2007-11-02 18:56:58 +00:00
|
|
|
/*
|
|
|
|
* $Source$
|
|
|
|
* $State$
|
|
|
|
*/
|
|
|
|
|
|
|
|
operation
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
: OP { emit4($1); }
|
|
|
|
| OP_BDA bda { emit4($1 | $2); }
|
|
|
|
| OP_BDL bdl { emit4($1 | $2); }
|
|
|
|
| OP_BF_BFA CR ',' CR { emit4($1 | ($2<<23) | ($4<<18)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_BF_FRA_FRB CR ',' FPR ',' FPR { emit4($1 | ($2<<23) | ($4<<16) | ($6<<11)); }
|
|
|
|
| OP_BF_L_RA_RB CR ',' u1 ',' GPR ',' GPR { emit4($1 | ($2<<23) | ($4<<21) | ($6<<16) | ($8<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_BF_L_RA_SI CR ',' u1 ',' GPR ',' e16 { emit_hl($1 | ($2<<23) | ($4<<21) | ($6<<16) | $8); }
|
|
|
|
| OP_BF_L_RA_UI CR ',' u1 ',' GPR ',' e16 { emit_hl($1 | ($2<<23) | ($4<<21) | ($6<<16) | $8); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_BF_RA_RB cr_opt GPR ',' GPR { emit4($1 | ($2<<23) | ($3<<16) | ($5<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_BF_RA_SI cr_opt GPR ',' e16 { emit_hl($1 | ($2<<23) | ($3<<16) | $5); }
|
|
|
|
| OP_BF_RA_UI cr_opt GPR ',' e16 { emit_hl($1 | ($2<<23) | ($3<<16) | $5); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_BF_U_C c CR ',' u4 { emit4($1 | $2 | ($3<<23) | ($5<<12)); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_BH { emit4($1); }
|
|
|
|
| OP_BH u2 { emit4($1 | ($2<<11)); }
|
|
|
|
| OP_BI_BDA u5 ',' bda { emit4($1 | ($2<<16) | $4); }
|
|
|
|
| OP_BI_BDL u5 ',' bdl { emit4($1 | ($2<<16) | $4); }
|
|
|
|
| OP_BI_BH u5 opt_bh { emit4($1 | ($2<<16) | $3); }
|
|
|
|
| OP_BICR_BDA cr_opt bda { emit4($1 | ($2<<18) | $3); }
|
|
|
|
| OP_BICR_BDL cr_opt bdl { emit4($1 | ($2<<18) | $3); }
|
|
|
|
| OP_BICR_BH { emit4($1); }
|
|
|
|
| OP_BICR_BH CR opt_bh { emit4($1 | ($2<<18) | $3); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_BO_BI_BDA u5 ',' u5 ',' bda { emit4($1 | ($2<<21) | ($4<<16) | $6); }
|
|
|
|
| OP_BO_BI_BDL u5 ',' u5 ',' bdl { emit4($1 | ($2<<21) | ($4<<16) | $6); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_BO_BI_BH u5 ',' u5 opt_bh { emit4($1 | ($2<<21) | ($4<<16) | $5); }
|
|
|
|
| OP_BT_BA_BA u5 ',' u5 { emit4($1 | ($2<<21) | ($4<<16) | ($4<<11)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_BT_BA_BB u5 ',' u5 ',' u5 { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_BT_BT_BT u5 { emit4($1 | ($2<<21) | ($2<<16) | ($2<<11)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_BT_C c u5 { emit4($1 | $2 | ($3<<21)); }
|
|
|
|
| OP_FLM_FRB_C c u8 ',' FPR { emit4($1 | $2 | ($3<<17) | ($5<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_FRS_RA_D FPR ',' e16 '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_FRS_RA_RB FPR ',' GPR ',' GPR { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
|
|
|
| OP_FRT_FRA_FRB_C c FPR ',' FPR ',' FPR { emit4($1 | $2 | ($3<<21) | ($5<<16) | ($7<<11)); }
|
2017-01-14 23:15:01 +00:00
|
|
|
| OP_FRT_FRA_FRC_FRB_C c FPR ',' FPR ',' FPR ',' FPR { emit4($1 | $2 | ($3<<21) | ($5<<16) | ($9<<11) | ($7<<6)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_FRT_FRA_FRC_C c FPR ',' FPR ',' FPR { emit4($1 | $2 | ($3<<21) | ($5<<16) | ($7<<6)); }
|
2017-01-14 23:15:01 +00:00
|
|
|
| OP_FRT_FRB_C c FPR ',' FPR { emit4($1 | $2 | ($3<<21) | ($5<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_FRT_RA_D FPR ',' e16 '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_FRT_RA_RB FPR ',' GPR ',' GPR { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
|
|
|
| OP_FRT_C c FPR { emit4($1 | $2 | ($3<<21)); }
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
| OP_L { emit4($1); }
|
|
|
|
| OP_L u2 { emit4($1 | ($2<<21)); }
|
|
|
|
| OP_LEV { emit4($1); }
|
|
|
|
| OP_LEV u7 { emit4($1 | ($2<<5)); }
|
|
|
|
| OP_RA_RB GPR ',' GPR
|
|
|
|
{ emit4($1 | ($2<<16) | ($4<<11)); }
|
|
|
|
| OP_RA_RB_TH GPR ',' GPR opt_bh
|
|
|
|
{ emit4($1 | $5 | ($2<<16) | ($4<<11)); }
|
|
|
|
/*
|
|
|
|
* For instructions with "mnemonic RS, RA, ..."
|
|
|
|
* OP_RA_RS_... swaps RS and RA to (RA<<21) || (RS<<16)
|
|
|
|
* OP_RS_RA_... keeps RS and RA as (RS<<21) || (RA<<16)
|
|
|
|
*/
|
|
|
|
| OP_RA_RS_C c GPR ',' GPR
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16)); }
|
|
|
|
| OP_RA_RS_RA_C c GPR ',' GPR
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | ($5<<11)); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_RA_RS_RB_C c GPR ',' GPR ',' GPR
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | ($7<<11)); }
|
|
|
|
| OP_RA_RS_RB_MB5_ME5_C c GPR ',' GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | ($7<<11) |
|
|
|
|
($9<<6) | ($11<<1)); }
|
|
|
|
| OP_RA_RS_RB_MB6_C c GPR ',' GPR ',' GPR ',' u6
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | ($7<<11) | MB6($9)); }
|
|
|
|
| OP_RA_RS_SH5_C c GPR ',' GPR ',' u5
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | ($7<<11)); }
|
|
|
|
| OP_RA_RS_SH5_MB5_ME5_C c GPR ',' GPR ',' u5 ',' u5 ',' u5
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
($7<<11) | ($9<<6) | ($11<<1)); }
|
|
|
|
| OP_RA_RS_SH6_C c GPR ',' GPR ',' u6
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6($7)); }
|
|
|
|
| OP_RA_RS_SH6_MB6_C c GPR ',' GPR ',' u6 ',' u6
|
|
|
|
{ emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6($7) | MB6($9)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_RA_RS_UI GPR ',' GPR ',' e16 { emit_hl($1 | ($4<<21) | ($2<<16) | $6); }
|
|
|
|
| OP_RA_RS_UI_CC C GPR ',' GPR ',' e16 { emit_hl($1 | ($5<<21) | ($3<<16) | $7); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_RT GPR { emit4($1 | ($2<<21)); }
|
|
|
|
| OP_RT_RA_C c GPR ',' GPR { emit4($1 | $2 | ($3<<21) | ($5<<16)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_RT_RA_D GPR ',' e16 '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
|
|
|
| OP_RT_RA_DS GPR ',' ds '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_RT_RA_NB GPR ',' GPR ',' nb { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
|
|
|
| OP_RT_RA_RB GPR ',' GPR ',' GPR { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_RT_RA_RB_C c GPR ',' GPR ',' GPR { emit4($1 | $2 | ($3<<21) | ($5<<16) | ($7<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_RT_RA_SI GPR ',' GPR ',' e16 { emit_hl($1 | ($2<<21) | ($4<<16) | $6); }
|
|
|
|
| OP_RT_RA_SI_addic c GPR ',' GPR ',' e16 { emit_hl($1 | ($2<<26) | ($3<<21) | ($5<<16) | $7); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_RT_RA_SI_subi GPR ',' GPR ',' negate16 { emit4($1 | ($2<<21) | ($4<<16) | $6); }
|
|
|
|
| OP_RT_RA_SI_subic c GPR ',' GPR ',' negate16 { emit4($1 | ($2<<26) | ($3<<21) | ($5<<16) | $7); }
|
|
|
|
| OP_RT_RB_RA_C c GPR ',' GPR ',' GPR { emit4($1 | $2 | ($3<<21) | ($7<<16) | ($5<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_RT_SI GPR ',' e16 { emit_hl($1 | ($2<<21) | $4); }
|
2016-12-07 22:28:00 +00:00
|
|
|
| OP_RT_SPR GPR ',' spr_num { emit4($1 | ($2<<21) | ($4<<11)); }
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
| OP_RT_TBR GPR opt_tbr { emit4($1 | ($2<<21) | ($3<<11)); }
|
|
|
|
| OP_RS GPR { emit4($1 | ($2<<21)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_RS_FXM u7 ',' GPR { emit4($1 | ($4<<21) | ($2<<12)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_RS_RA_D GPR ',' e16 '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
|
|
|
| OP_RS_RA_DS GPR ',' ds '(' GPR ')' { emit_hl($1 | ($2<<21) | ($6<<16) | $4); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_RS_RA_NB GPR ',' GPR ',' nb { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
|
|
|
| OP_RS_RA_RB GPR ',' GPR ',' GPR { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
| OP_RS_RA_RB_CC C GPR ',' GPR ',' GPR { emit4($1 | ($3<<21) | ($5<<16) | ($7<<11)); }
|
2016-12-07 22:28:00 +00:00
|
|
|
| OP_RS_SPR spr_num ',' GPR { emit4($1 | ($4<<21) | ($2<<11)); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_TO_RA_RB u5 ',' GPR ',' GPR { emit4($1 | ($2<<21) | ($4<<16) | ($6<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_TO_RA_SI u5 ',' GPR ',' e16 { emit_hl($1 | ($2<<21) | ($4<<16) | $6); }
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_TOX_RA_RB GPR ',' GPR { emit4($1 | ($2<<16) | ($4<<11)); }
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_TOX_RA_SI GPR ',' e16 { emit_hl($1 | ($2<<16) | $4); }
|
2007-11-02 18:56:58 +00:00
|
|
|
| OP_LIA lia { emit4($1 | $2); }
|
|
|
|
| OP_LIL lil { emit4($1 | $2); }
|
2016-09-18 21:03:23 +00:00
|
|
|
| OP_LI32 li32 /* emitted in subrule */
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
| OP_clrlsldi c GPR ',' GPR ',' u6 ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t mb = ($7 - $9) & 0x3f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($9 <= $7);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6($9) | MB6(mb));
|
|
|
|
}
|
|
|
|
| OP_clrldi c GPR ',' GPR ',' u6
|
|
|
|
{
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6(0) | MB6($7));
|
|
|
|
}
|
|
|
|
| OP_clrrdi c GPR ',' GPR ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = 63 - $7;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6(0) | MB6(me));
|
|
|
|
}
|
|
|
|
| OP_extldi c GPR ',' GPR ',' u6 ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = ($7 - 1) & 0x3f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6($9) | MB6(me));
|
|
|
|
}
|
|
|
|
| OP_extrdi c GPR ',' GPR ',' u6 ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = ($9 + $7) & 0x3f;
|
|
|
|
word_t mb = (64 - $7) & 0x3f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6(sh) | MB6(mb));
|
|
|
|
}
|
|
|
|
| OP_rotrdi c GPR ',' GPR ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (64 - $7) & 0x3f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6(sh) | MB6(0));
|
|
|
|
}
|
|
|
|
| OP_sldi c GPR ',' GPR ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = 63 - $7;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6($7) | MB6(me));
|
|
|
|
}
|
|
|
|
| OP_srdi c GPR ',' GPR ',' u6
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (64 - $7) & 0x3f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) | SH6(sh) | MB6($7));
|
|
|
|
}
|
|
|
|
| OP_clrlslwi c GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t mb = ($7 - $9) & 0x1f;
|
|
|
|
word_t me = 31 - $9;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($9 <= $7);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
($9<<11) | (mb<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_clrlwi c GPR ',' GPR ',' u5
|
|
|
|
{
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(0<<11) | ($7<<6) | (31<<1));
|
|
|
|
}
|
|
|
|
| OP_clrrwi c GPR ',' GPR ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = 31 - $7;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(0<<11) | (0<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_extlwi c GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = ($7 - 1) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
($9<<11) | (0<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_extrwi c GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = ($9 + $7) & 0x1f;
|
|
|
|
word_t mb = (32 - $7) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(sh<<11) | (mb<<6) | (31<<1));
|
|
|
|
}
|
|
|
|
| OP_inslwi c GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (32 - $9) & 0x1f;
|
|
|
|
word_t me = ($9 + $7 - 1) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(sh<<11) | ($9<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_insrwi c GPR ',' GPR ',' u5 ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (32 - $9 - $7) & 0x1f;
|
|
|
|
word_t me = ($9 + $7 - 1) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
fit($7 > 0);
|
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(sh<<11) | ($9<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_rotrwi c GPR ',' GPR ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (32 - $7) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(sh<<11) | (0<<6) | (31<<1));
|
|
|
|
}
|
|
|
|
| OP_slwi c GPR ',' GPR ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t me = 31 - $7;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
($7<<11) | (0<<6) | (me<<1));
|
|
|
|
}
|
|
|
|
| OP_srwi c GPR ',' GPR ',' u5
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t sh = (32 - $7) & 0x1f;
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
emit4($1 | $2 | ($5<<21) | ($3<<16) |
|
|
|
|
(sh<<11) | ($7<<6) | (31<<1));
|
|
|
|
}
|
2007-11-02 18:56:58 +00:00
|
|
|
;
|
|
|
|
|
|
|
|
c
|
|
|
|
: /* nothing */ { $$ = 0; }
|
|
|
|
| C { $$ = 1; }
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
e16
|
2016-09-17 10:43:15 +00:00
|
|
|
: absexp
|
2007-11-02 18:56:58 +00:00
|
|
|
{
|
2016-09-17 10:43:15 +00:00
|
|
|
/* Allow signed or unsigned 16-bit values. */
|
|
|
|
if (($1 < -0x8000) || ($1 > 0xffff))
|
|
|
|
serror("16-bit value out of range");
|
|
|
|
$$ = (uint16_t) $1;
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
no_hl();
|
2007-11-02 18:56:58 +00:00
|
|
|
}
|
Add RELOLIS for PowerPC lis with ha16 or hi16.
The new relocation type RELOLIS handles these instructions:
lis RT, ha16[expr] == addis RT, r0, ha16[expr]
lis RT, hi16[expr] == addis RT, r0, hi16[expr]
RELOLIS stores a 32-bit value in the program text. In this value, the
high bit is a ha16 flag, the next 5 bits are the target register RT,
and the low bits are a signed 26-bit offset. The linker replaces this
value with the lis instruction.
The old RELOPPC relocated a ha16/lo16 or hi16/lo16 pair. The new
RELOLIS relocates only a ha16 or hi16, so it is no longer necessary to
have a matching lo16 in the next instruction. The disadvantage is
that RELOLIS has only a signed 26-bit offset, not a 32-bit offset.
Switch the assembler to use RELOLIS for ha16 or hi16 and RELO2 for
lo16. The li32 instruction still uses the old RELOPPC relocation.
This is not the same as my RELOPPC change from my recent mail to
tack-devel (https://sourceforge.net/p/tack/mailman/message/35651528/).
This commit is on a different branch. Here I am throwing away my
RELOPPC change and instead trying RELOLIS.
2017-02-08 16:46:31 +00:00
|
|
|
| OP_HI ASC_LPAR expr ASC_RPAR { $$ = eval_hl(&$3, OP_HI); }
|
|
|
|
| OP_HA ASC_LPAR expr ASC_RPAR { $$ = eval_hl(&$3, OP_HA); }
|
|
|
|
| OP_LO ASC_LPAR expr ASC_RPAR { $$ = eval_hl(&$3, OP_LO); }
|
2007-11-02 18:56:58 +00:00
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
negate16
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
/* To encode subi, we negate the immediate value, then
|
|
|
|
* it must fit as signed 16-bit. */
|
|
|
|
$$ = -$1;
|
|
|
|
fit(fitx($$, 16));
|
|
|
|
$$ = (uint16_t) $$;
|
|
|
|
}
|
|
|
|
;
|
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u8
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0xFF))
|
|
|
|
serror("8-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u7
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0x7F))
|
|
|
|
serror("7-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u6
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0x3F))
|
|
|
|
serror("6-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u5
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0x1F))
|
|
|
|
serror("5-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u4
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0xF))
|
|
|
|
serror("4-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u1
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 1))
|
|
|
|
serror("1-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
u2
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0x3))
|
|
|
|
serror("2-bit unsigned value out of range");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
/* Optional comma, branch hint (or touch hint). */
|
Teach the assembler about PowerPC extended mnemonics.
Also make a few changes to basic mnemonics. Fix typo in name of the
basic "creqv". Add the basic "addc" and relatives, because it would
be odd to have the extended "subc" without "addc". Fix the basic
"rldicl", "rldicr", "rldic", "rldimi" to correctly encode the 6-bit MB
field. Fix "slw" and relatives to correctly swap their RA and RS
operands.
Add many, but not all, of the extended mnemonics from IBM's Power ISA
Version 2.06 Book I Appendix E. (I used 2.06, published 2009, just
because I already had the PDF of it.) This commit includes mnemonics
for branching, subtraction, traps, bit rotation, and a few others,
like "mflr" and "nop". The assembler now understands branches like
`beq cr7, label` and bit shifts like `slwi r7, r7, 2`. These encode
the same machine instructions as the basic "bc" and "rlwinm".
Some operands to basic names become optional. The assembler no longer
requires the level in "sc" or the branch hint in "bcctr" and "bclr";
they default to zero. Some extended names take an optional branch
hint or condition register.
Some extended names are still missing. I don't provide names with
static branch prediction, like "beq+" or "bge-", because the assembler
parses '+' and '-' as operators, not as part of an instruction name.
I also don't provide some names that 2.06 has for moving to or from
the condition register or some special purpose registers, names like
"mtcr" or "mfuamr".
This commit also deletes some unused tokens and one unused yacc rule.
2017-01-22 04:49:29 +00:00
|
|
|
opt_bh
|
|
|
|
: /* nothing */ { $$ = 0; }
|
|
|
|
| ',' u2 { $$ = ($2<<11); }
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Optional condition register, comma. This checks if the token is a
|
|
|
|
* CR register name. This wouldn't work if we allowed CR as a number.
|
|
|
|
*/
|
|
|
|
cr_opt
|
|
|
|
: /* nothing */ { $$ = 0; }
|
|
|
|
| CR ',' { $$ = $1; }
|
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
ds
|
|
|
|
: e16
|
|
|
|
{
|
|
|
|
if ($1 & 3)
|
|
|
|
serror("value must be 4-aligned");
|
|
|
|
$$ = $1;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
nb
|
|
|
|
: absexp
|
|
|
|
{
|
|
|
|
if (($1 < 1) || ($1 > 32))
|
|
|
|
serror("register count must be in the range 1..32");
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
if ($1 == 32)
|
|
|
|
$$ = 0;
|
|
|
|
else
|
2017-01-14 23:15:01 +00:00
|
|
|
$$ = $1;
|
2007-11-02 18:56:58 +00:00
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
bdl
|
|
|
|
: expr
|
|
|
|
{
|
|
|
|
int dist = $1.val - DOTVAL;
|
|
|
|
fit(fitx(dist, 25));
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
if (dist & 0x3)
|
|
|
|
serror("jump targets must be 4-aligned");
|
|
|
|
|
|
|
|
DOTVAL += 2;
|
|
|
|
newrelo($1.typ, RELO2 | RELPC | FIXUPFLAGS);
|
|
|
|
DOTVAL -= 2;
|
|
|
|
$$ = dist & 0xFFFD;
|
|
|
|
}
|
|
|
|
;
|
|
|
|
|
|
|
|
bda
|
|
|
|
: expr
|
|
|
|
{
|
|
|
|
int target = $1.val;
|
|
|
|
fit(fitx(target, 16));
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
if (target & 0x3)
|
|
|
|
serror("jump targets must be 4-aligned");
|
|
|
|
|
|
|
|
DOTVAL += 2;
|
|
|
|
newrelo($1.typ, RELO2 | FIXUPFLAGS);
|
|
|
|
DOTVAL -= 2;
|
|
|
|
$$ = target & 0xFFFD;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2016-09-18 21:03:23 +00:00
|
|
|
li32
|
2016-09-17 10:43:15 +00:00
|
|
|
: GPR ',' expr
|
|
|
|
{
|
2017-01-30 21:15:02 +00:00
|
|
|
word_t type = $3.typ & S_TYP;
|
|
|
|
word_t val = $3.val;
|
2016-10-17 22:21:32 +00:00
|
|
|
if ((type == S_ABS) && (val <= 0xffff))
|
|
|
|
emit4((14<<26) | ($1<<21) | (0<<16) | val); /* addi */
|
|
|
|
else
|
|
|
|
{
|
|
|
|
newrelo($3.typ, RELOPPC | FIXUPFLAGS);
|
|
|
|
emit4((15<<26) | ($1<<21) | (0<<16) | (val >> 16)); /* addis */
|
|
|
|
emit4((24<<26) | ($1<<21) | ($1<<16) | (val & 0xffff)); /* ori */
|
|
|
|
}
|
2016-09-17 10:43:15 +00:00
|
|
|
}
|
|
|
|
;
|
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
lil
|
|
|
|
: expr
|
|
|
|
{
|
|
|
|
int dist = $1.val - DOTVAL;
|
|
|
|
fit(fitx(dist, 26));
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
if (dist & 0x3)
|
|
|
|
serror("jump targets must be 4-aligned");
|
|
|
|
|
|
|
|
newrelo($1.typ, RELOPPC | RELPC | FIXUPFLAGS);
|
|
|
|
$$ = dist & 0x03FFFFFD;
|
|
|
|
}
|
|
|
|
;
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
lia
|
|
|
|
: expr
|
|
|
|
{
|
|
|
|
int target = $1.val;
|
|
|
|
fit(fitx(target, 26));
|
2017-01-14 23:15:01 +00:00
|
|
|
|
2007-11-02 18:56:58 +00:00
|
|
|
if (target & 0x3)
|
|
|
|
serror("jump targets must be 4-aligned");
|
|
|
|
|
|
|
|
newrelo($1.typ, RELOPPC | FIXUPFLAGS);
|
|
|
|
$$ = target & 0x03FFFFFD;
|
|
|
|
}
|
|
|
|
;
|
2016-12-07 22:28:00 +00:00
|
|
|
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
/*
|
|
|
|
* Instructions "mfspr", "mtspr", and "mftb" encode the 10-bit special
|
|
|
|
* purpose register (spr) or time base register (tbr) by swapping the
|
|
|
|
* low 5 bits with the high 5 bits. The value from an SPR token has
|
|
|
|
* already been swapped.
|
|
|
|
*/
|
|
|
|
|
2016-12-07 22:28:00 +00:00
|
|
|
spr_num
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
: SPR { $$ = $1; }
|
|
|
|
| tbr_num { $$ = $1; }
|
|
|
|
;
|
|
|
|
|
|
|
|
opt_tbr
|
|
|
|
: /* nothing */ { $$ = 8 | (12<<5); }
|
|
|
|
| ',' tbr_num { $$ = $2; }
|
|
|
|
;
|
|
|
|
|
|
|
|
tbr_num
|
|
|
|
: absexp
|
2016-12-07 22:28:00 +00:00
|
|
|
{
|
|
|
|
if (($1 < 0) || ($1 > 0x3ff))
|
Add instructions like "lwarx". Extend manual.
Add more page numbers from PowerPC version 2.01. Remove "xnop" not in
2.01, add "mtcr" from 2.01. Add "lwarx" and the other instructions
from Book II. I did not try all the newly added instructions, but
these seem to work: dcbt, dcbtst, icibi, isync, lwarx, stwcx., mftb,
mftbu
In man/powerpc_as.6 (not installed), add a summary of the registers
and addressing modes (like in i386_as.6), describe short forms, update
description of hi16/ha16, add CAVEATS about instructions that some
processors can't run.
2018-03-07 18:37:31 +00:00
|
|
|
serror("10-bit unsigned value out of range");
|
2016-12-07 22:28:00 +00:00
|
|
|
$$ = ($1 >> 5) | (($1 & 0x1f) << 5);
|
|
|
|
}
|
|
|
|
;
|