Fix potentially incorrect floating-point arithmetic on MinGW

[BrettDong]

Summary

Bugfixes "Fix potentially incorrect floating-point arithmetic on MinGW"

Purpose of change

The following three tests are failing on MinGW cross-compile test workflow:

• stamina burn for movement
• player_morale_decay
• throwing_skill_impact_test

The symptom is that floating point arithmetic in these three failing tests is producing different numeric results to that on other platforms.

Describe the solution

• player_morale_decay and throwing_skill_impact_test succeeds if I reset x87 FPU and SSE states at the beginning of main function. This seems to be a MinGW bug where floating point math package states are not properly initialised on a Windows thread: https://blog.zaita.com/mingw64-compiler-bug/
• stamina burn for movement failure is solved by adding std::round() to round a double number to int in burden_player, without which on MinGW 45000 * 1.01 => 45449.

Describe alternatives you've considered

I've tried to call standard library function std::fesetround( FE_TONEAREST ) to set floating point rounding mode to round to nearest representable value, but this does not actually solve the problem, the tests were still failing.

Compiling with -msse -mfpmath=sse -march=pentium4 mentioned by lemire in his blog https://lemire.me/blog/2020/06/26/gcc-not-nearest/ might help, but I haven't tried yet.

Testing

Expect to see MinGW cross-compile test workflow to pass.

Additional context

[BrettDong]

Expected output for the first two assertions:

Capacity: 45000000mg | Burden Proportion: 1.01 | Units: 45450
Burn Ratio (1): 15
Burn Ratio (2): 16
Stamina Multiplier: 1
Burn Ratio (3): 16
Stamina Move Cost Modifier: 1
Stamina:  10000 -> 9984

../tests/char_stamina_test.cpp:295: FAILED:
  CHECK( burdened_burn_rate( dummy, move_mode_walk, 1.01 ) == normal_burn_rate + 10 )
with expansion:
  16 == 16

---

Capacity: 45000000mg | Burden Proportion: 1.02 | Units: 45900
Burn Ratio (1): 15
Burn Ratio (2): 17
Stamina Multiplier: 1
Burn Ratio (3): 17
Stamina Move Cost Modifier: 1
Stamina:  10000 -> 9983
../tests/char_stamina_test.cpp:296: FAILED:
  CHECK( burdened_burn_rate( dummy, move_mode_walk, 1.02 ) == normal_burn_rate + 20 )
with expansion:
  17 == 17

[actual-nh]

Different rounding (again!)?

[BrettDong]

Yes it seems to be a rounding problem:

Capacity: 45000000mg | Burden Proportion: 1.01 | Units: 45449

45000 * 1.01 should be 45450 but was 45449 in MinGW workflow .

[BrettDong]

Setting rounding mode to nearest representable value doesn't help.

std::fesetround( FE_TONEAREST );

[BrettDong]

Stamina test is fixed by std::round. Next to fix is the failing morale decay test.

[BrettDong]

Now throwing test is failing:

 -------------------------------------------------------------------------------
 throwing_skill_impact_test
   hi_skill_basestats_rock
 -------------------------------------------------------------------------------
 ../tests/throwing_test.cpp:224
 ...............................................................................
 
 ../tests/throwing_test.cpp:144: FAILED:
   CHECK( dmg_thresh_met )
 with expansion:
   false
 with messages:
   Monster: 'mon_zombie' Item: 'rock
   Range: 5 Pstats: STR: 8 DEX: 8 PER: 8 SKL: 10
   Total throws: 10000
   Ratio: 100%
   Hit Lower: 99.9235% Hit Upper: 100%
   Hit Thresh: 90% - 110%
   Adj Wald error: 0.000764809
   Avg total damage: 12.8064
   Dmg Lower: 12.7758 Dmg Upper: 12.837
   Dmg Thresh: 13 - 23
   Margin of error: 0.000764809

Expected numbers should be:

  Monster: 'mon_zombie' Item: 'rock
  Range: 5 Pstats: STR: 8 DEX: 8 PER: 8 SKL: 10
  Total throws: 100
  Ratio: 100%
  Hit Lower: 93.2609% Hit Upper: 100%
  Hit Thresh: 90% - 110%
  Adj Wald error: 0.0673911
  Avg total damage: 13.99
  Dmg Lower: 13.6262 Dmg Upper: 14.3538
  Dmg Thresh: 13 - 23
  Margin of error: 0.0673911

[kevingranade]

I'm very concerned that this leaves the door open for other errors to persist in the codebase that aren't calling std::round(), this behavior needs to be the same in both builds (though if this is an incremental step to getting things unblocked disregard).

[BrettDong]

Even with fpreset this test is still failing, so I'm adding rounding at the moment.