Hi Andrew, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> writes: > On Thu, 19 Dec 2013 22:53:16 +0100 Arnaud Ebalard <arno@xxxxxxxxxxxx> wrote: > >> +static int isl12057_i2c_validate_chip(struct regmap *regmap) >> +{ >> + u8 regs[ISL12057_MEM_MAP_LEN]; >> + u8 mask[ISL12057_MEM_MAP_LEN] = { 0x80, 0x80, 0x80, 0xf8, >> + 0xc0, 0x60, 0x00, 0x00, >> + 0x00, 0x00, 0x00, 0x00, >> + 0x00, 0x00, 0x60, 0x7c }; >> + int ret, i; >> + >> + ret = regmap_bulk_read(regmap, 0, regs, ISL12057_MEM_MAP_LEN); >> + if (ret) >> + return ret; >> + >> + for (i = 0; i < ISL12057_MEM_MAP_LEN; ++i) { >> + if (regs[i] & mask[i]) /* check if bits are cleared */ >> + return -ENODEV; >> + } >> + >> + return 0; >> +} > > When I build this for x86_64, mask[] is assembled on-stack at runtime: > > leaq -48(%rbp), %r14 #, tmp130 > xorl %esi, %esi # > movl $16, %ecx #, > movq %rax, %rdi # regmap, > incq .LPBX1+552(%rip) # *.LPBX1 > movb $-128, -64(%rbp) #, mask > movq %r14, %rdx # tmp130, > movb $-128, -63(%rbp) #, mask > movb $-128, -62(%rbp) #, mask > movb $-8, -61(%rbp) #, mask > movb $-64, -60(%rbp) #, mask > movb $96, -59(%rbp) #, mask > movb $0, -58(%rbp) #, mask > movb $0, -57(%rbp) #, mask > movb $0, -56(%rbp) #, mask > movb $0, -55(%rbp) #, mask > movb $0, -54(%rbp) #, mask > movb $0, -53(%rbp) #, mask > movb $0, -52(%rbp) #, mask > movb $0, -51(%rbp) #, mask > movb $96, -50(%rbp) #, mask > movb $124, -49(%rbp) #, mask > call regmap_bulk_read # > > If we make `mask' static const, that all gets done at compile time instead: > > leaq -48(%rbp), %r14 #, tmp134 > xorl %esi, %esi # > movl $16, %ecx #, > movq %rax, %rdi # regmap, > incq .LPBX1+552(%rip) # *.LPBX1 > movq %r14, %rdx # tmp134, > call regmap_bulk_read # > > much nicer. Damned. That's a good argument for a v8 ;-) Give me some minutes to assemble one. Cheers, a+ -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html