[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 06/27] Correct ppc popcntb logic, implement popc
From: |
Alexander Graf |
Subject: |
Re: [Qemu-devel] [PATCH 06/27] Correct ppc popcntb logic, implement popcntw and popcntd |
Date: |
Fri, 1 Apr 2011 20:02:24 +0200 |
On 01.04.2011, at 19:58, Aurelien Jarno wrote:
> On Fri, Apr 01, 2011 at 03:15:13PM +1100, David Gibson wrote:
>> From: David Gibson <address@hidden>
>>
>> qemu already includes support for the popcntb instruction introduced
>> in POWER5 (although it doesn't actually allow you to choose POWER5).
>>
>> However, the logic is slightly incorrect: it will generate results
>> truncated to 32-bits when the CPU is in 32-bit mode. This is not
>> normal for powerpc - generally arithmetic instructions on a 64-bit
>> powerpc cpu will generate full 64 bit results, it's just that only the
>> low 32 bits will be significant for condition codes.
>>
>> This patch corrects this nit, which actually simplifies the code slightly.
>>
>> In addition, this patch implements the popcntw and popcntd
>> instructions added in POWER7, in preparation for allowing POWER7 as an
>> emulated CPU.
>>
>> Signed-off-by: David Gibson <address@hidden>
>> ---
>> target-ppc/cpu.h | 2 +
>> target-ppc/helper.h | 3 +-
>> target-ppc/op_helper.c | 55 +++++++++++++++++++++++++++++++++++++++++++----
>> target-ppc/translate.c | 20 +++++++++++++----
>> 4 files changed, 69 insertions(+), 11 deletions(-)
>>
>> diff --git a/target-ppc/cpu.h b/target-ppc/cpu.h
>> index f293f85..37dde39 100644
>> --- a/target-ppc/cpu.h
>> +++ b/target-ppc/cpu.h
>> @@ -1505,6 +1505,8 @@ enum {
>> PPC_DCRX = 0x2000000000000000ULL,
>> /* user-mode DCR access, implemented in PowerPC 460
>> */
>> PPC_DCRUX = 0x4000000000000000ULL,
>> + /* popcntw and popcntd instructions
>> */
>> + PPC_POPCNTWD = 0x8000000000000000ULL,
>> };
>>
>> /*****************************************************************************/
>> diff --git a/target-ppc/helper.h b/target-ppc/helper.h
>> index 2b4744d..7c02be9 100644
>> --- a/target-ppc/helper.h
>> +++ b/target-ppc/helper.h
>> @@ -38,10 +38,11 @@ DEF_HELPER_2(mulldo, i64, i64, i64)
>>
>> DEF_HELPER_FLAGS_1(cntlzw, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> DEF_HELPER_FLAGS_1(popcntb, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> +DEF_HELPER_FLAGS_1(popcntw, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> DEF_HELPER_2(sraw, tl, tl, tl)
>> #if defined(TARGET_PPC64)
>> DEF_HELPER_FLAGS_1(cntlzd, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> -DEF_HELPER_FLAGS_1(popcntb_64, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> +DEF_HELPER_FLAGS_1(popcntd, TCG_CALL_CONST | TCG_CALL_PURE, tl, tl)
>> DEF_HELPER_2(srad, tl, tl, tl)
>> #endif
>>
>> diff --git a/target-ppc/op_helper.c b/target-ppc/op_helper.c
>> index aa2e8ba..b1b883d 100644
>> --- a/target-ppc/op_helper.c
>> +++ b/target-ppc/op_helper.c
>> @@ -499,6 +499,50 @@ target_ulong helper_srad (target_ulong value,
>> target_ulong shift)
>> }
>> #endif
>>
>> +#if defined(TARGET_PPC64)
>> +target_ulong helper_popcntb (target_ulong val)
>> +{
>> + val = (val & 0x5555555555555555ULL) + ((val >> 1) &
>> + 0x5555555555555555ULL);
>> + val = (val & 0x3333333333333333ULL) + ((val >> 2) &
>> + 0x3333333333333333ULL);
>> + val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
>> + 0x0f0f0f0f0f0f0f0fULL);
>> + return val;
>> +}
>> +
>> +target_ulong helper_popcntw (target_ulong val)
>> +{
>> + val = (val & 0x5555555555555555ULL) + ((val >> 1) &
>> + 0x5555555555555555ULL);
>> + val = (val & 0x3333333333333333ULL) + ((val >> 2) &
>> + 0x3333333333333333ULL);
>> + val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
>> + 0x0f0f0f0f0f0f0f0fULL);
>> + val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) &
>> + 0x00ff00ff00ff00ffULL);
>> + val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) &
>> + 0x0000ffff0000ffffULL);
>> + return val;
>> +}
>> +
>> +target_ulong helper_popcntd (target_ulong val)
>> +{
>> + val = (val & 0x5555555555555555ULL) + ((val >> 1) &
>> + 0x5555555555555555ULL);
>> + val = (val & 0x3333333333333333ULL) + ((val >> 2) &
>> + 0x3333333333333333ULL);
>> + val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
>> + 0x0f0f0f0f0f0f0f0fULL);
>> + val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) &
>> + 0x00ff00ff00ff00ffULL);
>> + val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) &
>> + 0x0000ffff0000ffffULL);
>> + val = (val & 0x00000000ffffffffULL) + ((val >> 32) &
>> + 0x00000000ffffffffULL);
>> + return val;
>> +}
>
> I probably arrive a bit late, but note that for this one you can use
> ctpop64() (from host-utils.h), which also uses a GCC builtin when
> available.
Ah, nice. David - sounds like a good chance for a follow-up patch :)
Alex
- [Qemu-devel] [0/27] Implement emulation of pSeries logical partitions (v6), David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 02/27] Allow qemu_devtree_setprop() to take arbitrary values, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 03/27] Add a hook to allow hypercalls to be emulated on PowerPC, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 01/27] Clean up PowerPC SLB handling code, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 12/27] Add POWER7 support for ppc, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 04/27] Implement PowerPC slbmfee and slbmfev instructions, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 11/27] Support 1T segments on ppc, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 06/27] Correct ppc popcntb logic, implement popcntw and popcntd, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 17/27] Implement assorted pSeries hcalls and RTAS methods, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 08/27] Parse SDR1 on mtspr instead of at translate time, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 13/27] Start implementing pSeries logical partition machine, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 15/27] Virtual hash page table handling on pSeries machine, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 16/27] Implement hcall based RTAS for pSeries machines, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 25/27] Add a PAPR TCE-bypass mechanism for the pSeries machine, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 05/27] Implement missing parts of the logic for the POWER PURR, David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 22/27] Implement sPAPR Virtual LAN (ibmveth), David Gibson, 2011/04/01
- [Qemu-devel] [PATCH 14/27] Implement the bus structure for PAPR virtual IO, David Gibson, 2011/04/01