P4 to Git Change 1086296 by emankov@em-hsa-amd on 2014/10/10 07:46:11

ECR #333753 - HSA RT/Performance: Elimination of HSAIL text usage in RT (part 2)

	(as it's already done for gpuprogram.cpp)

	Extracting HSAIL from the binary and parsing it for the kernel names in RT were replaced with aclQueryInfo call for RT_KERNEL_NAMES. Kernel names are obtained now from the corresponding metadata symbols names, which are already presented in BIF at kernel finalization stage.

	Side effect: performance improvement

	Next Step: Performance: Stop obligatory BRIG disassembling to HSAIL and insertion into BIF (previously was needed only by RT).

	testing: pre check-in

	Reviewers: Stanislav Mekhanoshin, German Andryeyev, Brian Sumner

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/hsa/hsaprogram.cpp#37 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/hsa_foundation/hsaprogram.cpp#6 edit


[ROCm/clr commit: 3e74ee1f88]

projects/clr/rocclr/runtime/device/hsa/hsaprogram.cpp

@@ -13,7 +13,7 @@
 #include "runtime/device/hsa/hsacompilerlib.hpp"
 #include "runtime/device/hsa/oclhsa_common.hpp"
 #include "utils/bif_section_labels.hpp"
-
+#include "utils/libUtils.h"
 
 #include <string>
 #include <vector>
@@ -625,50 +625,32 @@ namespace oclhsa {
             return true;
         }
 
-        size_t fsailSize;
-        const void *hsailText = g_complibApi._aclExtractSection(device().compiler(),
-            binaryElf_,
-            &fsailSize,
-            aclCODEGEN,
-            &errorCode);
-        if (errorCode != ACL_SUCCESS) {
-            buildLog_ += "Error while reading out the HSAIL from the ELF" ;
-            return false;
-        }
         const HsaDevice *hsaDevice = dev().getBackendDevice();
         if (!loadBrig()) {
             buildLog_ += "Error while loading BRIG" ;
             return false;
         }
 
-        std::string hsailProgram((char*)hsailText, fsailSize);
-        fsailProgram_ = hsailProgram;
-        // We pull out all the kernel names in a very ugly manner
-        //Todo(sramalin) : check if this has been fixed in the compiler library
-        // If not file an EPR
-        if (!fsailProgram_.empty()) {
-            // Find out the name of the kernel. Works for multiple kernels
-            std::vector<std::string> kernelNames;
-            int pos = 0;
-            while (true) {
-                std::string findString = "kernel &";
-                size_t kernelNPos = fsailProgram_.find(findString, pos);
-                if (kernelNPos == std::string::npos) {
-                    break;
-                }
-                size_t kernelEndNPos = fsailProgram_.find("l(", kernelNPos);
-                pos = kernelEndNPos + 1;
-                if (kernelEndNPos == std::string::npos) {
-                    break;
-                }
-                // "kernel &" is 8
-                // "__OpenCL_" is 9
-                // "_kerne" is 6
-                // We can drop all this with a compiler tweak later
-                std::string kernelName = fsailProgram_.substr(kernelNPos + 8 + 9,
-                    kernelEndNPos -
-                    (kernelNPos + 8 + 9) - 6);
-                kernelNames.push_back(kernelName);
+        size_t kernelNamesSize = 0;
+        errorCode = aclQueryInfo(dev().compiler(), binaryElf_, RT_KERNEL_NAMES, NULL, NULL, &kernelNamesSize);
+        if (errorCode != ACL_SUCCESS) {
+            buildLog_ += "Error while Finalization phase: kernel names query from the ELF failed\n";
+            return false;
+        }
+        if (kernelNamesSize > 0) {
+            char* kernelNames = new char[kernelNamesSize];
+            errorCode = aclQueryInfo(dev().compiler(), binaryElf_, RT_KERNEL_NAMES, NULL, kernelNames, &kernelNamesSize);
+            if (errorCode != ACL_SUCCESS) {
+                buildLog_ += "Error while Finalization phase: kernel's Metadata is corrupted in the ELF\n";
+                delete kernelNames;
+                return false;
+            }
+            std::vector<std::string> vKernels = splitSpaceSeparatedString(kernelNames);
+            delete kernelNames;
+            std::vector<std::string>::iterator it = vKernels.begin();
+            bool dynamicParallelism = false;
+            for (it; it != vKernels.end(); ++it) {
+                std::string kernelName = *it;
                 Kernel *aKernel = new oclhsa::Kernel(kernelName,
                     this,
                     &brig_,
@@ -676,21 +658,18 @@ namespace oclhsa {
                 if (!aKernel->init() ) {
                     return false;
                 }
-                aKernel->setUniformWorkGroupSize(options
-                 ->oVariables->UniformWorkGroupSize);
-                //Update the binary in the FSAILProgram to save the ISA and debug
-                // information. This is so the debugger and the profiler can use the
-                // a single aclBinary for all their needs.
+                aKernel->setUniformWorkGroupSize(options->oVariables->UniformWorkGroupSize);
+                // Update the binary in the FSAILProgram to save the ISA and debug information.
+                // This is so the debugger and the profiler can use the a single aclBinary for all their needs.
                 if (!updateAclBinaryWithKernelIsaAndDebug(kernelName)) {
-                  return false;
+                    return false;
                 }
                 kernels()[kernelName] = aKernel;
             }
-            saveBinaryAndSetType(TYPE_EXECUTABLE);
-            buildLog_ += g_complibApi._aclGetCompilerLog(device().compiler());
-            return true;
         }
-        return false;
+        saveBinaryAndSetType(TYPE_EXECUTABLE);
+        buildLog_ += g_complibApi._aclGetCompilerLog(device().compiler());
+        return true;
     }
 
     bool FSAILProgram::createBinary(amd::option::Options *options) {