created for the new loop are maintained in VectorLoopValueMap. The change allows for several simplifications. Previously, when an instruction was scalarized, we had to insert the scalar values into vectors in order to maintain the mapping in WidenMap. Then, if a user of the scalarized value was also scalar, we had to extract the scalar values from the temporary vector we created. We now aovid these unnecessary scalar-to-vector-to-scalar conversions. If a scalarized value is used by a scalar instruction, the scalar value is used directly. However, if the scalarized value is needed by a vector instruction, we generate the needed insertelement instructions on-demand. A common idiom in several locations in the code (including the scalarization code), is to first get the vector values an instruction from the original loop maps to, and then extract a particular scalar value. This patch adds getScalarValue for this purpose along side getVectorValue as an interface into VectorLoopValueMap. These functions work together to return the requested values if they're available or to produce them if they're not. The mapping has also be made less permissive. Entries can be added to VectorLoopValue map with the new initVector and initScalar functions. getVectorValue has been modified to return a constant reference to the mapped entries. There's no real functional change with this patch; however, in some cases we will generate slightly different code. For example, instead of an insertelement sequence following the definition of an instruction, it will now precede the first use of that instruction. This can be seen in the test case changes. Differential Revision: https://reviews.llvm.org/D23169 llvm-svn: 279649
2016-08-25 02:23:17 +08:00
; VEC: %[[v0:.+]] = add i64 %index, 0
; VEC: %[[v2:.+]] = getelementptr inbounds i32, i32* %f, i64 %[[v0]]
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; VEC: %[[v8:.+]] = icmp sgt <2 x i32> %{{.*}}, <i32 100, i32 100>
; VEC: %[[v11:.+]] = extractelement <2 x i1> %[[v8]], i32 0
; VEC: br i1 %[[v11]], label %[[cond:.+]], label %[[else:.+]]
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
;
; VEC: [[cond]]:
; VEC: %[[v13:.+]] = extractelement <2 x i32> %wide.load, i32 0
; VEC: %[[v9a:.+]] = add nsw i32 %[[v13]], 20
; VEC: store i32 %[[v9a]], i32* %[[v2]], align 4
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; VEC: br label %[[else:.+]]
;
; VEC: [[else]]:
; VEC: %[[v15:.+]] = extractelement <2 x i1> %[[v8]], i32 1
; VEC: br i1 %[[v15]], label %[[cond2:.+]], label %[[else2:.+]]
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
;
; VEC: [[cond2]]:
; VEC: %[[v17:.+]] = extractelement <2 x i32> %wide.load, i32 1
; VEC: %[[v9b:.+]] = add nsw i32 %[[v17]], 20
; VEC: %[[v1:.+]] = add i64 %index, 1
; VEC: %[[v4:.+]] = getelementptr inbounds i32, i32* %f, i64 %[[v1]]
; VEC: store i32 %[[v9b]], i32* %[[v4]], align 4
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; VEC: br label %[[else2:.+]]
;
; VEC: [[else2]]:
; UNROLL-LABEL: test
; UNROLL: vector.body:
; UNROLL: %[[IND:[a-zA-Z0-9]+]] = add i64 %{{.*}}, 0
; UNROLL: %[[IND1:[a-zA-Z0-9]+]] = add i64 %{{.*}}, 1
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
; UNROLL: %[[v0:[a-zA-Z0-9]+]] = getelementptr inbounds i32, i32* %f, i64 %[[IND]]
; UNROLL: %[[v1:[a-zA-Z0-9]+]] = getelementptr inbounds i32, i32* %f, i64 %[[IND1]]
; UNROLL: %[[v2:[a-zA-Z0-9]+]] = load i32, i32* %[[v0]], align 4
; UNROLL: %[[v3:[a-zA-Z0-9]+]] = load i32, i32* %[[v1]], align 4
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; UNROLL: %[[v4:[a-zA-Z0-9]+]] = icmp sgt i32 %[[v2]], 100
; UNROLL: %[[v5:[a-zA-Z0-9]+]] = icmp sgt i32 %[[v3]], 100
; UNROLL: br i1 %[[v4]], label %[[cond:[a-zA-Z0-9.]+]], label %[[else:[a-zA-Z0-9.]+]]
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
;
; UNROLL: [[cond]]:
; UNROLL: %[[v6:[a-zA-Z0-9]+]] = add nsw i32 %[[v2]], 20
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; UNROLL: store i32 %[[v6]], i32* %[[v0]], align 4
; UNROLL: br label %[[else]]
;
; UNROLL: [[else]]:
; UNROLL: br i1 %[[v5]], label %[[cond2:[a-zA-Z0-9.]+]], label %[[else2:[a-zA-Z0-9.]+]]
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
;
; UNROLL: [[cond2]]:
; UNROLL: %[[v7:[a-zA-Z0-9]+]] = add nsw i32 %[[v3]], 20
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
; UNROLL: store i32 %[[v7]], i32* %[[v1]], align 4
; UNROLL: br label %[[else2]]
;
; UNROLL: [[else2]]:
for.body:
%indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%arrayidx = getelementptr inbounds i32, i32* %f, i64 %indvars.iv
%0 = load i32, i32* %arrayidx, align 4
LoopVectorize: Support conditional stores by scalarizing The vectorizer takes a loop like this and widens all instructions except for the store. The stores are scalarized/unrolled and hidden behind an "if" block. for (i = 0; i < 128; ++i) { if (a[i] < 10) a[i] += val; } for (i = 0; i < 128; i+=2) { v = a[i:i+1]; v0 = (extract v, 0) + 10; v1 = (extract v, 1) + 10; if (v0 < 10) a[i] = v0; if (v1 < 10) a[i] = v1; } The vectorizer relies on subsequent optimizations to sink instructions into the conditional block where they are anticipated. The flag "vectorize-num-stores-pred" controls whether and how many stores to handle this way. Vectorization of conditional stores is disabled per default for now. This patch also adds a change to the heuristic when the flag "enable-loadstore-runtime-unroll" is enabled (off by default). It unrolls small loops until load/store ports are saturated. This heuristic uses TTI's getMaxUnrollFactor as a measure for load/store ports. I also added a second flag -enable-cond-stores-vec. It will enable vectorization of conditional stores. But there is no cost model for vectorization of conditional stores in place yet so this will not do good at the moment. rdar://15892953 Results for x86-64 -O3 -mavx +/- -mllvm -enable-loadstore-runtime-unroll -vectorize-num-stores-pred=1 (before the BFI change): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.35% (maze3() is identical but 10% slower) Applications/siod/siod 2.18% Performance improvements: mesa -4.42% libquantum -4.15% With a patch that slightly changes the register heuristics (by subtracting the induction variable on both sides of the register pressure equation, as the induction variable is probably not really unrolled): Performance Regressions: Benchmarks/Ptrdist/yacr2/yacr2 7.73% Applications/siod/siod 1.97% Performance Improvements: libquantum -13.05% (we now also unroll quantum_toffoli) mesa -4.27% llvm-svn: 200270
2014-01-28 09:01:53 +08:00
%cmp1 = icmp sgt i32 %0, 100
br i1 %cmp1, label %if.then, label %for.inc
if.then:
%add = add nsw i32 %0, 20
store i32 %add, i32* %arrayidx, align 4
br label %for.inc
for.inc:
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%exitcond = icmp eq i64 %indvars.iv.next, 128
br i1 %exitcond, label %for.end, label %for.body
for.end:
ret i32 0
}
; Track basic blocks when unrolling conditional blocks. This code used to assert
; because we did not update the phi nodes with the proper predecessor in the
; vectorized loop body.
; PR18724
; UNROLL-NOSIMPLIFY-LABEL: bug18724
; UNROLL-NOSIMPLIFY: store i32
; UNROLL-NOSIMPLIFY: store i32
define void @bug18724() {
entry:
br label %for.body9
for.body9:
br i1 undef, label %for.inc26, label %for.body14
for.body14:
%indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc23 ], [ undef, %for.body9 ]
%iNewChunks.120 = phi i32 [ %iNewChunks.2, %for.inc23 ], [ undef, %for.body9 ]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%arrayidx16 = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 %indvars.iv3
%tmp = load i32, i32* %arrayidx16, align 4
br i1 undef, label %if.then18, label %for.inc23
if.then18:
store i32 2, i32* %arrayidx16, align 4
%inc21 = add nsw i32 %iNewChunks.120, 1
br label %for.inc23
for.inc23:
%iNewChunks.2 = phi i32 [ %inc21, %if.then18 ], [ %iNewChunks.120, %for.body14 ]
%indvars.iv.next4 = add nsw i64 %indvars.iv3, 1
%tmp1 = trunc i64 %indvars.iv3 to i32
%cmp13 = icmp slt i32 %tmp1, 0
br i1 %cmp13, label %for.body14, label %for.inc26
for.inc26:
%iNewChunks.1.lcssa = phi i32 [ undef, %for.body9 ], [ %iNewChunks.2, %for.inc23 ]
unreachable
}
; VEC-LABEL: @minimal_bit_widths(
;
; In the test below, it's more profitable for the expression feeding the
; conditional store to remain scalar. Since we can only type-shrink vector
; types, we shouldn't try to represent the expression in a smaller type.
;
; VEC: vector.body:
; VEC: %wide.load = load <2 x i8>, <2 x i8>* {{.*}}, align 1
; VEC: br i1 {{.*}}, label %[[IF0:.+]], label %[[CONT0:.+]]
; VEC: [[IF0]]:
; VEC: %[[E0:.+]] = extractelement <2 x i8> %wide.load, i32 0
; VEC: %[[Z0:.+]] = zext i8 %[[E0]] to i32
; VEC: %[[T0:.+]] = trunc i32 %[[Z0]] to i8
; VEC: store i8 %[[T0]], i8* {{.*}}, align 1
; VEC: br label %[[CONT0]]
; VEC: [[CONT0]]:
; VEC: br i1 {{.*}}, label %[[IF1:.+]], label %[[CONT1:.+]]
; VEC: [[IF1]]:
; VEC: %[[E1:.+]] = extractelement <2 x i8> %wide.load, i32 1
; VEC: %[[Z1:.+]] = zext i8 %[[E1]] to i32
; VEC: %[[T1:.+]] = trunc i32 %[[Z1]] to i8
; VEC: store i8 %[[T1]], i8* {{.*}}, align 1
; VEC: br label %[[CONT1]]
; VEC: [[CONT1]]:
; VEC: br i1 {{.*}}, label %middle.block, label %vector.body
;
define void @minimal_bit_widths(i1 %c) {
entry:
br label %for.body
for.body:
%tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]
%tmp1 = phi i64 [ %tmp7, %for.inc ], [ undef, %entry ]
%tmp2 = getelementptr i8, i8* undef, i64 %tmp0
%tmp3 = load i8, i8* %tmp2, align 1
br i1 %c, label %if.then, label %for.inc
if.then:
%tmp4 = zext i8 %tmp3 to i32
%tmp5 = trunc i32 %tmp4 to i8
store i8 %tmp5, i8* %tmp2, align 1
br label %for.inc
for.inc:
%tmp6 = add nuw nsw i64 %tmp0, 1
%tmp7 = add i64 %tmp1, -1
%tmp8 = icmp eq i64 %tmp7, 0
br i1 %tmp8, label %for.end, label %for.body
for.end:
ret void
}