Metrics for simple fma test

analyze1.0ms (0%)

Algorithm

1×

Search

Probability	Valid	Unknown	Precondition	Infinite	Domain	Can't	Iter
0%	0%	99.9%	0.1%	0%	0%	0%	0
100%	99.9%	0%	0.1%	0%	0%	0%	1

Compiler

Compiled 17 to 9 computations (47.1% saved)

Precisions

Click to see histograms. Total time spent on operations: 0.0ms

ival-sub: 0.0ms (0% of total)

ival-add: 0.0ms (0% of total)

ival-mult: 0.0ms (0% of total)

const: 0.0ms (0% of total)

backward-pass: 0.0ms (0% of total)

sample1.9s (64.5%)

Results

697.0ms	2947×	1	valid
586.0ms	3103×	2	valid
196.0ms	2116×	0	valid
24.0ms	90×	3	valid

Precisions

Click to see histograms. Total time spent on operations: 967.0ms

ival-add: 326.0ms (33.7% of total)

backward-pass: 313.0ms (32.4% of total)

ival-mult: 224.0ms (23.2% of total)

ival-sub: 89.0ms (9.2% of total)

const: 14.0ms (1.4% of total)

Bogosity

preprocess338.0ms (11.5%)

Algorithm

1×	egg-herbie

Rules

911×	`fmm-def`
788×	`unsub-neg`
476×	`sub-neg`
379×	`fma-define`
287×	`distribute-neg-in`

Iterations

Useful iterations: 2 (0.0ms)

Iter	Nodes	Cost
0	52	553
1	137	541
2	378	11
3	1556	11
4	3637	11
5	5091	11
6	6021	11
7	7055	11
8	7693	11
9	7777	11
10	7777	11
11	7885	11
12	7972	11
13	7993	11

Stop Event

1×	node limit

Calls

Call 1

Inputs
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))
(-.f64 (fma.f64 (neg.f64 x) y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 (neg.f64 x) y) z)))
(-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x (neg.f64 y)) z)))
(-.f64 (fma.f64 x y (neg.f64 z)) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) (neg.f64 z))))
(neg.f64 (-.f64 (fma.f64 (neg.f64 x) y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 (neg.f64 x) y) z))))
(neg.f64 (-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x (neg.f64 y)) z))))
(neg.f64 (-.f64 (fma.f64 x y (neg.f64 z)) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) (neg.f64 z)))))
(-.f64 (fma.f64 y x z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 y x) z)))
(-.f64 (fma.f64 z y x) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 z y) x)))
(-.f64 (fma.f64 x z y) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x z) y)))

Outputs
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 (neg.f64 x) y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 (neg.f64 x) y) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x (neg.f64 y)) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 x y (neg.f64 z)) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) (neg.f64 z))))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(neg.f64 (-.f64 (fma.f64 (neg.f64 x) y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 (neg.f64 x) y) z))))
(neg.f64 (-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (fma.f64 x (neg.f64 y) z))))
(+.f64 (-.f64 (.f64 x y) z) (+.f64 z (-.f64 #s(literal 1 binary64) (.f64 x y))))
#s(literal 1 binary64)
(neg.f64 (-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x (neg.f64 y)) z))))
(neg.f64 (-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (fma.f64 x (neg.f64 y) z))))
(+.f64 (-.f64 (.f64 x y) z) (+.f64 z (-.f64 #s(literal 1 binary64) (.f64 x y))))
#s(literal 1 binary64)
(neg.f64 (-.f64 (fma.f64 x y (neg.f64 z)) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) (neg.f64 z)))))
(neg.f64 (-.f64 (fma.f64 x (neg.f64 y) z) (+.f64 #s(literal 1 binary64) (fma.f64 x (neg.f64 y) z))))
(+.f64 (-.f64 (.f64 x y) z) (+.f64 z (-.f64 #s(literal 1 binary64) (.f64 x y))))
#s(literal 1 binary64)
(-.f64 (fma.f64 y x z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 y x) z)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 z y x) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 z y) x)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)
(-.f64 (fma.f64 x z y) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x z) y)))
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (fma.f64 x y z)))
#s(literal -1 binary64)

Symmetry

(abs x)

(abs y)

(abs z)

(sort x y z)

explain274.0ms (9.3%)

FPErrors

Click to see full error table

Ground Truth	Overpredictions	Example	Example	Subexpression
217	3	(3.056883780299492e-66 2.0394030565318493e-13 117255.08413958021)	-	`(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))`
0	0	-	-	`x`
0	0	-	-	`y`
0	0	-	-	`(*.f64 x y)`
0	0	-	-	`z`
0	0	-	-	`(+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z))`
0	0	-	-	`#s(literal 1 binary64)`
0	0	-	-	`(fma.f64 x y z)`
0	0	-	-	`(+.f64 (*.f64 x y) z)`

Results

109.0ms	240×	2	valid
30.0ms	196×	1	valid
7.0ms	76×	0	valid

Compiler

Compiled 97 to 27 computations (72.2% saved)

Precisions

Click to see histograms. Total time spent on operations: 88.0ms

ival-add: 61.0ms (69.2% of total)

backward-pass: 12.0ms (13.6% of total)

ival-mult: 7.0ms (7.9% of total)

ival-sub: 7.0ms (7.9% of total)

const: 1.0ms (1.1% of total)

eval0.0ms (0%)

Compiler: Compiled 15 to 8 computations (46.7% saved)

prune1.0ms (0%)

Alt Table

Click to see full alt table

Status	Accuracy	Program
▶	100.0%	#s(literal -1 binary64)

Compiler

Compiled 4 to 4 computations (0% saved)

simplify5.0ms (0.2%)

Algorithm

1×	egg-herbie

Iterations

Useful iterations: 0 (0.0ms)

Iter	Nodes	Cost
0	1	1

Stop Event

1×

saturated

Calls

Call 1

Inputs
#s(literal -1 binary64)

Outputs
#s(literal -1 binary64)

localize11.0ms (0.4%)

Results

8.0ms

256×

valid

Compiler

Compiled 5 to 5 computations (0% saved)

Precisions

Click to see histograms. Total time spent on operations: 0.0ms

const: 0.0ms (0% of total)

backward-pass: 0.0ms (0% of total)

eval0.0ms (0%)

Compiler: Compiled 3 to 3 computations (0% saved)

prune1.0ms (0%)

Pruning

1 alts after pruning (0 fresh and 1 done)

	Kept	Total
New	0	0
Fresh	0	0
Picked	1	1
Done	0	0
Total	1	1

Accuracy

100.0%

Counts

1 → 1

Alt Table

Click to see full alt table

Status	Accuracy	Program
✓	100.0%	#s(literal -1 binary64)

Compiler

Compiled 20 to 14 computations (30% saved)

regimes6.0ms (0.2%)

Accuracy

Total -52.6b remaining (-∞%)

Threshold costs -52.6b (-∞%)

Counts

2 → 1

Calls

Call 1

Inputs
#s(literal -1 binary64)
(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))

Outputs
#s(literal -1 binary64)

Calls

4 calls:

2.0ms	z
1.0ms	y
1.0ms	x
1.0ms	(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))

Results

Accuracy	Segments	Branch
100.0%	1	`x`
100.0%	1	`y`
100.0%	1	`z`
100.0%	1	`(-.f64 (fma.f64 x y z) (+.f64 #s(literal 1 binary64) (+.f64 (*.f64 x y) z)))`

Compiler

Compiled 27 to 18 computations (33.3% saved)

simplify1.0ms (0%)

Algorithm

1×	egg-herbie

Iterations

Useful iterations: 0 (0.0ms)

Iter	Nodes	Cost
0	1	1

Stop Event

1×

saturated

Calls

Call 1

Inputs
#s(literal -1 binary64)

Outputs
#s(literal -1 binary64)

soundness347.0ms (11.8%)

Rules

911×	`fmm-def`
788×	`unsub-neg`
476×	`sub-neg`
379×	`fma-define`
287×	`distribute-neg-in`

Iterations

Useful iterations: 2 (0.0ms)

Iter	Nodes	Cost
0	52	553
1	137	541
2	378	11
3	1556	11
4	3637	11
5	5091	11
6	6021	11
7	7055	11
8	7693	11
9	7777	11
10	7777	11
11	7885	11
12	7972	11
13	7993	11

Stop Event

1×	done
1×	node limit

Compiler

Compiled 62 to 21 computations (66.1% saved)

preprocess55.0ms (1.9%)

Remove

(sort x y z)

(abs z)

(abs y)

(abs x)

Compiler

Compiled 110 to 98 computations (10.9% saved)

simple fma test

analyze1.0ms (0%)

sample1.9s (64.5%)

preprocess338.0ms (11.5%)

explain274.0ms (9.3%)

eval0.0ms (0%)

prune1.0ms (0%)

simplify5.0ms (0.2%)

localize11.0ms (0.4%)

eval0.0ms (0%)

prune1.0ms (0%)

regimes6.0ms (0.2%)

simplify1.0ms (0%)

soundness347.0ms (11.8%)

preprocess55.0ms (1.9%)

end0.0ms (0%)

Profiling