Re: [PATCH net-next] net: ethernet: ti: cpsw: replace unnecessarily macroses on functions

From: Joe Perches
Date: Fri Jul 27 2018 - 16:04:28 EST


On Fri, 2018-07-27 at 22:36 +0300, Ivan Khoronzhuk wrote:
> On Fri, Jul 27, 2018 at 12:21:07PM -0700, Joe Perches wrote:
> > On Fri, 2018-07-27 at 22:13 +0300, Ivan Khoronzhuk wrote:
> > > Replace ugly macroses on functions.
> >
> > Careful, see below.
> >
> > > diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
> >
> > []
> > > @@ -565,40 +565,40 @@ static const struct cpsw_stats cpsw_gstrings_ch_stats[] = {
> > > (func)(slave++, ##arg); \
> > > } while (0)
> > >
> > > -#define cpsw_dual_emac_src_port_detect(cpsw, status, ndev, skb) \
> > > - do { \
> > > - if (!cpsw->data.dual_emac) \
> > > - break; \
> > > - if (CPDMA_RX_SOURCE_PORT(status) == 1) { \
> > > - ndev = cpsw->slaves[0].ndev; \
> > > - skb->dev = ndev; \
> > > - } else if (CPDMA_RX_SOURCE_PORT(status) == 2) { \
> > > - ndev = cpsw->slaves[1].ndev; \
> > > - skb->dev = ndev; \
> > > - } \
> > > - } while (0)
> > > -#define cpsw_add_mcast(cpsw, priv, addr) \
> > > - do { \
> > > - if (cpsw->data.dual_emac) { \
> > > - struct cpsw_slave *slave = cpsw->slaves + \
> > > - priv->emac_port; \
> > > - int slave_port = cpsw_get_slave_port( \
> > > - slave->slave_num); \
> > > - cpsw_ale_add_mcast(cpsw->ale, addr, \
> > > - 1 << slave_port | ALE_PORT_HOST, \
> > > - ALE_VLAN, slave->port_vlan, 0); \
> > > - } else { \
> > > - cpsw_ale_add_mcast(cpsw->ale, addr, \
> > > - ALE_ALL_PORTS, \
> > > - 0, 0, 0); \
> > > - } \
> > > - } while (0)
> > > -
> > > static inline int cpsw_get_slave_port(u32 slave_num)
> > > {
> > > return slave_num + 1;
> > > }
> > >
> > > +static inline void cpsw_src_port_detect(struct cpsw_common *cpsw, int status,
> > > + struct sk_buff *skb)
> > > +{
> > > + if (!cpsw->data.dual_emac)
> > > + return;
> > > +
> > > + if (CPDMA_RX_SOURCE_PORT(status) == 1)
> > > + skb->dev = cpsw->slaves[0].ndev;
> > > + else if (CPDMA_RX_SOURCE_PORT(status) == 2)
> > > + skb->dev = cpsw->slaves[1].ndev;
> > > +}
> >
> > perhaps better as a switch/case
>
> not better, it's shorter.

True for the source code but it compiles to more object code.

$ cat foo.c
struct cpsw_common {
struct {
int dual_emac;
} data;
struct {
int ndev;
} slaves[2];
};

struct sk_buff {
int dev;
};

#define CPDMA_RX_SOURCE_PORT(__status__) ((__status__ >> 16) & 0x7)

#if defined SWITCH

void foo(struct cpsw_common *cpsw, int status, struct sk_buff *skb)
{
if (!cpsw->data.dual_emac)
return;

switch (CPDMA_RX_SOURCE_PORT(status)) {
case 1:
skb->dev = cpsw->slaves[0].ndev;
break;
case 2:
skb->dev = cpsw->slaves[1].ndev;
break;
}
}

#else

void foo(struct cpsw_common *cpsw, int status, struct sk_buff *skb)
{
if (!cpsw->data.dual_emac)
return;

if (CPDMA_RX_SOURCE_PORT(status) == 1)
skb->dev = cpsw->slaves[0].ndev;
else if (CPDMA_RX_SOURCE_PORT(status) == 2)
skb->dev = cpsw->slaves[1].ndev;
}

#endif
$ gcc -c -O2 -DSWITCH foo.c
$ size foo.o
text data bss dec hex filename
94 0 0 94 5e foo.o
$ objdump -d foo.o

foo.o: file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <foo>:
0: 8b 07 mov (%rdi),%eax
2: 85 c0 test %eax,%eax
4: 74 15 je 1b <foo+0x1b>
6: c1 fe 10 sar $0x10,%esi
9: 83 e6 07 and $0x7,%esi
c: 83 fe 01 cmp $0x1,%esi
f: 74 17 je 28 <foo+0x28>
11: 83 fe 02 cmp $0x2,%esi
14: 75 0a jne 20 <foo+0x20>
16: 8b 47 08 mov 0x8(%rdi),%eax
19: 89 02 mov %eax,(%rdx)
1b: f3 c3 repz retq
1d: 0f 1f 00 nopl (%rax)
20: f3 c3 repz retq
22: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1)
28: 8b 47 04 mov 0x4(%rdi),%eax
2b: 89 02 mov %eax,(%rdx)
2d: c3 retq
$ gcc -c -O2 foo.c
$ size foo.o
text data bss dec hex filename
102 0 0 102 66 foo.o
$ objdump -d foo.o

foo.o: file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <foo>:
0: 8b 07 mov (%rdi),%eax
2: 85 c0 test %eax,%eax
4: 74 10 je 16 <foo+0x16>
6: c1 fe 10 sar $0x10,%esi
9: 83 e6 07 and $0x7,%esi
c: 83 fe 01 cmp $0x1,%esi
f: 74 0f je 20 <foo+0x20>
11: 83 fe 02 cmp $0x2,%esi
14: 74 1a je 30 <foo+0x30>
16: f3 c3 repz retq
18: 0f 1f 84 00 00 00 00 nopl 0x0(%rax,%rax,1)
1f: 00
20: 8b 47 04 mov 0x4(%rdi),%eax
23: 89 02 mov %eax,(%rdx)
25: c3 retq
26: 66 2e 0f 1f 84 00 00 nopw %cs:0x0(%rax,%rax,1)
2d: 00 00 00
30: 8b 47 08 mov 0x8(%rdi),%eax
33: 89 02 mov %eax,(%rdx)
35: c3 retq