source: icGREP/icgrep-devel/icgrep/kernels/radix64.cpp @ 5831

Last change on this file since 5831 was 5831, checked in by nmedfort, 16 months ago

Potential bug fix for 32-bit

File size: 17.5 KB
Line 
1/*
2 *  Copyright (c) 2016 International Characters.
3 *  This software is licensed to the public under the Open Software License 3.0.
4 */
5#include "radix64.h"
6#include <kernels/streamset.h>
7#include <kernels/kernel_builder.h>
8
9using namespace llvm;
10
11namespace kernel {
12
13// This kernel produces an expanded input stream by duplicating every third byte.
14// It is implemented using SIMD shufflevector operations.  With 16-byte registers,
15// a single shufflevector operation produces 16 bytes of output data from the
16// 12 bytes of input data.   With 32-byte registers, 32 bytes of output data are
17// produced from 24 bytes of input data.
18//
19// Using aligned SIMD loads, an inner loop processes three registers full of input
20// data (i.e., three BytePacks) to produce four registers full of output.   This is
21// a 3 step process.
22// Step 1:  Load input_pack0, apply the shuffle operation to produce output_pack0.
23//          At this point 3/4 of the data in input_pack0 has been processed.
24// Step 2:  Load input_pack1, apply a shuffle operation to use the remaining
25//          1/4 of input_pack0 and 1/2 of input_pack1 to produce output_pack1.
26//          At this point 1/2 of the data in input_pack1 has been processed.
27// Step 3:  Load input_pack2, apply a shuffle operation to use the remaining 1/2
28//          of input_pack1 and 1/4 of input_pack2 to produce output_pack2.
29//          Then apply a further shuffle opertaion to use the remaining 3/4 of
30//          input_pack2 to produce output_pack3.
31
32// The MultiBlockLogic is based on a natural stride taking 3 packs at a time.
33// In this case, the output produced is exactly 4 packs or 4 blocks, with no pending
34// data maintained in the kernel state.
35//
36// When processing the final partial stride of data, the kernel performs full
37// triple-pack processing for each full or partial triple-pack remaining,
38// relying on the MultiBlockKernel builder to only copy the correct number
39// of bytes to the actual output stream.
40
41void expand3_4Kernel::generateMultiBlockLogic(const std::unique_ptr<KernelBuilder> &iBuilder, Value * const numOfStrides) {
42
43    BasicBlock * expand2_3entry = iBuilder->GetInsertBlock();
44    BasicBlock * expand_3_4_loop = iBuilder->CreateBasicBlock("expand_3_4_loop");
45    BasicBlock * expand3_4_exit = iBuilder->CreateBasicBlock("expand3_4_exit");
46   
47    // Determine the require shufflevector constants.
48    const unsigned PACK_SIZE = iBuilder->getBitBlockWidth()/8;
49   
50    // Construct a list of indexes in  the form
51    // 0, 1, 2, 2, 3, 4, 5, 5, 6, 7, 8, 8, ...
52    unsigned sourceByteIndex = 0;
53    unsigned expand3_4_index[PACK_SIZE];
54    for (unsigned i = 0; i < PACK_SIZE; i++) {
55        expand3_4_index[i] = sourceByteIndex;
56        if (i % 4 != 2) sourceByteIndex++;
57    }
58    unsigned const expand3_4_offset[4] = {PACK_SIZE, 3*PACK_SIZE/4, PACK_SIZE/2, PACK_SIZE/4};
59    Value * expand_3_4_shuffle[4];
60    for (unsigned j = 0; j < 4; j++) {
61        std::vector<Constant *> Idxs;
62        for (unsigned i = 0; i < PACK_SIZE; i++) {
63            Idxs.push_back(ConstantInt::get(iBuilder->getInt32Ty(), expand3_4_offset[j] + expand3_4_index[i]));
64        }
65        expand_3_4_shuffle[j] = ConstantVector::get(Idxs);
66    }
67
68    Constant * triplePackSize = iBuilder->getSize(3 * PACK_SIZE); // 3 packs per loop.
69    UndefValue * undefPack = UndefValue::get(iBuilder->fwVectorType(8));
70   
71    const unsigned packAlign = iBuilder->getBitBlockWidth()/8;
72
73    Value * itemsToDo = mAvailableItemCount[0];
74
75    Value * sourceStream = iBuilder->getInputStreamBlockPtr("sourceStream", iBuilder->getInt32(0));
76    Value * expandedStream = iBuilder->getOutputStreamBlockPtr("expand34Stream", iBuilder->getInt32(0));
77
78    // The main loop processes 3 packs of data at a time.
79    // The initial pack offsets may be nonzero.
80    sourceStream = iBuilder->CreatePointerCast(sourceStream, iBuilder->getInt8PtrTy());
81    expandedStream = iBuilder->CreatePointerCast(expandedStream, iBuilder->getInt8PtrTy());
82    Value * offset = iBuilder->CreateURem(iBuilder->getProcessedItemCount("sourceStream"), iBuilder->getSize(iBuilder->getBitBlockWidth()));
83    Value * sourcePackPtr = iBuilder->CreatePointerCast(iBuilder->CreateGEP(sourceStream, offset), iBuilder->getBitBlockType()->getPointerTo());
84    offset = iBuilder->CreateURem(iBuilder->getProducedItemCount("expand34Stream"), iBuilder->getSize(iBuilder->getBitBlockWidth()));
85    Value * outputPackPtr = iBuilder->CreatePointerCast(iBuilder->CreateGEP(expandedStream, offset), iBuilder->getBitBlockType()->getPointerTo());
86    iBuilder->CreateCondBr(iBuilder->CreateICmpSGT(itemsToDo, iBuilder->getSize(0)), expand_3_4_loop, expand3_4_exit);
87   
88    iBuilder->SetInsertPoint(expand_3_4_loop);
89    PHINode * loopInput_ptr = iBuilder->CreatePHI(sourcePackPtr->getType(), 2);
90    PHINode * loopOutput_ptr = iBuilder->CreatePHI(outputPackPtr->getType(), 2);
91    PHINode * loopItemsRemain = iBuilder->CreatePHI(iBuilder->getSizeTy(), 2);
92
93    loopInput_ptr->addIncoming(sourcePackPtr, expand2_3entry);
94    loopOutput_ptr->addIncoming(outputPackPtr, expand2_3entry);
95    loopItemsRemain->addIncoming(itemsToDo, expand2_3entry);
96
97
98    // Step 1 of the main loop.
99    Value * pack0 = iBuilder->fwCast(8, iBuilder->CreateAlignedLoad(loopInput_ptr, packAlign));
100    Value * expand0 = iBuilder->bitCast(iBuilder->CreateShuffleVector(undefPack, pack0, expand_3_4_shuffle[0]));
101    iBuilder->CreateBlockAlignedStore(expand0, loopOutput_ptr);
102    // Step 2 of the main loop.
103    Value * inPack1_ptr = iBuilder->CreateGEP(loopInput_ptr, iBuilder->getInt32(1));
104    Value * outPack1_ptr = iBuilder->CreateGEP(loopOutput_ptr, iBuilder->getInt32(1));
105    Value * pack1 = iBuilder->fwCast(8, iBuilder->CreateAlignedLoad(inPack1_ptr, packAlign));
106    Value * expand1 = iBuilder->bitCast(iBuilder->CreateShuffleVector(pack0, pack1, expand_3_4_shuffle[1]));
107    iBuilder->CreateBlockAlignedStore(expand1, outPack1_ptr);
108    // Step 3 of the main loop.
109    Value * inPack2_ptr = iBuilder->CreateGEP(loopInput_ptr, iBuilder->getInt32(2));
110    Value * outPack2_ptr = iBuilder->CreateGEP(loopOutput_ptr, iBuilder->getInt32(2));
111    Value * pack2 = iBuilder->fwCast(8, iBuilder->CreateAlignedLoad(inPack2_ptr, packAlign));
112    Value * expand2 = iBuilder->bitCast(iBuilder->CreateShuffleVector(pack1, pack2, expand_3_4_shuffle[2]));
113    iBuilder->CreateBlockAlignedStore(expand2, outPack2_ptr);
114    Value * outPack3_ptr = iBuilder->CreateGEP(loopOutput_ptr, iBuilder->getInt32(3));
115    Value * expand3 = iBuilder->bitCast(iBuilder->CreateShuffleVector(pack2, undefPack, expand_3_4_shuffle[3]));
116    iBuilder->CreateBlockAlignedStore(expand3, outPack3_ptr);
117
118    Value * loopNextInputPack = iBuilder->CreateGEP(loopInput_ptr, iBuilder->getInt32(3));
119    Value * remainingItems = iBuilder->CreateSub(loopItemsRemain, triplePackSize);
120
121    Value * loopNextOutputPack;
122    loopNextOutputPack = iBuilder->CreateGEP(loopOutput_ptr, iBuilder->getInt32(4));
123
124    loopInput_ptr->addIncoming(loopNextInputPack, expand_3_4_loop);
125    loopOutput_ptr->addIncoming(loopNextOutputPack, expand_3_4_loop);
126    loopItemsRemain->addIncoming(remainingItems, expand_3_4_loop);
127
128    Value * continueLoop = iBuilder->CreateICmpSGT(remainingItems, iBuilder->getSize(0));
129    iBuilder->CreateCondBr(continueLoop, expand_3_4_loop, expand3_4_exit);
130   
131    iBuilder->SetInsertPoint(expand3_4_exit);
132
133}
134
135
136// Radix 64 determination, converting 3 bytes to 4 6-bit values.
137//
138//  00000000|zyxwvuts|rqpmnlkj|hgfedcba    Original
139//           zy                            bits to move 6 positions right
140//             xwvuts                      bits to move 8 positions left
141//                    rqpm                 bits to move 4 positions right
142//                        nlkj             bits to move 10 positions left
143//                             hqfedc      bits to move 2 positions right
144//                                   ba    bits to move 12 positions left
145//    xwvuts|  nlkjzy|  barqpm|  hgfedc    Target
146inline Value * radix64Kernel::processPackData(const std::unique_ptr<KernelBuilder> & iBuilder, llvm::Value * bytepack) const {
147
148    Value * step_right_6 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x00C00000));
149    Value * right_6_result = iBuilder->simd_srli(32, iBuilder->simd_and(bytepack, step_right_6), 6);
150
151    Value * step_left_8 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x003F0000));
152    Value * left_8_result = iBuilder->simd_slli(32, iBuilder->simd_and(bytepack, step_left_8), 8);
153    Value * mid = iBuilder->simd_or(right_6_result, left_8_result);
154
155    Value * step_right_4 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x0000F000));
156    Value * right_4_result = iBuilder->simd_srli(32, iBuilder->simd_and(bytepack, step_right_4), 4);
157    mid = iBuilder->simd_or(mid, right_4_result);
158
159    Value * step_left_10 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x00000F00));
160    Value * left_10_result = iBuilder->simd_slli(32, iBuilder->simd_and(bytepack, step_left_10), 10);
161    mid = iBuilder->simd_or(mid, left_10_result);
162
163    Value * step_right_2 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x000000FC));
164    Value * right_2_result = iBuilder->simd_srli(32, iBuilder->simd_and(bytepack, step_right_2), 2);
165    mid = iBuilder->simd_or(mid, right_2_result);
166
167    Value * step_left_12 = iBuilder->simd_fill(32, ConstantInt::get(iBuilder->getInt32Ty(), 0x00000003));
168    Value * left_12_result = iBuilder->simd_slli(32, iBuilder->simd_and(bytepack, step_left_12), 12);
169    mid = iBuilder->simd_or(mid, left_12_result);
170
171    return iBuilder->bitCast(mid);
172}
173
174void radix64Kernel::generateDoBlockMethod(const std::unique_ptr<KernelBuilder> & iBuilder) {
175    for (unsigned i = 0; i < 8; i++) {
176        Value * bytepack = iBuilder->loadInputStreamPack("expandedStream", iBuilder->getInt32(0), iBuilder->getInt32(i));
177        Value * radix64pack = processPackData(iBuilder, bytepack);
178        iBuilder->storeOutputStreamPack("radix64stream", iBuilder->getInt32(0), iBuilder->getInt32(i), radix64pack);
179    }
180}
181
182void radix64Kernel::generateFinalBlockMethod(const std::unique_ptr<KernelBuilder> & iBuilder, Value * remainingBytes) {
183
184    BasicBlock * entry = iBuilder->GetInsertBlock();
185    BasicBlock * radix64_loop = iBuilder->CreateBasicBlock("radix64_loop");
186    BasicBlock * fbExit = iBuilder->CreateBasicBlock("fbExit");
187   
188    const unsigned PACK_SIZE = iBuilder->getStride()/8;
189    Constant * packSize = iBuilder->getSize(PACK_SIZE);
190
191    // Enter the loop only if there is at least one byte remaining to process.
192    iBuilder->CreateCondBr(iBuilder->CreateICmpEQ(remainingBytes, iBuilder->getSize(0)), fbExit, radix64_loop);
193
194    iBuilder->SetInsertPoint(radix64_loop);
195    PHINode * idx = iBuilder->CreatePHI(iBuilder->getInt32Ty(), 2);
196    PHINode * loopRemain = iBuilder->CreatePHI(iBuilder->getSizeTy(), 2);
197    idx->addIncoming(ConstantInt::getNullValue(iBuilder->getInt32Ty()), entry);
198    loopRemain->addIncoming(remainingBytes, entry);
199
200    Value * bytepack = iBuilder->loadInputStreamPack("expandedStream", iBuilder->getInt32(0), idx);
201    Value * radix64pack = processPackData(iBuilder, bytepack);
202    iBuilder->storeOutputStreamPack("radix64stream", iBuilder->getInt32(0), idx, radix64pack);
203
204    Value* nextIdx = iBuilder->CreateAdd(idx, ConstantInt::get(iBuilder->getInt32Ty(), 1));
205    idx->addIncoming(nextIdx, radix64_loop);
206    Value* remainAfterLoop = iBuilder->CreateSub(loopRemain, packSize);
207    loopRemain->addIncoming(remainAfterLoop, radix64_loop);
208
209    Value* continueLoop = iBuilder->CreateICmpSGT(remainAfterLoop, iBuilder->getSize(0));
210
211    iBuilder->CreateCondBr(continueLoop, radix64_loop, fbExit);
212
213    iBuilder->SetInsertPoint(fbExit);
214}
215
216inline llvm::Value* base64Kernel::processPackData(const std::unique_ptr<KernelBuilder> & iBuilder, llvm::Value* bytepack) const {
217    Value * mask_gt_25 = iBuilder->simd_ugt(8, bytepack, iBuilder->simd_fill(8, iBuilder->getInt8(25)));
218    Value * mask_gt_51 = iBuilder->simd_ugt(8, bytepack, iBuilder->simd_fill(8, iBuilder->getInt8(51)));
219    Value * mask_eq_62 = iBuilder->simd_eq(8, bytepack, iBuilder->simd_fill(8, iBuilder->getInt8(62)));
220    Value * mask_eq_63 = iBuilder->simd_eq(8, bytepack, iBuilder->simd_fill(8, iBuilder->getInt8(63)));
221    // Strategy:
222    // 1. add ord('A') = 65 to all radix64 values, this sets the correct values for entries 0 to 25.
223    // 2. add ord('a') - ord('A') - (26 - 0) = 6 to all values >25, this sets the correct values for entries 0 to 51
224    // 3. subtract ord('a') - ord('0') + (52 - 26) = 75 to all values > 51, this sets the correct values for entries 0 to 61
225    // 4. subtract ord('0') - ord('+') + (62 - 52) = 15 for all values = 62
226    // 4. add ord('/') - ord('0') - (63 - 52) = 3 for all values = 63
227    Value * t0_25 = iBuilder->simd_add(8, bytepack, iBuilder->simd_fill(8, iBuilder->getInt8('A')));
228    Value * t0_51 = iBuilder->simd_add(8, t0_25, iBuilder->simd_and(mask_gt_25, iBuilder->simd_fill(8, iBuilder->getInt8(6))));
229    Value * t0_61 = iBuilder->simd_sub(8, t0_51, iBuilder->simd_and(mask_gt_51, iBuilder->simd_fill(8, iBuilder->getInt8(75))));
230    Value * t0_62 = iBuilder->simd_sub(8, t0_61, iBuilder->simd_and(mask_eq_62, iBuilder->simd_fill(8, iBuilder->getInt8(15))));
231    return iBuilder->bitCast(iBuilder->simd_sub(8, t0_62, iBuilder->simd_and(mask_eq_63, iBuilder->simd_fill(8, iBuilder->getInt8(12)))));
232}
233
234void base64Kernel::generateDoBlockMethod(const std::unique_ptr<KernelBuilder> & iBuilder) {
235    for (unsigned i = 0; i < 8; i++) {
236        Value * bytepack = iBuilder->loadInputStreamPack("radix64stream", iBuilder->getInt32(0), iBuilder->getInt32(i));
237        Value * base64pack = processPackData(iBuilder, bytepack);
238        iBuilder->storeOutputStreamPack("base64stream", iBuilder->getInt32(0), iBuilder->getInt32(i), base64pack);
239    }
240}
241
242// Special processing for the base 64 format.   The output must always contain a multiple
243// of 4 bytes.   When the number of radix 64 values is not a multiple of 4
244// number of radix 64 values
245void base64Kernel::generateFinalBlockMethod(const std::unique_ptr<KernelBuilder> & iBuilder, Value * remainingBytes) {
246
247    BasicBlock * entry = iBuilder->GetInsertBlock();
248    BasicBlock * base64_loop = iBuilder->CreateBasicBlock("base64_loop");
249    BasicBlock * loopExit = iBuilder->CreateBasicBlock("loopExit");
250    BasicBlock * doPadding = iBuilder->CreateBasicBlock("doPadding");
251    BasicBlock * doPadding2 = iBuilder->CreateBasicBlock("doPadding2");
252    BasicBlock * fbExit = iBuilder->CreateBasicBlock("fbExit");
253
254    Value * remainMod4 = iBuilder->CreateAnd(remainingBytes, iBuilder->getSize(3));
255    Value * padBytes = iBuilder->CreateSub(iBuilder->getSize(4), remainMod4);
256    padBytes = iBuilder->CreateAnd(padBytes, iBuilder->getSize(3));
257
258    Constant * packSize = iBuilder->getSize(iBuilder->getStride() / 8);
259
260    // Enter the loop only if there is at least one byte remaining to process.
261    iBuilder->CreateCondBr(iBuilder->CreateICmpEQ(remainingBytes, iBuilder->getSize(0)), fbExit, base64_loop);
262
263    iBuilder->SetInsertPoint(base64_loop);
264    PHINode * idx = iBuilder->CreatePHI(iBuilder->getInt32Ty(), 2);
265    PHINode * loopRemain = iBuilder->CreatePHI(iBuilder->getSizeTy(), 2);
266    idx->addIncoming(ConstantInt::getNullValue(iBuilder->getInt32Ty()), entry);
267    loopRemain->addIncoming(remainingBytes, entry);
268    Value * bytepack = iBuilder->loadInputStreamPack("radix64stream", iBuilder->getInt32(0), idx);
269    Value * base64pack = processPackData(iBuilder, bytepack);
270    iBuilder->storeOutputStreamPack("base64stream", iBuilder->getInt32(0), idx, base64pack);
271    idx->addIncoming(iBuilder->CreateAdd(idx, ConstantInt::get(iBuilder->getInt32Ty(), 1)), base64_loop);
272    Value* remainAfterLoop = iBuilder->CreateSub(loopRemain, packSize);
273    loopRemain->addIncoming(remainAfterLoop, base64_loop);
274
275    Value* continueLoop = iBuilder->CreateICmpSGT(remainAfterLoop, iBuilder->getSize(0));
276    iBuilder->CreateCondBr(continueLoop, base64_loop, loopExit);
277
278    iBuilder->SetInsertPoint(loopExit);
279    iBuilder->CreateCondBr(iBuilder->CreateICmpEQ(padBytes, iBuilder->getSize(0)), fbExit, doPadding);
280
281    iBuilder->SetInsertPoint(doPadding);
282    Value * i8output_ptr = iBuilder->getOutputStreamBlockPtr("base64stream", iBuilder->getInt32(0));
283    i8output_ptr = iBuilder->CreatePointerCast(i8output_ptr, iBuilder->getInt8PtrTy());
284    iBuilder->CreateStore(ConstantInt::get(iBuilder->getInt8Ty(), '='), iBuilder->CreateGEP(i8output_ptr, remainingBytes));
285    iBuilder->CreateCondBr(iBuilder->CreateICmpEQ(remainMod4, iBuilder->getSize(3)), fbExit, doPadding2);
286    iBuilder->SetInsertPoint(doPadding2);
287    Value * finalPadPos = iBuilder->CreateAdd(remainingBytes, iBuilder->getSize(1));
288    iBuilder->CreateStore(ConstantInt::get(iBuilder->getInt8Ty(), '='), iBuilder->CreateGEP(i8output_ptr, finalPadPos));
289    iBuilder->CreateBr(fbExit);
290    iBuilder->SetInsertPoint(fbExit);
291}
292
293expand3_4Kernel::expand3_4Kernel(const std::unique_ptr<kernel::KernelBuilder> & iBuilder)
294: MultiBlockKernel("expand3_4",
295            {Binding{iBuilder->getStreamSetTy(1, 8), "sourceStream", FixedRate(3)}},
296            {Binding{iBuilder->getStreamSetTy(1, 8), "expand34Stream", FixedRate(4)}},
297            {}, {}, {}) {
298
299}
300
301radix64Kernel::radix64Kernel(const std::unique_ptr<kernel::KernelBuilder> & iBuilder)
302: BlockOrientedKernel("radix64",
303            {Binding{iBuilder->getStreamSetTy(1, 8), "expandedStream"}},
304            {Binding{iBuilder->getStreamSetTy(1, 8), "radix64stream"}},
305            {}, {}, {}) {
306}
307
308base64Kernel::base64Kernel(const std::unique_ptr<kernel::KernelBuilder> & iBuilder)
309: BlockOrientedKernel("base64",
310            {Binding{iBuilder->getStreamSetTy(1, 8), "radix64stream"}},
311            {Binding{iBuilder->getStreamSetTy(1, 8), "base64stream", FixedRate(1), RoundUpTo(4)}},
312            {}, {}, {}) {
313}
314
315}
Note: See TracBrowser for help on using the repository browser.