| Message ID | 1426648281-13955-1-git-send-email-aik@ozlabs.ru |
|---|---|
| State | New |
| Headers | show |
s/existed/existing/ or "that never existed" Am 18.03.2015 um 04:11 schrieb Alexey Kardashevskiy: > IBM uses low 16bits to specify a chip version of a POWER CPU. "specify the chip version of a" > So there has never been an actual silicon with PVR = 0x003B0000. > The first silicon would have PVR 0x003B0100 but it is very unlikely > to find it in any machine shipped to any customer as it is was too raw. "as it was" > > This removes CPU_POWERPC_POWER5P_v00 definition and changes Maybe worth pointing out that the POWER5+_v0.0 QOM type was introduced in recent commit d7586dc426472b5ad0f5c01b5c7c551eeb5a6003 (target-ppc: Add versions to server CPU descriptions)? > POWER5+ and POWERgs aliases (which are synonyms) to point to > POWER5+_v2.1 which can still be found in real machines. FTR, 171777a4b38a0f6331ae60c2546a5baf84c4b359 (target-ppc: Turn POWER5gs CPU into alias for POWER5+) set the POWER5+ alias up as conflict resolution before I enabled the #ifdef TODO'ed POWER5P code in 35ebcb2b7a469739e6452d27379181bfbfc0388d (target-ppc: Prepare POWER5P CPU family). > > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> > --- > > > I asked Paul. He suggested that there has never been an actual > POWER5 silicon with PVR which low 16 bits are zeroes, > the first one would be 0x003B0100 but it would be so buggy so > it would not be shipped to any real customer. > And then he suggested to look at the real POWER5+ machine, > we looked around and found one: > > cpu : POWER5+ (gs) > clock : 1898.100000MHz > revision : 2.0 (pvr 003b 0200) > > I believe 3b 0201 is also something real and it is defined already in QEMU > so here is a patch. Yes, I have a 2.1, so having POWER5+ point to it should be fine, it being the latest and now only one we have. Reviewed-by: Andreas Färber <afaerber@suse.de> Alex, can you tweak the commit message and apply this for 2.3 please, as fixup for the too hastily applied fix? > Yes, this does not touch the cpu family class registration issue, just > a tiny cleanup :) We can fix that for 2.4. :) Same for Alex' class cache that I just saw again. Thanks, Andreas
On 03/18/2015 02:36 PM, Andreas Färber wrote: > s/existed/existing/ or "that never existed" > > Am 18.03.2015 um 04:11 schrieb Alexey Kardashevskiy: >> IBM uses low 16bits to specify a chip version of a POWER CPU. > "specify the chip version of a" > >> So there has never been an actual silicon with PVR = 0x003B0000. >> The first silicon would have PVR 0x003B0100 but it is very unlikely >> to find it in any machine shipped to any customer as it is was too raw. > "as it was" > >> This removes CPU_POWERPC_POWER5P_v00 definition and changes > Maybe worth pointing out that the POWER5+_v0.0 QOM type was introduced > in recent commit d7586dc426472b5ad0f5c01b5c7c551eeb5a6003 (target-ppc: > Add versions to server CPU descriptions)? > >> POWER5+ and POWERgs aliases (which are synonyms) to point to >> POWER5+_v2.1 which can still be found in real machines. > FTR, 171777a4b38a0f6331ae60c2546a5baf84c4b359 (target-ppc: Turn POWER5gs > CPU into alias for POWER5+) set the POWER5+ alias up as conflict > resolution before I enabled the #ifdef TODO'ed POWER5P code in > 35ebcb2b7a469739e6452d27379181bfbfc0388d (target-ppc: Prepare POWER5P > CPU family). > >> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> >> --- >> >> >> I asked Paul. He suggested that there has never been an actual >> POWER5 silicon with PVR which low 16 bits are zeroes, >> the first one would be 0x003B0100 but it would be so buggy so >> it would not be shipped to any real customer. >> And then he suggested to look at the real POWER5+ machine, >> we looked around and found one: >> >> cpu : POWER5+ (gs) >> clock : 1898.100000MHz >> revision : 2.0 (pvr 003b 0200) >> >> I believe 3b 0201 is also something real and it is defined already in QEMU >> so here is a patch. > Yes, I have a 2.1, so having POWER5+ point to it should be fine, it > being the latest and now only one we have. > > Reviewed-by: Andreas Färber <afaerber@suse.de> > > Alex, can you tweak the commit message and apply this for 2.3 please, as > fixup for the too hastily applied fix? Sure. Thanks, applied to ppc-next. Alex
diff --git a/target-ppc/cpu-models.c b/target-ppc/cpu-models.c index 2b560a4..4d5ab4b 100644 --- a/target-ppc/cpu-models.c +++ b/target-ppc/cpu-models.c @@ -1124,8 +1124,6 @@ POWERPC_DEF("POWER5", CPU_POWERPC_POWER5, POWER5, "POWER5") #endif - POWERPC_DEF("POWER5+_v0.0", CPU_POWERPC_POWER5P_v00, POWER5P, - "POWER5+ v0.0") POWERPC_DEF("POWER5+_v2.1", CPU_POWERPC_POWER5P_v21, POWER5P, "POWER5+ v2.1") #if defined(TODO) @@ -1387,8 +1385,8 @@ PowerPCCPUAlias ppc_cpu_aliases[] = { { "Dino", "POWER3" }, { "POWER3+", "631" }, { "POWER5gr", "POWER5" }, - { "POWER5+", "POWER5+_v0.0" }, - { "POWER5gs", "POWER5+_v0.0" }, + { "POWER5+", "POWER5+_v2.1" }, + { "POWER5gs", "POWER5+_v2.1" }, { "POWER7", "POWER7_v2.3" }, { "POWER7+", "POWER7+_v2.1" }, { "POWER8E", "POWER8E_v1.0" }, diff --git a/target-ppc/cpu-models.h b/target-ppc/cpu-models.h index ee693af..9d80e72 100644 --- a/target-ppc/cpu-models.h +++ b/target-ppc/cpu-models.h @@ -547,7 +547,6 @@ enum { CPU_POWERPC_POWER4P = 0x00380000, /* XXX: missing 0x003A0201 */ CPU_POWERPC_POWER5 = 0x003A0203, - CPU_POWERPC_POWER5P_v00 = 0x003B0000, CPU_POWERPC_POWER5P_v21 = 0x003B0201, CPU_POWERPC_POWER6 = 0x003E0000, CPU_POWERPC_POWER6_5 = 0x0F000001, /* POWER6 in POWER5 mode */
IBM uses low 16bits to specify a chip version of a POWER CPU. So there has never been an actual silicon with PVR = 0x003B0000. The first silicon would have PVR 0x003B0100 but it is very unlikely to find it in any machine shipped to any customer as it is was too raw. This removes CPU_POWERPC_POWER5P_v00 definition and changes POWER5+ and POWERgs aliases (which are synonyms) to point to POWER5+_v2.1 which can still be found in real machines. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> --- I asked Paul. He suggested that there has never been an actual POWER5 silicon with PVR which low 16 bits are zeroes, the first one would be 0x003B0100 but it would be so buggy so it would not be shipped to any real customer. And then he suggested to look at the real POWER5+ machine, we looked around and found one: cpu : POWER5+ (gs) clock : 1898.100000MHz revision : 2.0 (pvr 003b 0200) I believe 3b 0201 is also something real and it is defined already in QEMU so here is a patch. Yes, this does not touch the cpu family class registration issue, just a tiny cleanup :) --- target-ppc/cpu-models.c | 6 ++---- target-ppc/cpu-models.h | 1 - 2 files changed, 2 insertions(+), 5 deletions(-)