[Intel-gfx] [RFC 00/15] Selectable platform support

[Chris Wilson]

Quoting Chris Wilson (2018-02-09 11:19:44)
> Ideas? A long time ago, we wanted a static INTEL_INFO. I think now we
> have driver_caps, we try to kill off mkwrite_intel_info() and where
> need be use INTEL_INFO() && DRIVER_CAPS().
> 
> Tvrtko, how easy do you think it will be to go from Kconfig to
> 
> #define INTEL_INFO(dev_priv) (&platform_info)
> ?

Fwiw, I've been playing around with getting INTEL_INFO() back to being a
static blob, but the DCE elimination for a single platform_info does
require LTO. (And no, I can't build Andi Keen's lto-415-2 branch either.)

--- ctl.h ---
#include <stdbool.h>

struct control {
	bool exists;
};

--- ctl.c ---
#include "ctl.h"

const struct control ctl = {
	.exists = false,
};

--- test.c ---
#include "ctl.h"

static void unwanted(int *x)
{
	*x = 0;
}

extern const struct control ctl;

int main(void)
{
	if (ctl.exists)
		unwanted(0);

	return 0;
}

Without LTO, gcc -Wall -g test.c ctl.c -O2
Dump of assembler code for function main:
   0x00000000004003e0 <+0>:	cmpb   $0x0,0x1ad(%rip)        # 0x400594 <ctl>
   0x00000000004003e7 <+7>:	je     0x400400 <main+32>
   0x00000000004003e9 <+9>:	movl   $0x0,0x0
   0x00000000004003f4 <+20>:	ud2    
   0x00000000004003f6 <+22>:	nopw   %cs:0x0(%rax,%rax,1)
   0x0000000000400400 <+32>:	xor    %eax,%eax
   0x0000000000400402 <+34>:	retq   

With LTO, gcc -Wall -g test.c ctl.c -flto -O2
Dump of assembler code for function main:
   0x00000000004003e0 <+0>:	xor    %eax,%eax
   0x00000000004003e2 <+2>:	retq   

Let's hope that magic scales!
-Chris