[PATCH ARM]Use stricter constraint for pattern arm_movqi_insn to avoid constant pool
Ramana Radhakrishnan
ramana.gcc@googlemail.com
Thu Sep 4 15:21:00 GMT 2014
On Thu, Sep 4, 2014 at 6:33 AM, Bin Cheng <bin.cheng@arm.com> wrote:
> Hi,
> This patch is to fix a potential bug in arm pattern "arm_movqi_insn".
>
> For the pattern,
> (define_insn "*arm_movqi_insn"
> [(set (match_operand:QI 0 "nonimmediate_operand" "=r,r,r,l,r,l,Uu,r,m")
> (match_operand:QI 1 "general_operand" "r,r,I,Py,K,Uu,l,m,r"))]
> "TARGET_32BIT
> && ( register_operand (operands[0], QImode)
> || register_operand (operands[1], QImode))"
> "@
> mov%?\\t%0, %1
> mov%?\\t%0, %1
> mov%?\\t%0, %1
> mov%?\\t%0, %1
> mvn%?\\t%0, #%B1
> ldr%(b%)\\t%0, %1
> str%(b%)\\t%1, %0
> ldr%(b%)\\t%0, %1
> str%(b%)\\t%1, %0"
> [(set_attr "type"
> "mov_reg,mov_reg,mov_imm,mov_imm,mvn_imm,load1,store1,load1,store1")
> (set_attr "predicable" "yes")
> (set_attr "predicable_short_it" "yes,yes,yes,no,no,no,no,no,no")
> (set_attr "arch" "t2,any,any,t2,any,t2,t2,any,any")
> (set_attr "length" "2,4,4,2,4,2,2,4,4")]
> )
>
> Both predicate "general_operand" and constraint "m" of the 2nd operand
> support load from constant pool. Problem is both attribute pool_range and
> neg_pool_range are default to 0 because we don't set it explicitly. When
> GCC comes to situation that it generates load from constant pool for this
> pattern, function push_minipool_fix runs into assertion failure.
>
> This patch uses stricter constraint "Uh" just like r212303. This patch is
> necessary for inlining memset calls on ARM target.
>
> It passes tests on
> arm-none-eabi/arm-none-linux-gnueabi/arm-none-linux-gnueabihf with different
> test variants. It also passes bootstrap and glibc build on
> arm-none-linux-gnueabihf. Is it OK?
Ok if no regressions. Thanks,
Ramana
>
> Thanks,
> bin
>
> 2014-09-04 Bin Cheng <bin.cheng@arm.com>
>
> * config/arm/arm.md (arm_movqi_insn): Use Uh instead of m
> constraint.
More information about the Gcc-patches
mailing list