search for: g_ptr_add

Displaying 3 results from an estimated 3 matches for "g_ptr_add".

GlobalISel round table follow up: register bank select
2020 Oct 07
2
GlobalISel round table follow up: register bank select
...ankSelect goes top-down, it never looks at all available mappings for the operands when considering which of the mappings to apply to the current instruction. In our architecture we have one register bank dedicated to pointers and another one for anything else. We often see code where we have a G_PTR_ADD with a constant. Since the constant is not a pointer, we put it on the other register bank. We could put it on the address regbank and do provide alternative mappings for that, but since the greedy algorithm doesn't actually check the usage of the constant, it is always put on the other ba...
GlobalISel round table follow up: register bank select
2020 Oct 08
2
GlobalISel round table follow up: register bank select
...oks at all available mappings for the operands when >> considering which of the mappings to apply to the current >> instruction. In our architecture we have one register bank dedicated >> to pointers and another one for anything else. We often see code >> where we have a G_PTR_ADD with a constant. Since the constant is not >> a pointer, we put it on the other register bank. We could put it on >> the address regbank and do provide alternative mappings for that, but >> since the greedy algorithm doesn't actually check the usage of the >> constan...
GlobalISel round table follow up: register bank select
2020 Oct 09
2
GlobalISel round table follow up: register bank select
...r the operands >>>> when considering which of the mappings to apply to the current >>>> instruction. In our architecture we have one register bank >>>> dedicated to pointers and another one for anything else. We often >>>> see code where we have a G_PTR_ADD with a constant. Since the >>>> constant is not a pointer, we put it on the other register bank. We >>>> could put it on the address regbank and do provide alternative >>>> mappings for that, but since the greedy algorithm doesn't actually >>>&gt...