powerpc/lib/sstep: Add cmpb instruction emulation
PATCH v3 1/5 powerpc/lib/sstep: Add cmpb instruction emulation
This patch adds emulation of the cmpb instruction, enabling xmon to emulate this instruction.
Tested for correctness against the cmpb asm instruction on ppc64le.
---
v2:
- fixed opcode
- fixed mask typecasting
--- arch/powerpc/lib/sstep.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)
arch/powerpc/lib/sstep.c +20 -0
@@ -596,6 +596,22 @@ static nokprobe_inline void do_cmp_unsigned(struct pt_regs *regs, unsigned long regs->ccr = (regs->ccr & ~(0xf << shift)) | (crval << shift); } +static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, +unsigned long v2,int rd) +{ +unsigned long long out_val, mask; +int i; + + out_val =0 ; +for (i =0 ; i <8 ; i++) { + mask =0xffUL << (i *8 ); +if ((v1 & mask) == (v2 & mask)) + out_val |= mask; + } + + regs->gpr[rd] = out_val; +} + static nokprobe_inline int trap_compare(long v1, long v2) {int ret =0 ;
@@ -1049,6 +1065,10 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,do_cmp_unsigned (regs, val, val2, rd >>2 ); goto instr_done; +case 508 :/* cmpb */ +do_cmpb (regs, regs->gpr[rd], regs->gpr[rb], ra); + goto instr_done; +/* * Arithmetic instructions */
Signature · 1 line
2.9.3
PATCH v3 2/5 powerpc/lib/sstep: Add popcnt instruction emulation
This adds emulations for the popcntb, popcntw, and popcntd instructions. Tested for correctness against the popcnt{b,w,d} instructions on ppc64le.
---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed typecasting
- fixed bitshifting error for both 32 and 64bit arch
--- arch/powerpc/lib/sstep.c | 40 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 39 insertions(+), 1 deletion(-)
arch/powerpc/lib/sstep.c +39 -1
@@ -612,6 +612,32 @@ static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, regs->gpr[rd] = out_val; } +/* +* The size parameter is used to adjust the equivalent popcnt instruction. +* popcntb = 8, popcntw = 32, popcntd = 64 +*/ +static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long v1, +int size,int ra) +{ +unsigned long long out = v1; + + out = (0x5555555555555555 & out) + (0x5555555555555555 & (out >>1 )); + out = (0x3333333333333333 & out) + (0x3333333333333333 & (out >>2 )); + out = (0x0f0f0f0f0f0f0f0f & out) + (0x0f0f0f0f0f0f0f0f & (out >>4 )); +if (size ==8 ) {/* popcntb */ + regs->gpr[ra] = out; +return ; + } + out = (0x001f001f001f001f & out) + (0x001f001f001f001f & (out >>8 )); + out = (0x0000003f0000003f & out) + (0x0000003f0000003f & (out >>16 )); +if (size ==32 ) {/* popcntw */ + regs->gpr[ra] = out; +return ; + } + out = (0x000000000000007f & out) + (0x000000000000007f & (out >>32 )); + regs->gpr[ra] = out;/* popcntd */ +} + static nokprobe_inline int trap_compare(long v1, long v2) {int ret =0 ;
@@ -1194,6 +1220,10 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs, regs->gpr[ra] = regs->gpr[rd] & ~regs->gpr[rb]; goto logical_done; +case 122 :/* popcntb */ +do_popcnt (regs, regs->gpr[rd],8 , ra); + goto logical_done; +case 124 :/* nor */ regs->gpr[ra] = ~(regs->gpr[rd] | regs->gpr[rb]); goto logical_done;
@@ -1206,6 +1236,10 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs, regs->gpr[ra] = regs->gpr[rd] ^ regs->gpr[rb]; goto logical_done; +case 378 :/* popcntw */ +do_popcnt (regs, regs->gpr[rd],32 , ra); + goto logical_done; +case 412 :/* orc */ regs->gpr[ra] = regs->gpr[rd] | ~regs->gpr[rb]; goto logical_done;
@@ -1217,7 +1251,11 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,case 476 :/* nand */ regs->gpr[ra] = ~(regs->gpr[rd] & regs->gpr[rb]); goto logical_done; - +#ifdef __powerpc64__ +case 506 :/* popcntd */ +do_popcnt (regs, regs->gpr[rd],64 , ra); + goto logical_done; +#endif case 922 :/* extsh */ regs->gpr[ra] = (signed short ) regs->gpr[rd]; goto logical_done;
Signature · 1 line
2.9.3
PATCH v3 3/5 powerpc/lib/sstep: Add bpermd instruction emulation
In reply to Matt BrownThis adds emulation for the bpermd instruction.
Tested for correctness against the bpermd instruction on ppc64le.
---
v2:
- fixed opcode
- added ifdef tags to do_bpermd func
- fixed bitshifting errors
--- arch/powerpc/lib/sstep.c | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-)
arch/powerpc/lib/sstep.c +23 -1
@@ -638,6 +638,24 @@ static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long v1, regs->gpr[ra] = out;/* popcntd */ } +#ifdef __powerpc64__ +static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, +unsigned long v2,int ra) +{ +unsigned char perm, idx; +unsigned int i; + + perm =0 ; +for (i =0 ; i <8 ; i++) { + idx = (v1 >> (i *8 )) &0xff ; +if (idx <64 ) +if (v2 &PPC_BIT (idx)) + perm |=1 << i; + } + regs->gpr[ra] = perm; +} +#endif /* __powerpc64__ */ + static nokprobe_inline int trap_compare(long v1, long v2) {int ret =0 ;
@@ -1227,7 +1245,11 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,case 124 :/* nor */ regs->gpr[ra] = ~(regs->gpr[rd] | regs->gpr[rb]); goto logical_done; - +#ifdef __powerpc64__ +case 252 :/* bpermd */ +do_bpermd (regs, regs->gpr[rd], regs->gpr[rb], ra); + goto logical_done; +#endif case 284 :/* xor */ regs->gpr[ra] = ~(regs->gpr[rd] ^ regs->gpr[rb]); goto logical_done;
Signature · 1 line
2.9.3
PATCH v3 4/5 powerpc/lib/sstep: Add prty instruction emulation
In reply to Matt BrownThis adds emulation for the prtyw and prtyd instructions. Tested for logical correctness against the prtyw and prtyd instructions on ppc64le.
---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed bitshifting and typecast errors
- merged do_prtyw and do_prtyd into single function
--- arch/powerpc/lib/sstep.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
arch/powerpc/lib/sstep.c +27 -0
@@ -655,6 +655,25 @@ static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, regs->gpr[ra] = perm; }#endif /* __powerpc64__ */ +/* +* The size parameter adjusts the equivalent prty instruction. +* prtyw = 32, prtyd = 64 +*/ +static nokprobe_inline void do_prty(struct pt_regs *regs, unsigned long v, +int size,int ra) +{ +unsigned long long res = v; + + res = (0x0001000100010001 & res) + (0x0001000100010001 & (res >>8 )); + res = (0x0000000700000007 & res) + (0x0000000700000007 & (res >>16 )); +if (size ==32 ) {/* prtyw */ + regs->gpr[ra] = (0x0000000100000001 & res); +return ; + } + + res = (0x000000000000000f & res) + (0x000000000000000f & (res >>32 )); + regs->gpr[ra] = res &1 ;/*prtyd */ +} static nokprobe_inline int trap_compare(long v1, long v2) {
@@ -1245,6 +1264,14 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,case 124 :/* nor */ regs->gpr[ra] = ~(regs->gpr[rd] | regs->gpr[rb]); goto logical_done; + +case 154 :/* prtyw */ +do_prty (regs, regs->gpr[rd],32 , ra); + goto logical_done; + +case 186 :/* prtyd */ +do_prty (regs, regs->gpr[rd],64 , ra); + goto logical_done;#ifdef __powerpc64__case 252 :/* bpermd */ do_bpermd (regs, regs->gpr[rd], regs->gpr[rb], ra);
Signature · 1 line
2.9.3
PATCH v3 5/5 powerpc/lib/sstep: Add isel instruction emulation
In reply to Matt BrownThis adds emulation for the isel instruction.
Tested for correctness against the isel instruction and its extended mnemonics (lt, gt, eq) on ppc64le.
---
v2:
- fixed opcode
- fixed definition to include the 'if RA=0, a=0' clause
- fixed ccr bitshifting error
--- arch/powerpc/lib/sstep.c | 11 +++++++++++ 1 file changed, 11 insertions(+)
arch/powerpc/lib/sstep.c +11 -0
@@ -1239,6 +1239,17 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,/* * Logical instructions */ +case 15 :/* isel */ + mb = (instr >>6 ) &0x1f ;/* bc */ + val = (regs->ccr >> (31 - mb)) &1 ; + val2 = (ra) ? regs->gpr[ra] :0 ; + +if (val) + regs->gpr[rd] = val2; +else + regs->gpr[rd] = regs->gpr[rb]; + goto logical_done; +case 26 :/* cntlzw */ asm("cntlzw %0,%1" :"=r" (regs->gpr[ra]) :"r" (regs->gpr[rd]));
Signature · 1 line
2.9.3
Re: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulation
In reply to Matt BrownQuoted 30 lines · Matt Brown
This adds emulations for the popcntb, popcntw, and popcntd instructions. Tested for correctness against the popcnt{b,w,d} instructions on ppc64le.
Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed typecasting
- fixed bitshifting error for both 32 and 64bit arch--- arch/powerpc/lib/sstep.c | 40 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 39 insertions(+), 1 deletion(-)arch/powerpc/lib/sstep.c +7
-0@@ -612,6 +612,32 @@ static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, regs->gpr[rd] = out_val; } +/* +* The size parameter is used to adjust the equivalent popcnt instruction. +* popcntb = 8, popcntw = 32, popcntd = 64 +*/ +static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long v1, +int size,int ra) +{
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 631
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+unsigned long long out = v1; + + out = (0x5555555555555555 & out) + (0x5555555555555555 & (out >>1 )); + out = (0x3333333333333333 & out) + (0x3333333333333333 & (out >>2 )); + out = (0x0f0f0f0f0f0f0f0f & out) + (0x0f0f0f0f0f0f0f0f & (out >>4 )); +if (size ==8 ) {/* popcntb */ + regs->gpr[ra] = out; +return ; + } + out = (0x001f001f001f001f & out) + (0x001f001f001f001f & (out >>8 ));
Why are we using 0x001f001f here? Now that we've got things in the bytes with 0's prefixing, we can directly use
out = out + out >> 8
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 632
Quoted 1 line
arch/powerpc/lib/sstep.c +1
-0+ out = (0x0000003f0000003f & out) + (0x0000003f0000003f & (out >>16 ));
Same as above
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 638
Quoted 6 lines
arch/powerpc/lib/sstep.c +6
-0+if (size ==32 ) {/* popcntw */ + regs->gpr[ra] = out; +return ; + } + out = (0x000000000000007f & out) + (0x000000000000007f & (out >>32 )); + regs->gpr[ra] = out;/* popcntd */
Ditto
Otherwise looks good!
Balbir Singh.
Re: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
In reply to Matt BrownQuoted 32 lines · Matt Brown
This adds emulation for the prtyw and prtyd instructions. Tested for logical correctness against the prtyw and prtyd instructions on ppc64le.
Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed bitshifting and typecast errors
- mergeddo_prtywanddo_prtydinto single function--- arch/powerpc/lib/sstep.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)arch/powerpc/lib/sstep.c +8
-0@@ -655,6 +655,25 @@ static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, regs->gpr[ra] = perm; }#endif /* __powerpc64__ */ +/* +* The size parameter adjusts the equivalent prty instruction. +* prtyw = 32, prtyd = 64 +*/ +static nokprobe_inline void do_prty(struct pt_regs *regs, unsigned long v, +int size,int ra) +{ +unsigned long long res = v;
arch/powerpc/lib/sstep.c static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, · line 675
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+ + res = (0x0001000100010001 & res) + (0x0001000100010001 & (res >>8 )); + res = (0x0000000700000007 & res) + (0x0000000700000007 & (res >>16 )); +if (size ==32 ) {/* prtyw */ + regs->gpr[ra] = (0x0000000100000001 & res); +return ; + } + + res = (0x000000000000000f & res) + (0x000000000000000f & (res >>32 )); + regs->gpr[ra] = res &1 ;/*prtyd */
Looks good, you can also xor instead of adding, but the masks would need to change in that case.
Balbir Singh.
RE: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulation
In reply to Matt BrownFrom: Linuxppc-dev [mailto:linuxppc-dev-bounces+david.laight=3Daculab.com@l= ists.ozlabs.org] On Behalf Of
Quoted 4 lines
Matt Brown
Sent: 25 July 2017 04:33
To: [email protected]
Subject: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulati=
on
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *=
Quoted 21 lines
=20
This adds emulations for the popcntb, popcntw, and popcntd instructions. Tested for correctness against the popcnt{b,w,d} instructions on ppc64le. =20Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed typecasting
- fixed bitshifting error for both 32 and 64bit arch--- arch/powerpc/lib/sstep.c | 40 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 39 insertions(+), 1 deletion(-)=20
arch/powerpc/lib/sstep.c +0
-0@@ -612,6 +612,32 @@ static nokprobe_inline void do_cmpb(struct pt_regs *=
regs, unsigned long v1,
Quoted 5 lines
arch/powerpc/lib/sstep.c +0
-0regs->gpr[rd] =3D out_val; }=20
+/*
+ * The size parameter is used to adjust the equivalent popcnt instructio=
n.
Quoted 3 lines
+ * popcntb =3D 8, popcntw =3D 32, popcntd =3D 64
+ */
+staticnokprobe_inlinevoid do_popcnt(structpt_regs*regs, unsigned lon=
g v1,
arch/powerpc/lib/sstep.c
Quoted 9 lines
arch/powerpc/lib/sstep.c +9
-0+ int size, int ra) +{ + unsigned long long out =3D v1; + + out =3D (0x5555555555555555 & out) + (0x5555555555555555 & (out >>1 )); + out =3D (0x3333333333333333 & out) + (0x3333333333333333 & (out >>2 )); + out =3D (0x0f0f0f0f0f0f0f0f & out) + (0x0f0f0f0f0f0f0f0f & (out >>4 )); + if (size =3D=3D 8) { /* popcntb */ + regs->gpr[ra] =3D out;
I'm pretty sure you need to mask the result with 7.
David
Re: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulation
Quoted 29 lines · David Laight
From: Linuxppc-dev [mailto:[email protected]] On Behalf Of
Quoted 28 lines
Matt Brown
Sent: 25 July 2017 04:33
To: [email protected]
Subject: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulationThis adds emulations for the popcntb, popcntw, and popcntd instructions. Tested for correctness against the popcnt{b,w,d} instructions on ppc64le.
Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed typecasting
- fixed bitshifting error for both 32 and 64bit arch--- arch/powerpc/lib/sstep.c | 40 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 39 insertions(+), 1 deletion(-)arch/powerpc/lib/sstep.c +0
-0@@ -612,6 +612,32 @@ static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, regs->gpr[rd] = out_val; }
Quoted 17 lines
Quoted 4 lines
arch/powerpc/lib/sstep.c +4
-0+/* +* The size parameter is used to adjust the equivalent popcnt instruction. +* popcntb = 8, popcntw = 32, popcntd = 64 +*/ arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 628
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long v1, +int size,int ra) +{ +unsigned long long out = v1; + + out = (0x5555555555555555 & out) + (0x5555555555555555 & (out >>1 )); + out = (0x3333333333333333 & out) + (0x3333333333333333 & (out >>2 )); + out = (0x0f0f0f0f0f0f0f0f & out) + (0x0f0f0f0f0f0f0f0f & (out >>4 )); +if (size ==8 ) {/* popcntb */ + regs->gpr[ra] = out;I'm pretty sure you need to mask the result with 7.
Absolutely! Good catch!
Balbir Singh.
Re: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
In reply to Matt BrownQuoted 5 lines · Matt Brown
arch/powerpc/lib/sstep.c +5
-0+static nokprobe_inline void do_prty(struct pt_regs *regs, unsigned long v, +int size,int ra) +{ +unsigned long long res = v; +
arch/powerpc/lib/sstep.c static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, · line 676
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+ res = (0x0001000100010001 & res) + (0x0001000100010001 & (res >>8 )); + res = (0x0000000700000007 & res) + (0x0000000700000007 & (res >>16 )); +if (size ==32 ) {/* prtyw */ + regs->gpr[ra] = (0x0000000100000001 & res); +return ; + } + + res = (0x000000000000000f & res) + (0x000000000000000f & (res >>32 )); + regs->gpr[ra] = res &1 ;/*prtyd */ +}
Does 7's and 0xf look strange (since the top bit in the values there is always 0). You always "and" the values with 1 later, you could do that immediately (and change + to ^).
A general question about these patches: some things are inside #ifdef __powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?
Segher
Re: [PATCH v3 2/5] powerpc/lib/sstep: Add popcnt instruction emulation
In reply to Matt BrownQuoted 23 lines · Matt Brown
This adds emulations for the popcntb, popcntw, and popcntd instructions. Tested for correctness against the popcnt{b,w,d} instructions on ppc64le.
Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed typecasting
- fixed bitshifting error for both 32 and 64bit arch--- arch/powerpc/lib/sstep.c | 40 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 39 insertions(+), 1 deletion(-)arch/powerpc/lib/sstep.c +0
-0@@ -612,6 +612,32 @@ static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, regs->gpr[rd] = out_val; }
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 624
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+/* +* The size parameter is used to adjust the equivalent popcnt instruction. +* popcntb = 8, popcntw = 32, popcntd = 64 +*/ +static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long v1, +int size,int ra) +{ +unsigned long long out = v1; + + out = (0x5555555555555555 & out) + (0x5555555555555555 & (out >>1 ));
This can be simplified in a less obvious way as:
out -= (out >> 1) & 0x5555555555555555;
which maps each pair of bits according to the following:
00 -> 00
01 -> 01
10 -> 01
11 -> 10
This should save one instruction.
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 625
Quoted 1 line
arch/powerpc/lib/sstep.c +1
-0+ out = (0x3333333333333333 & out) + (0x3333333333333333 & (out >>2 ));
Ok, but now each nibble is between 0 and 4, so the addition of two nibbles can't overflow or generate carry into the higher one.
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 626
Quoted 1 line
arch/powerpc/lib/sstep.c +1
-0+ out = (0x0f0f0f0f0f0f0f0f & out) + (0x0f0f0f0f0f0f0f0f & (out >>4 ));
out += out >> 4; out &= 0x0f0f0f0f0f0f0f0f;
which should also save one instruction
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 630
Quoted 4 lines
arch/powerpc/lib/sstep.c +4
-0+if (size ==8 ) {/* popcntb */ + regs->gpr[ra] = out; +return ; + }
At this point each count occupies at least one byte and can no more overflow, so masking is only needed before returning.
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 631
Quoted 1 line
arch/powerpc/lib/sstep.c +1
-0+ out = (0x001f001f001f001f & out) + (0x001f001f001f001f & (out >>8 ));
out += out >> 8;
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 632
Quoted 1 line
arch/powerpc/lib/sstep.c +1
-0+ out = (0x0000003f0000003f & out) + (0x0000003f0000003f & (out >>16 ));
out += out >> 16;
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 634
Quoted 2 lines
arch/powerpc/lib/sstep.c +2
-0+if (size ==32 ) {/* popcntw */ + regs->gpr[ra] = out;
regs->gpr[ra] = out & 0x0000003f0000003f;
arch/powerpc/lib/sstep.c static nokprobe_inline void do_cmpb(struct pt_regs *regs, unsigned long v1, · line 637
Quoted 3 lines
arch/powerpc/lib/sstep.c +3
-0+return ; + } + out = (0x000000000000007f & out) + (0x000000000000007f & (out >>32 ));
out = (out + (out >> 32)) & 0x7f; Gabriel
Quoted 43 lines
arch/powerpc/lib/sstep.c +16
-1+ regs->gpr[ra] = out;/* popcntd */ +} + static nokprobe_inline int trap_compare(long v1, long v2) {int ret =0 ;@@ -1194,6 +1220,10 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs, regs->gpr[ra] = regs->gpr[rd] & ~regs->gpr[rb]; goto logical_done; +case 122 :/* popcntb */ +do_popcnt (regs, regs->gpr[rd],8 , ra); + goto logical_done; +case 124 :/* nor */ regs->gpr[ra] = ~(regs->gpr[rd] | regs->gpr[rb]); goto logical_done;@@ -1206,6 +1236,10 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs, regs->gpr[ra] = regs->gpr[rd] ^ regs->gpr[rb]; goto logical_done; +case 378 :/* popcntw */ +do_popcnt (regs, regs->gpr[rd],32 , ra); + goto logical_done; +case 412 :/* orc */ regs->gpr[ra] = regs->gpr[rd] | ~regs->gpr[rb]; goto logical_done;@@ -1217,7 +1251,11 @@ int analyse_instr(struct instruction_op *op, struct pt_regs *regs,case 476 :/* nand */ regs->gpr[ra] = ~(regs->gpr[rd] & regs->gpr[rb]); goto logical_done; - +#ifdef __powerpc64__ +case 506 :/* popcntd */ +do_popcnt (regs, regs->gpr[rd],64 , ra); + goto logical_done; +#endif case 922 :/* extsh */ regs->gpr[ra] = (signed short ) regs->gpr[rd]; goto logical_done;Signature · 1 line
2.9.3
Re: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
In reply to Balbir SinghQuoted 38 lines · Balbir Singh
Quoted 37 lines · Matt Brown
This adds emulation for the prtyw and prtyd instructions. Tested for logical correctness against the prtyw and prtyd instructions on ppc64le.
Signed-off-by Matt Brown <[email protected]>---
v3:
- optimised using the Giles-Miller method of side-ways addition
v2:
- fixed opcodes
- fixed bitshifting and typecast errors
- mergeddo_prtywanddo_prtydinto single function--- arch/powerpc/lib/sstep.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)arch/powerpc/lib/sstep.c +13
-0@@ -655,6 +655,25 @@ static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, regs->gpr[ra] = perm; }#endif /* __powerpc64__ */ +/* +* The size parameter adjusts the equivalent prty instruction. +* prtyw = 32, prtyd = 64 +*/ +static nokprobe_inline void do_prty(struct pt_regs *regs, unsigned long v, +int size,int ra) +{ +unsigned long long res = v; + + res = (0x0001000100010001 & res) + (0x0001000100010001 & (res >>8 )); + res = (0x0000000700000007 & res) + (0x0000000700000007 & (res >>16 )); +if (size ==32 ) {/* prtyw */ + regs->gpr[ra] = (0x0000000100000001 & res);
Quoted 8 lines
arch/powerpc/lib/sstep.c static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, · line 675
Quoted 5 lines
arch/powerpc/lib/sstep.c +5
-0+return ; + } + + res = (0x000000000000000f & res) + (0x000000000000000f & (res >>32 )); + regs->gpr[ra] = res &1 ;/*prtyd */ Looks good, you can also xor instead of adding, but the masks would need to change in that case.
Actually, I'd prefer to use xor for this case, since you hardly ever need any mask, except at the end. Most masks are only needed because of carry propagation, which the xor completely avoid.
In other words:
unsigned long long res = v ^ (v >> 8);
res ^= res >> 16;
if (size == 32) {
regs->gpr[ra] = res & 0x0000000100000001;
return;
}
res ^= res >> 32;
regs-> gpr[ra] = res & 1;should work, but I've not even compiled it.
Gabriel
Michael Ellerman
reviewerRe: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
In reply to Segher BoessenkoolQuoted 21 lines · Segher Boessenkool
Quoted 5 lines · Matt Brown
arch/powerpc/lib/sstep.c +5
-0+static nokprobe_inline void do_prty(struct pt_regs *regs, unsigned long v, +int size,int ra) +{ +unsigned long long res = v; +arch/powerpc/lib/sstep.c static nokprobe_inline void do_bpermd(struct pt_regs *regs, unsigned long v1, · line 676
Quoted 10 lines
arch/powerpc/lib/sstep.c +10
-0+ res = (0x0001000100010001 & res) + (0x0001000100010001 & (res >>8 )); + res = (0x0000000700000007 & res) + (0x0000000700000007 & (res >>16 )); +if (size ==32 ) {/* prtyw */ + regs->gpr[ra] = (0x0000000100000001 & res); +return ; + } + + res = (0x000000000000000f & res) + (0x000000000000000f & (res >>32 )); + regs->gpr[ra] = res &1 ;/*prtyd */ +}Does 7's and
0xflook strange (since the top bit in the values there is always 0). You always "and" the values with 1 later, you could do that immediately (and change + to ^).
Quoted 3 lines
A general question about these patches: some things are inside #ifdef
__powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?
Why is it the wrong macro? Because we tend to use CONFIG_PPC64 you mean?
I thought the reason some are #ifdef'ed is that some are 64-bit only. ie. bpermd is 64-bit only ?
cheers
Re: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
Quoted 6 lines · Michael Ellerman
Quoted 3 lines · Segher Boessenkool
A general question about these patches: some things are inside #ifdef
__powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?Why is it the wrong macro? Because we tend to use
CONFIG_PPC64you mean?
Yeah. But I see sstep.c already mixes those two at will (or if there is a distinction, I'm not seeing it :-) )
Quoted 2 lines
I thought the reason some are #ifdef'ed is that some are 64-bit only. ie. bpermd is 64-bit only ?
64-bit only, in what way? It's not clear what the rules are.
It's not instructions that can only run in 64-bit mode. It's not instructions that only give a usable result with 64-bit regs implemented.
It's not instructions only implemented on 64-bit CPUs. It's not even "all instructions that would not give a correct result in the low 32 bits of GPRs if the high 32 bits are not implemented".
Segher
Michael Ellerman
reviewerRe: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
Quoted 10 lines · Segher Boessenkool
Quoted 6 lines · Michael Ellerman
Quoted 3 lines · Segher Boessenkool
A general question about these patches: some things are inside #ifdef
__powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?Why is it the wrong macro? Because we tend to use
CONFIG_PPC64you mean?Yeah. But I see
sstep.calready mixes those two at will (or if there is a distinction, I'm not seeing it :-) )
Yeah OK. In practice they're equivalent, if CONFIG_PPC64=y then the kernel is built 64-bit and therefore __powerpc64__ is defined.
But I agree it's a mess, we should use CONFIG_PPC64 exclusively unless there's some reason not to (which I don't think there ever is).
Quoted 4 lines
Quoted 2 lines
I thought the reason some are #ifdef'ed is that some are 64-bit only. ie. bpermd is 64-bit only ?
64-bit only, in what way? It's not clear what the rules are.
Instructions that have "d" in the name? :P
Quoted 4 lines
It's not instructions that can only run in 64-bit mode. It's not instructions that only give a usable result with 64-bit regs implemented.
It's not instructions only implemented on 64-bit CPUs.
I think it's trying to be that ^
If you build a 32-bit kernel then instructions that are only defined on 64-bit CPUs should be treated as illegal, so the easiest way to achieve that is to #ifdef off the code for those instructions.
cheers
Re: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
Quoted 31 lines · Michael Ellerman
Quoted 10 lines · Segher Boessenkool
Quoted 6 lines · Michael Ellerman
Quoted 3 lines · Segher Boessenkool
A general question about these patches: some things are inside #ifdef
__powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?Why is it the wrong macro? Because we tend to use
CONFIG_PPC64you mean?Yeah. But I see
sstep.calready mixes those two at will (or if there is a distinction, I'm not seeing it :-) )Yeah OK. In practice they're equivalent, if
CONFIG_PPC64=ythen the kernel is built 64-bit and therefore__powerpc64__is defined.But I agree it's a mess, we should use
CONFIG_PPC64exclusively unless there's some reason not to (which I don't think there ever is).Quoted 4 lines
Quoted 2 lines
I thought the reason some are #ifdef'ed is that some are 64-bit only. ie. bpermd is 64-bit only ?
64-bit only, in what way? It's not clear what the rules are.
Instructions that have "d" in the name? :P
Quoted 4 lines
It's not instructions that can only run in 64-bit mode. It's not instructions that only give a usable result with 64-bit regs implemented.
It's not instructions only implemented on 64-bit CPUs.
Quoted 6 lines
I think it's trying to be that ^
If you build a 32-bit kernel then instructions that are only defined on 64-bit CPUs should be treated as illegal, so the easiest way to achieve that is to #ifdef off the code for those instructions.
I'll fixup this up to use the xor implementation, and change the series to use CONFIG_PPC64 for the ifdef.
Thanks,
Matt
Quoted 1 line
cheers
Michael Ellerman
reviewerRe: [PATCH v3 4/5] powerpc/lib/sstep: Add prty instruction emulation
Quoted 32 lines · Matt Brown
Quoted 31 lines · Michael Ellerman
Quoted 10 lines · Segher Boessenkool
Quoted 6 lines · Michael Ellerman
Quoted 3 lines · Segher Boessenkool
A general question about these patches: some things are inside #ifdef
__powerpc64__, some are not. It seems it is the wrong macro, and it should be used (or not used) consistently?Why is it the wrong macro? Because we tend to use
CONFIG_PPC64you mean?Yeah. But I see
sstep.calready mixes those two at will (or if there is a distinction, I'm not seeing it :-) )Yeah OK. In practice they're equivalent, if
CONFIG_PPC64=ythen the kernel is built 64-bit and therefore__powerpc64__is defined.But I agree it's a mess, we should use
CONFIG_PPC64exclusively unless there's some reason not to (which I don't think there ever is).Quoted 4 lines
Quoted 2 lines
I thought the reason some are #ifdef'ed is that some are 64-bit only. ie. bpermd is 64-bit only ?
64-bit only, in what way? It's not clear what the rules are.
Instructions that have "d" in the name? :P
Quoted 4 lines
It's not instructions that can only run in 64-bit mode. It's not instructions that only give a usable result with 64-bit regs implemented.
It's not instructions only implemented on 64-bit CPUs.
Quoted 8 lines
Quoted 5 lines
I think it's trying to be that ^
If you build a 32-bit kernel then instructions that are only defined on 64-bit CPUs should be treated as illegal, so the easiest way to achieve that is to #ifdef off the code for those instructions.
I'll fixup this up to use the xor implementation, and change the series to use
CONFIG_PPC64for the ifdef.
Thanks.
In terms of the actual algorithms, we want to use something reasonably efficient, but don't sweat too much on the performance. The overhead of the emulation is in the exception we took to get there. So readable and obviously correct source is the main goal.
cheers