Bug 113 - FCVT unit tests have too great a dynamic range
Summary: FCVT unit tests have too great a dynamic range
Status: RESOLVED FIXED
Alias: None
Product: Libre-SOC's first SoC
Classification: Unclassified
Component: ALU (including IEEE754 16/32/64-bit FPU) (show other bugs)
Version: unspecified
Hardware: PC Linux
: --- enhancement
Assignee: Luke Kenneth Casson Leighton
URL:
Depends on:
Blocks: 48
  Show dependency tree / graph
 
Reported: 2019-07-21 10:31 BST by Luke Kenneth Casson Leighton
Modified: 2020-12-04 20:36 GMT (History)
1 user (show)

See Also:
NLnet milestone: NLnet.2019.02.012
total budget (EUR) for completion of task and all subtasks: 250
budget (EUR) for this task, excluding subtasks' budget: 250
parent task for budget allocation: 48
child tasks for budget allocation:
The table of payments (in EUR) for this task; TOML format:
lkcl={amount=250, paid=2020-03-12}

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Luke Kenneth Casson Leighton 2019-07-21 10:31:20 BST 
the range on the higher-sized to lower-sized randomised unit tests
give input values that are so large, they're almost exclusively
converted to +INF (or hit the -ve limit - 0 for unsigned, -INF for signed)

better tests are needed.
Comment 1 Luke Kenneth Casson Leighton 2019-08-01 01:00:34 BST 
this for:

* fcvt float2int (large FP to smaller int)
* int2float (for FP16/32/64 to FP16)
* fcvt FP to FP (large FP to smaller FP)

int input is fairly easy to deal with: test within "close" range.
FP however needs *FP* numbers to be generated that will (or will
almost) fit into the target int range.
Comment 2 Luke Kenneth Casson Leighton 2020-02-10 15:10:56 GMT 
https://git.libre-riscv.org/?p=ieee754fpu.git;a=commitdiff;h=6ea789c9e5a54e7dfbfbd96cb0c620e52b01c2a4

added appx-range within chosen int (-INTMAX*2 to +INTMAX*2) now need one
that is very close to the limits.
Comment 3 Luke Kenneth Casson Leighton 2020-02-11 22:40:39 GMT 
also added floating-point conversion tests which convert numbers just at
the limits of accuracy (or INF) of the target number.