radeon/llvm: Run standard optimization passes on conpute shader modules

The SROA and function inliner passes are espically important, because
they optimize away unsupported features: functions and indirect
private memory access.

diff --git a/src/gallium/drivers/radeon/radeon_llvm_util.c b/src/gallium/drivers/radeon/radeon_llvm_util.c
index e3b0d63..25be245 100644 (file)
--- a/src/gallium/drivers/radeon/radeon_llvm_util.c
+++ b/src/gallium/drivers/radeon/radeon_llvm_util.c
@@ -29,6 +29,7 @@
 
 #include <llvm-c/BitReader.h>
 #include <llvm-c/Core.h>
+#include <llvm-c/Transforms/PassManagerBuilder.h>
 
 LLVMModuleRef radeon_llvm_parse_bitcode(const unsigned char * bitcode,
                                                        unsigned bitcode_len)
@@ -50,6 +51,19 @@ unsigned radeon_llvm_get_num_kernels(const unsigned char *bitcode,
        return LLVMGetNamedMetadataNumOperands(mod, "opencl.kernels");
 }
 
+static void radeon_llvm_optimize(LLVMModuleRef mod)
+{
+       LLVMPassManagerBuilderRef builder = LLVMPassManagerBuilderCreate();
+       LLVMPassManagerRef pass_manager = LLVMCreatePassManager();
+
+       LLVMPassManagerBuilderUseInlinerWithThreshold(builder, 1000000000);
+       LLVMPassManagerBuilderPopulateModulePassManager(builder, pass_manager);
+
+       LLVMRunPassManager(pass_manager, mod);
+       LLVMPassManagerBuilderDispose(builder);
+       LLVMDisposePassManager(pass_manager);
+}
+
 LLVMModuleRef radeon_llvm_get_kernel_module(unsigned index,
                const unsigned char *bitcode, unsigned bitcode_len)
 {
@@ -72,5 +86,6 @@ LLVMModuleRef radeon_llvm_get_kernel_module(unsigned index,
                LLVMDeleteFunction(kernel_function);
        }
        FREE(kernel_metadata);
+       radeon_llvm_optimize(mod);
        return mod;
 }