[Mesa-dev] [PATCH 5/9] gallium/radeon: always add +DumpCode to the LLVM target machine for LLVM <= 3.5

[Nicolai Hähnle]

What's the reason for always having +DumpCode? Generating the assembly 
is some overhead that's usually unnecessary. Even if it's a small part 
of the profiles I've seen, it still seems like a natural thing to just 
skip. From what I can tell it should be dependent on any of the shader 
dumping flags + DBG_CHECK_VM being set. In any case, I suppose that 
would be for a separate commit.

Cheers,
Nicolai

On 01.01.2016 09:13, Marek Olšák wrote:
> From: Marek Olšák <marek.olsak at amd.com>
>
> It's the same behavior that we use for later LLVM.
> ---
>   src/gallium/drivers/r600/r600_llvm.c          | 2 +-
>   src/gallium/drivers/radeon/radeon_llvm_emit.c | 5 ++---
>   src/gallium/drivers/radeon/radeon_llvm_emit.h | 2 +-
>   src/gallium/drivers/radeonsi/si_shader.c      | 2 +-
>   4 files changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/src/gallium/drivers/r600/r600_llvm.c b/src/gallium/drivers/r600/r600_llvm.c
> index 1cc3031..7d93658 100644
> --- a/src/gallium/drivers/r600/r600_llvm.c
> +++ b/src/gallium/drivers/r600/r600_llvm.c
> @@ -922,7 +922,7 @@ unsigned r600_llvm_compile(
>   	const char * gpu_family = r600_get_llvm_processor_name(family);
>
>   	memset(&binary, 0, sizeof(struct radeon_shader_binary));
> -	r = radeon_llvm_compile(mod, &binary, gpu_family, dump, dump, NULL);
> +	r = radeon_llvm_compile(mod, &binary, gpu_family, dump, NULL);
>
>   	r = r600_create_shader(bc, &binary, use_kill);
>
> diff --git a/src/gallium/drivers/radeon/radeon_llvm_emit.c b/src/gallium/drivers/radeon/radeon_llvm_emit.c
> index 61ed940..f8c7f54 100644
> --- a/src/gallium/drivers/radeon/radeon_llvm_emit.c
> +++ b/src/gallium/drivers/radeon/radeon_llvm_emit.c
> @@ -141,7 +141,7 @@ static void radeonDiagnosticHandler(LLVMDiagnosticInfoRef di, void *context)
>    * @returns 0 for success, 1 for failure
>    */
>   unsigned radeon_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binary,
> -			     const char *gpu_family, bool dump_ir, bool dump_asm,
> +			     const char *gpu_family, bool dump_ir,
>   			     LLVMTargetMachineRef tm)
>   {
>
> @@ -165,8 +165,7 @@ unsigned radeon_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binar
>   		}
>   		strncpy(cpu, gpu_family, CPU_STRING_LEN);
>   		memset(fs, 0, sizeof(fs));
> -		if (dump_asm)
> -			strncpy(fs, "+DumpCode", FS_STRING_LEN);
> +		strncpy(fs, "+DumpCode", FS_STRING_LEN);
>   		tm = LLVMCreateTargetMachine(target, triple, cpu, fs,
>   				  LLVMCodeGenLevelDefault, LLVMRelocDefault,
>   						  LLVMCodeModelDefault);
> diff --git a/src/gallium/drivers/radeon/radeon_llvm_emit.h b/src/gallium/drivers/radeon/radeon_llvm_emit.h
> index e20aed9..5f956dd 100644
> --- a/src/gallium/drivers/radeon/radeon_llvm_emit.h
> +++ b/src/gallium/drivers/radeon/radeon_llvm_emit.h
> @@ -38,7 +38,7 @@ void radeon_llvm_shader_type(LLVMValueRef F, unsigned type);
>   LLVMTargetRef radeon_llvm_get_r600_target(const char *triple);
>
>   unsigned radeon_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binary,
> -			     const char *gpu_family, bool dump_ir, bool dump_asm,
> +			     const char *gpu_family, bool dump_ir,
>   			     LLVMTargetMachineRef tm);
>
>   #endif /* RADEON_LLVM_EMIT_H */
> diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
> index a9297a5..4044961 100644
> --- a/src/gallium/drivers/radeonsi/si_shader.c
> +++ b/src/gallium/drivers/radeonsi/si_shader.c
> @@ -3884,7 +3884,7 @@ int si_compile_llvm(struct si_screen *sscreen, struct si_shader *shader,
>   	bool dump_ir = dump_asm && !(sscreen->b.debug_flags & DBG_NO_IR);
>
>   	r = radeon_llvm_compile(mod, &shader->binary,
> -		r600_get_llvm_processor_name(sscreen->b.family), dump_ir, dump_asm, tm);
> +		r600_get_llvm_processor_name(sscreen->b.family), dump_ir, tm);
>   	if (r)
>   		return r;
>
>