Re: ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

On Sat, Sep 24, 2016 at 5:36 AM, kugan wrote: > Hi Richard, > > There is also one more issue as reported by Pat Haugen. That is, in > converting value_range of ssa (which we get from get_range_info) to > argument type, my implementation is too simplistic and

Re: ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

Hi Richard, There is also one more issue as reported by Pat Haugen. That is, in converting value_range of ssa (which we get from get_range_info) to argument type, my implementation is too simplistic and wrong at times. I can check TYPE_UNSIGNED here but that would be pessimistic. tree-vrp

Re: ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

On Fri, Sep 23, 2016 at 10:58 AM, kugan wrote: > Hi Richard, > > Thanks for the review. > > On 23/09/16 17:19, Richard Biener wrote: >> >> On Fri, Sep 23, 2016 at 12:24 AM, kugan >> wrote: >>> >>> Hi, >>> As Richard pointed

Re: ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

Hi Richard, Thanks for the review. On 23/09/16 17:19, Richard Biener wrote: On Fri, Sep 23, 2016 at 12:24 AM, kugan wrote: Hi, As Richard pointed out in PR77677, TREE_OVERFLOW is not cleared in IPA-VRP. There are three places in which we set value_range:

Re: ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

On Fri, Sep 23, 2016 at 12:24 AM, kugan wrote: > Hi, > As Richard pointed out in PR77677, TREE_OVERFLOW is not cleared in IPA-VRP. > There are three places in which we set value_range: > > > 1. When value ranges are obtained from SSA_NAME with get_range_info

ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu

Hi, As Richard pointed out in PR77677, TREE_OVERFLOW is not cleared in IPA-VRP. There are three places in which we set value_range: 1. When value ranges are obtained from SSA_NAME with get_range_info with wide_int_to_tree. In this case we will not have TREE_OVERFLOW set. 2. When we