Ted Roby wrote:
> Consider the following structs. The first struct uses 1 bit wide unsigned
> integers
> for two flags set by the user during runtime. The second struct performs
> the same
> function, but used a regular signed integer instead of an unsigned integer
> with
> a defined bit width.
>
> typedef struct {
> char *dev; /* the device we will
> monitor */
> pcap_t *dev_handle; /* the handle to this
> device after opening */
> struct pcap_stat ps; /* packet statistics
> struct
> */
> pcap_dumper_t *pd; /* dump file pointer */
> int link_int; /* datalink int of device
> */
> int timeout; /* datalink timeout per
> packet */
> int pcount; /* packets captured */
> const char *link_desc; /* datalink description of
> device */
> const char *link_name; /* datalink name of device
> */
> char errbuf[PCAP_ERRBUF_SIZE]; /* pcap err buffer */
> bpf_u_int32 net; /* network of sniffing
> interface */
> bpf_u_int32 mask; /* netmask of sniffing
> interface */
> char netstr[INET_ADDRSTRLEN]; /* network string */
> char maskstr[INET_ADDRSTRLEN]; /* netmask string */
> unsigned int use_dumper : 1; /* flag to use dump file
> */
> unsigned int use_filter : 1; /* flag to compile filter
> */
> char dumpfile[FILESIZE]; /* file to dump packets */
> struct bpf_program fp; /* compiled filter
> expression */
> char filter_exp[FILTER_SIZE]; /* filter expression */
> } ether_dev_t;
>
> typedef struct {
> char *dev; /* the device we will
> monitor */
> pcap_t *dev_handle; /* the handle to this
> device after opening */
> struct pcap_stat ps; /* packet statistics
> struct
> */
> pcap_dumper_t *pd; /* dump file pointer */
> int link_int; /* datalink int of device
> */
> int timeout; /* datalink timeout per
> packet */
> int pcount; /* packets captured */
> const char *link_desc; /* datalink description of
> device */
> const char *link_name; /* datalink name of device
> */
> char errbuf[PCAP_ERRBUF_SIZE]; /* pcap err buffer */
> bpf_u_int32 net; /* network of sniffing
> interface */
> bpf_u_int32 mask; /* netmask of sniffing
> interface */
> char netstr[INET_ADDRSTRLEN]; /* network string */
> char maskstr[INET_ADDRSTRLEN]; /* netmask string */
> int use_dumper; /* flag to use dump file
> */
> int use_filter; /* flag to compile filter
> */
> char dumpfile[FILESIZE]; /* file to dump packets */
> struct bpf_program fp; /* compiled filter
> expression */
> char filter_exp[FILTER_SIZE]; /* filter expression */
> } ether_dev_t;
>
> when I run size on the compiled binary and its stripped version there
> seems to be a difference of 32 bits in the text portion and overall size.
>
> $ size regular_int regular_int.stripped
> text data bss dec hex
> 9023 744 352 10119 2787 regular_int
> 9023 744 352 10119 2787 regular_int.stripped
>
> $ size one_bit_u_int one_bit_u_int.stripped
> text data bss dec hex
> 8991 744 352 10087 2767 one_bit_u_int
> 8991 744 352 10087 2767 one_bit_u_int.stripped
>
>
> However, when I run ls -l on these files the non-stripped
> one bit integer binary is larger, which I did not expect.
>
> $ ls -l one_bit_u_int{,.stripped}
> -rwxr-xr-x 1 user user 29014 Jan 10 16:58 one_bit_u_int
> -rwxr-xr-x 1 user user 12440 Jan 10 16:59 one_bit_u_int.stripped
>
> $ ls -l regular_int{,.stripped}
> -rwxr-xr-x 1 user user 28942 Jan 10 16:57 regular_int
> -rwxr-xr-x 1 user user 12472 Jan 10 16:57 regular_int.stripped
>
> Why is the non-stripped, one bit wide integer binary larger than
> the non-stripped, regular integer binary?
>
> These were compiled on amd64 -current
For regular integers, usually having a size that match the size of processor
registers and places in memory, I expect the compiler to generate very little
code, probably just a single machine instruction, to deal with those
integers.
Maybe the compiler generates extra code to deal with your one bit sized
variables. It may not be able to generate single machine instructions for
dealing with such oddly sized integers.
