| Message ID | 0E3A13C8-561C-4587-AEAE-A067F9DD356D@comcast.net |
|---|---|
| State | New |
| Headers | show |
> This patch allows a target to increase the cost of anti-deps to better > reflect the actual cost on the machine. But it can already do it via the TARGET_SCHED_ADJUST_COST hook, can't it?
On 12/11/2015 02:22 AM, Eric Botcazou wrote: >> This patch allows a target to increase the cost of anti-deps to better >> reflect the actual cost on the machine. > > But it can already do it via the TARGET_SCHED_ADJUST_COST hook, can't it? And can't this be done with define_bypass as well? Jeff
On Dec 11, 2015, at 1:22 AM, Eric Botcazou <ebotcazou@adacore.com> wrote: >> This patch allows a target to increase the cost of anti-deps to better >> reflect the actual cost on the machine. > > But it can already do it via the TARGET_SCHED_ADJUST_COST hook, can't it? The undocumented TARGET_SCHED_ADJUST_COST_2 seems a better fit. Yes, that works, I can use it. I’m assuming the lack of documentation is a simple error.
On Dec 11, 2015, at 6:09 AM, Jeff Law <law@redhat.com> wrote: > On 12/11/2015 02:22 AM, Eric Botcazou wrote: >>> This patch allows a target to increase the cost of anti-deps to better >>> reflect the actual cost on the machine. >> >> But it can already do it via the TARGET_SCHED_ADJUST_COST hook, can't it? > And can't this be done with define_bypass as well? if (dep_type == REG_DEP_ANTI) cost = 0; else if (dep_type == REG_DEP_OUTPUT) { cost = (insn_default_latency (insn) - insn_default_latency (used)); if (cost <= 0) cost = 1; } else if (bypass_p (insn)) cost = insn_latency (insn, used); I don’t see how if the first case is true, one gets into the third without code mods. I opted for the adjust_cost_2 hook.
Index: defaults.h =================================================================== --- defaults.h (revision 231539) +++ defaults.h (working copy) @@ -1486,6 +1486,10 @@ #define TARGET_VTABLE_USES_DESCRIPTORS 0 #endif +#ifndef TARGET_ANTI_DEP_COST +#define TARGET_ANTI_DEP_COST 0 +#endif + #endif /* GCC_INSN_FLAGS_H */ #endif /* ! GCC_DEFAULTS_H */ Index: doc/tm.texi =================================================================== --- doc/tm.texi (revision 231539) +++ doc/tm.texi (working copy) @@ -6970,6 +6970,13 @@ the hook implementation for how different fusion types are supported. @end deftypefn +@defmac TARGET_ANTI_DEP_COST +The cost in cycles for an anti-dependency. Defaults to 0. On non-OOO +multi-issue machines that can't issue instructions that have +overlapping registers in the same cycle, a value of 1 will better +reflect the actual cost of the instruction sequence. +@end defmac + @node Sections @section Dividing the Output into Sections (Texts, Data, @dots{}) @c the above section title is WAY too long. maybe cut the part between Index: doc/tm.texi.in =================================================================== --- doc/tm.texi.in (revision 231539) +++ doc/tm.texi.in (working copy) @@ -4852,6 +4852,13 @@ @hook TARGET_SCHED_FUSION_PRIORITY +@defmac TARGET_ANTI_DEP_COST +The cost in cycles for an anti-dependency. Defaults to 0. On non-OOO +multi-issue machines that can't issue instructions that have +overlapping registers in the same cycle, a value of 1 will better +reflect the actual cost of the instruction sequence. +@end defmac + @node Sections @section Dividing the Output into Sections (Texts, Data, @dots{}) @c the above section title is WAY too long. maybe cut the part between Index: haifa-sched.c =================================================================== --- haifa-sched.c (revision 231539) +++ haifa-sched.c (working copy) @@ -1470,7 +1470,7 @@ if (INSN_CODE (insn) >= 0) { if (dep_type == REG_DEP_ANTI) - cost = 0; + cost = TARGET_ANTI_DEP_COST; else if (dep_type == REG_DEP_OUTPUT) { cost = (insn_default_latency (insn)