src/gallium/drivers/radeon/radeon_llvm_emit.cpp

   1 /*
   2  * Copyright 2011 Advanced Micro Devices, Inc.
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  *
  23  * Authors: Tom Stellard <thomas.stellard@amd.com>
  24  *
  25  */
  26 #include "radeon_llvm_emit.h"
  27
  28 #include <llvm/LLVMContext.h>
  29 #include <llvm/Module.h>
  30 #include <llvm/PassManager.h>
  31 #include <llvm/ADT/Triple.h>
  32 #include <llvm/Support/FormattedStream.h>
  33 #include <llvm/Support/Host.h>
  34 #include <llvm/Support/IRReader.h>
  35 #include <llvm/Support/SourceMgr.h>
  36 #include <llvm/Support/TargetRegistry.h>
  37 #include <llvm/Support/TargetSelect.h>
  38 #include <llvm/Support/Threading.h>
  39 #include <llvm/Target/TargetMachine.h>
  40 #include <llvm/Transforms/Scalar.h>
  41 #include <llvm-c/Target.h>
  42
  43 #if HAVE_LLVM < 0x0302
  44 #include <llvm/Target/TargetData.h>
  45 #else
  46 #include <llvm/DataLayout.h>
  47 #endif
  48
  49 #include <iostream>
  50 #include <stdlib.h>
  51 #include <stdio.h>
  52
  53 using namespace llvm;
  54
  55 #ifndef EXTERNAL_LLVM
  56 extern "C" {
  57
  58 void LLVMInitializeAMDGPUAsmPrinter(void);
  59 void LLVMInitializeAMDGPUTargetMC(void);
  60 void LLVMInitializeAMDGPUTarget(void);
  61 void LLVMInitializeAMDGPUTargetInfo(void);
  62 }
  63 #endif
  64
  65 namespace {
  66
  67 class LLVMEnsureMultithreaded {
  68 public:
  69    LLVMEnsureMultithreaded()
  70    {
  71       llvm_start_multithreaded();
  72    }
  73 };
  74
  75 static LLVMEnsureMultithreaded lLVMEnsureMultithreaded;
  76
  77 }
  78
  79 /**
  80  * Compile an LLVM module to machine code.
  81  *
  82  * @param bytes This function allocates memory for the byte stream, it is the
  83  * caller's responsibility to free it.
  84  */
  85 extern "C" unsigned
  86 radeon_llvm_compile(LLVMModuleRef M, unsigned char ** bytes,
  87                  unsigned * byte_count, const char * gpu_family,
  88                  unsigned dump) {
  89
  90    Triple AMDGPUTriple(sys::getDefaultTargetTriple());
  91
  92 #if HAVE_LLVM == 0x0302
  93    LLVMInitializeAMDGPUTargetInfo();
  94    LLVMInitializeAMDGPUTarget();
  95    LLVMInitializeAMDGPUTargetMC();
  96    LLVMInitializeAMDGPUAsmPrinter();
  97 #else
  98    LLVMInitializeR600TargetInfo();
  99    LLVMInitializeR600Target();
 100    LLVMInitializeR600TargetMC();
 101    LLVMInitializeR600AsmPrinter();
 102 #endif
 103
 104    std::string err;
 105    const Target * AMDGPUTarget = TargetRegistry::lookupTarget("r600", err);
 106    if(!AMDGPUTarget) {
 107       fprintf(stderr, "Can't find target: %s\n", err.c_str());
 108       return 1;
 109    }
 110
 111    Triple::ArchType Arch = Triple::getArchTypeForLLVMName("r600");
 112    if (Arch == Triple::UnknownArch) {
 113       fprintf(stderr, "Unknown Arch\n");
 114    }
 115    AMDGPUTriple.setArch(Arch);
 116
 117    Module * mod = unwrap(M);
 118    std::string FS;
 119    TargetOptions TO;
 120
 121    if (dump) {
 122       mod->dump();
 123       FS += "+DumpCode";
 124    }
 125
 126    std::auto_ptr<TargetMachine> tm(AMDGPUTarget->createTargetMachine(
 127                      AMDGPUTriple.getTriple(), gpu_family, FS,
 128                      TO, Reloc::Default, CodeModel::Default,
 129                      CodeGenOpt::Default
 130                      ));
 131    TargetMachine &AMDGPUTargetMachine = *tm.get();
 132    PassManager PM;
 133 #if HAVE_LLVM < 0x0302
 134    PM.add(new TargetData(*AMDGPUTargetMachine.getTargetData()));
 135 #else
 136    PM.add(new DataLayout(*AMDGPUTargetMachine.getDataLayout()));
 137 #endif
 138    PM.add(createPromoteMemoryToRegisterPass());
 139    AMDGPUTargetMachine.setAsmVerbosityDefault(true);
 140
 141    std::string CodeString;
 142    raw_string_ostream oStream(CodeString);
 143    formatted_raw_ostream out(oStream);
 144
 145    /* Optional extra paramater true / false to disable verify */
 146    if (AMDGPUTargetMachine.addPassesToEmitFile(PM, out, TargetMachine::CGFT_ObjectFile,
 147                                                true)){
 148       fprintf(stderr, "AddingPasses failed.\n");
 149       return 1;
 150    }
 151    PM.run(*mod);
 152
 153    out.flush();
 154    std::string &data = oStream.str();
 155
 156    *bytes = (unsigned char*)malloc(data.length() * sizeof(unsigned char));
 157    memcpy(*bytes, data.c_str(), data.length() * sizeof(unsigned char));
 158    *byte_count = data.length();
 159
 160    return 0;
 161 }