forked from Imagelibrary/rtems
SPARC: syscall optimizations and PSR-write fix
The last optimization missed was incorrect in regards to PSR write instruction delay must be 3 instructions. New optimizations: * align to 32-byte cache line. * rearrange code into three "blocks" of 4 instructions that is executed by syscall 2 and 3. This is to optimize for 16/32 byte cache lines. * use delay-slot instruction in trap table to reduce by one instruction. * use the fact that "wr %PSR" implements XOR to reduce by one instruction.
This commit is contained in:
@@ -35,27 +35,27 @@
|
||||
* g3 = additional exit code 2
|
||||
*/
|
||||
|
||||
.align 32 ! Align to 32-byte cache-line
|
||||
PUBLIC(syscall)
|
||||
|
||||
SYM(syscall):
|
||||
|
||||
subcc %g1, 2, %g0 ! syscall 2, disable interrupts
|
||||
bne 3f
|
||||
subcc %g1, 3, %g0 ! syscall 3, enable interrupts
|
||||
or %l0, 0x0f00, %l4 ! set PIL=15
|
||||
ba 9f
|
||||
or %l0, SPARC_PSR_ET_MASK, %i0 ! return old psr with ET=1
|
||||
3:
|
||||
bne 1f
|
||||
! "subcc, %g1, 3, %g0" done in trap table
|
||||
bne 2f ! syscall 3? enable interrupt
|
||||
and %i0, SPARC_PSR_PIL_MASK, %l4
|
||||
andn %l0, SPARC_PSR_PIL_MASK, %l5
|
||||
or %l5, %l4, %l4
|
||||
9: ! leave
|
||||
mov %l4, %psr ! Update PSR according to Syscall 2 or 3
|
||||
wr %l4, %l5, %psr ! Update PSR according to syscall 3
|
||||
1: ! leave, with 3 inst PSR-write delay
|
||||
mov 0, %g1 ! clear %g1
|
||||
jmpl %l2, %g0
|
||||
rett %l2 + 4
|
||||
1:
|
||||
or %l0, SPARC_PSR_ET_MASK, %i0 ! return old psr with ET=1. No
|
||||
! effect on syscall 3
|
||||
jmpl %l2, %g0
|
||||
rett %l2 + 4
|
||||
|
||||
2: or %l0, 0x0f00, %l4 ! set PIL=15
|
||||
subcc %g1, 2, %g0 ! syscall 2? disable interrupts
|
||||
beq,a 1b ! Annul delay-slot for syscall 1
|
||||
mov %l4, %psr ! Update PSR according to Syscall 2
|
||||
ta 0 ! syscall 1 (not 2 or 3), halt
|
||||
|
||||
PUBLIC(sparc_disable_interrupts)
|
||||
|
||||
Reference in New Issue
Block a user