Re: [Piglit] [PATCH] glsl-1.30: Reproduce a bug in the i965 backend optimizer

2018-09-12 Thread Ilia Mirkin
On Wed, Sep 12, 2018 at 7:59 PM, Ian Romanick wrote: > On 09/12/2018 04:33 PM, Ilia Mirkin wrote: >> On Wed, Sep 12, 2018 at 7:29 PM, Ian Romanick wrote: >>> From: Ian Romanick >>> >>> The optimizer recently added the ability to replace a compare with a >>> subtraction under certain

Re: [Piglit] [PATCH] glsl-1.30: Reproduce a bug in the i965 backend optimizer

2018-09-12 Thread Ian Romanick
On 09/12/2018 04:33 PM, Ilia Mirkin wrote: > On Wed, Sep 12, 2018 at 7:29 PM, Ian Romanick wrote: >> From: Ian Romanick >> >> The optimizer recently added the ability to replace a compare with a >> subtraction under certain circumstances. This can fail for integers. >> For inputs a =

Re: [Piglit] [PATCH] glsl-1.30: Reproduce a bug in the i965 backend optimizer

2018-09-12 Thread Ilia Mirkin
On Wed, Sep 12, 2018 at 7:29 PM, Ian Romanick wrote: > From: Ian Romanick > > The optimizer recently added the ability to replace a compare with a > subtraction under certain circumstances. This can fail for integers. > For inputs a = 0x8000, b = 4, int(0x8000) < 4, but >

[Piglit] [PATCH] glsl-1.30: Reproduce a bug in the i965 backend optimizer

2018-09-12 Thread Ian Romanick
From: Ian Romanick The optimizer recently added the ability to replace a compare with a subtraction under certain circumstances. This can fail for integers. For inputs a = 0x8000, b = 4, int(0x8000) < 4, but int(0x8000) - 4 overflows and results in 0x7ffc. That's not less than