freebsd-dev/sys/contrib/octeon-sdk/cvmx-log-arc.S
Juli Mallett 219d14fe5f Import the Cavium Simple Executive from the Cavium Octeon SDK. The Simple
Executive is a library that can be used by standalone applications and kernels
to abstract access to Octeon SoC and board-specific hardware and facilities.
The FreeBSD port to Octeon will be updated to use this where possible.
2010-07-20 07:19:43 +00:00

177 lines
5.5 KiB
ArmAsm

/***********************license start***************
* Copyright (c) 2003-2008 Cavium Networks (support@cavium.com). All rights
* reserved.
*
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials provided
* with the distribution.
*
* * Neither the name of Cavium Networks nor the names of
* its contributors may be used to endorse or promote products
* derived from this software without specific prior written
* permission.
*
* TO THE MAXIMUM EXTENT PERMITTED BY LAW, THE SOFTWARE IS PROVIDED "AS IS"
* AND WITH ALL FAULTS AND CAVIUM NETWORKS MAKES NO PROMISES, REPRESENTATIONS
* OR WARRANTIES, EITHER EXPRESS, IMPLIED, STATUTORY, OR OTHERWISE, WITH
* RESPECT TO THE SOFTWARE, INCLUDING ITS CONDITION, ITS CONFORMITY TO ANY
* REPRESENTATION OR DESCRIPTION, OR THE EXISTENCE OF ANY LATENT OR PATENT
* DEFECTS, AND CAVIUM SPECIFICALLY DISCLAIMS ALL IMPLIED (IF ANY) WARRANTIES
* OF TITLE, MERCHANTABILITY, NONINFRINGEMENT, FITNESS FOR A PARTICULAR
* PURPOSE, LACK OF VIRUSES, ACCURACY OR COMPLETENESS, QUIET ENJOYMENT, QUIET
* POSSESSION OR CORRESPONDENCE TO DESCRIPTION. THE ENTIRE RISK ARISING OUT
* OF USE OR PERFORMANCE OF THE SOFTWARE LIES WITH YOU.
*
*
* For any questions regarding licensing please contact marketing@caviumnetworks.com
*
***********************license end**************************************/
//
// The function defined here is called for every function as it is executed.
// These calls are automatically inserted by GCC when the switch "-pg" is
// used. This allows cvmx-log to add a PC entry as each function is executed.
// This information, along with the timestamps can give the user a good idea
// of the performance characteristics of their program. This function normally
// takes about 22 cycles to execute.
//
#ifdef __linux__
#include <asm/asm.h>
#include <asm/regdef.h>
#define LA dla
#else
#include <machine/asm.h>
#include <machine/regdef.h>
#define LA la
#endif
.set noreorder
.set noat
LEAF(_mcount)
//
// All registers we use must be saved since calls are added by gcc
// after register allocation. The at register ($3) will contain the
// original ra register before the _mcount call. Also the compiler
// automatically performs a "dsubu sp, sp, 16" before we're called.
// At the end of this function all registers must have their original
// values and the stack pointr must be adjusted by 16. This code is
// pretty unreadable since it has been arranged to promote dual issue.
//
#ifdef __linux__
dsubu sp, sp, 32
#else
dsubu sp, sp, 16
#endif
sd s3, 24(sp) // Save register
rdhwr s3, $31 // Read the cycle count
sd s0, 0(sp) // Save register
LA s0, cvmx_log_buffer_end_ptr // Load the address of the end of the log buffer
sd s1, 8(sp) // Save register
LA s1, cvmx_log_buffer_write_ptr // Load the address of the location in the log buffer
sd s2, 16(sp) // Save register
ld s0, 0(s0) // Get value of the current log buffer end location
ld s2, 0(s1) // Get value of the current log buffer location
dsubu s0, s0, s2 // Subtract the end pointer and the write pointer
sltiu s0, s0, 16 // Check if there are at least 16 bytes
bne s0, $0, call_c_pc // Call the slow C function if we don't have room in the log
li s0, 0x001 // 11 bit constant that matches the first 11 bits of a CVMX_LOG_TYPE_PC header
sd ra, 8(s2) // Write the pc to the log
dins s3, s0, 53, 11 // Overwrite the upper cycle count bits with the CVMX_LOG_TYPE_PC header
sd s3, 0(s2) // Write the log header
daddu s2, s2, 16 // Increment the write location ptr
sd s2, 0(s1) // Store the write location ptr
return_c_pc:
ld s0, 0(sp) // Restore register
ld s1, 8(sp) // Restore register
ld s2, 16(sp) // Restore register
ld s3, 24(sp) // Restore register
daddu sp, sp, 32 // Pop everything off the stack, even the 16 bytes done by gcc
jr ra // Return to the caller and
or ra, $1, $1 // make sure the ra is back to its original value
call_c_pc:
// The registers used by the C code may change based on optimizations. To be
// safe, I'll save all registers. We're in the slow path case anyway.
dsubu sp, sp, 216
sd $1, 0(sp)
sd $2, 8(sp)
sd $3, 16(sp)
sd $4, 24(sp)
sd $5, 32(sp)
sd $6, 40(sp)
sd $7, 48(sp)
sd $8, 56(sp)
sd $9, 64(sp)
sd $10, 72(sp)
sd $11, 80(sp)
sd $12, 88(sp)
sd $13, 96(sp)
sd $14, 104(sp)
sd $15, 112(sp)
// s0, s1, s2, s3 are already saved
sd $20, 120(sp)
sd $21, 128(sp)
sd $22, 136(sp)
sd $23, 144(sp)
sd $24, 152(sp)
sd $25, 160(sp)
sd $26, 168(sp)
sd $27, 176(sp)
sd $28, 184(sp)
sd $29, 192(sp)
sd $30, 200(sp)
sd $31, 208(sp)
or a0, ra, ra
jal cvmx_log_pc
nop
ld $1, 0(sp)
ld $2, 8(sp)
ld $3, 16(sp)
ld $4, 24(sp)
ld $5, 32(sp)
ld $6, 40(sp)
ld $7, 48(sp)
ld $8, 56(sp)
ld $9, 64(sp)
ld $10, 72(sp)
ld $11, 80(sp)
ld $12, 88(sp)
ld $13, 96(sp)
ld $14, 104(sp)
ld $15, 112(sp)
// s0, s1, s2, s3 will be restored later
ld $20, 120(sp)
ld $21, 128(sp)
ld $22, 136(sp)
ld $23, 144(sp)
ld $24, 152(sp)
ld $25, 160(sp)
ld $26, 168(sp)
ld $27, 176(sp)
ld $28, 184(sp)
ld $29, 192(sp)
ld $30, 200(sp)
ld $31, 208(sp)
b return_c_pc
daddu sp, sp, 216
END(_mcount)