Explorar o código

Recalibrate gemm application with the default 4*960 z axis size

Samuel Thibault %!s(int64=10) %!d(string=hai) anos
pai
achega
fa93ac1390

+ 12 - 12
tools/perfmodels/sampling/codelets/44/starpu_dgemm_gemm.attila

@@ -35,9 +35,9 @@
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	1.325410e+05   	6.551456e+03   	1.020565e+07   	1.355972e+12   	77
-19d944cc	302284800      	3.538944e+10   	3.351379e+06   	5.140990e+04   	2.681103e+08   	8.987506e+14   	80
-9e87ad36	134348800      	1.048576e+10   	1.018200e+06   	1.999744e+04   	8.145602e+07   	8.297054e+13   	80
+492beed5	66355200       	7.077888e+09   	6.739553e+05   	1.180373e+04   	5.391643e+07   	3.634841e+13   	80
+0b0b0ce8	7372800        	2.621440e+08   	2.919185e+04   	3.106016e+03   	2.335348e+06   	6.894489e+10   	80
+4220e23d	29491200       	2.097152e+09   	2.058556e+05   	4.934163e+03   	1.646845e+07   	3.392071e+12   	80
 
 ####################
 # COMB_2
@@ -67,9 +67,9 @@ a904baa5	33587200       	1.310720e+09   	1.325410e+05   	6.551456e+03   	1.02056
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.299792e+03   	3.590016e+02   	5.511784e+05   	2.934534e+09   	104
-19d944cc	302284800      	3.538944e+10   	1.430442e+05   	1.747241e+04   	7.867431e+06   	1.142181e+12   	55
-9e87ad36	134348800      	1.048576e+10   	3.517587e+04   	1.024947e+02   	3.623115e+06   	1.274473e+11   	103
+492beed5	66355200       	7.077888e+09   	2.313178e+04   	3.749098e+01   	2.451969e+06   	5.671856e+10   	106
+0b0b0ce8	7372800        	2.621440e+08   	1.039822e+03   	4.572723e+01   	9.982293e+04   	1.039988e+08   	96
+4220e23d	29491200       	2.097152e+09   	7.017044e+03   	9.150160e+00   	7.367896e+05   	5.170094e+09   	105
 
 ####################
 # COMB_1
@@ -99,9 +99,9 @@ a904baa5	33587200       	1.310720e+09   	5.299792e+03   	3.590016e+02   	5.51178
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.301676e+03   	3.783139e+02   	5.460726e+05   	2.909842e+09   	103
-19d944cc	302284800      	3.538944e+10   	1.433132e+05   	1.648772e+04   	7.738910e+06   	1.123767e+12   	54
-9e87ad36	134348800      	1.048576e+10   	3.516032e+04   	6.444220e+01   	3.551193e+06   	1.248615e+11   	101
+492beed5	66355200       	7.077888e+09   	2.311999e+04   	3.792305e+01   	2.427599e+06   	5.612623e+10   	105
+0b0b0ce8	7372800        	2.621440e+08   	1.036523e+03   	7.589914e+00   	8.810443e+04   	9.132713e+07   	85
+4220e23d	29491200       	2.097152e+09   	7.017425e+03   	3.431116e+01   	7.298122e+05   	5.121525e+09   	104
 
 ####################
 # COMB_0
@@ -131,7 +131,7 @@ a904baa5	33587200       	1.310720e+09   	5.301676e+03   	3.783139e+02   	5.46072
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.174822e+03   	5.540579e+01   	5.433563e+05   	2.812095e+09   	105
-19d944cc	302284800      	3.538944e+10   	1.362048e+05   	1.589418e+04   	6.401628e+06   	8.838060e+11   	47
-9e87ad36	134348800      	1.048576e+10   	3.562387e+04   	2.173490e+03   	3.598011e+06   	1.286522e+11   	101
+492beed5	66355200       	7.077888e+09   	2.312215e+04   	3.532854e+01   	2.427826e+06   	5.613668e+10   	105
+0b0b0ce8	7372800        	2.621440e+08   	1.036998e+03   	4.343669e+01   	1.078478e+05   	1.120342e+08   	104
+4220e23d	29491200       	2.097152e+09   	7.010229e+03   	3.598880e+01   	7.360740e+05   	5.160183e+09   	105
 

+ 12 - 12
tools/perfmodels/sampling/codelets/44/starpu_dgemm_gemm.mirage

@@ -35,9 +35,9 @@
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	1.389650e+05   	2.221518e+03   	1.111720e+07   	1.545297e+12   	80
-19d944cc	302284800      	3.538944e+10   	3.522057e+06   	7.908573e+04   	2.852866e+08   	1.005302e+15   	81
-9e87ad36	134348800      	1.048576e+10   	1.082134e+06   	2.331687e+04   	8.657069e+07   	9.372454e+13   	80
+0b0b0ce8	7372800        	2.621440e+08   	2.783376e+04   	1.016266e+03   	2.254534e+06   	6.283582e+10   	81
+492beed5	66355200       	7.077888e+09   	7.068870e+05   	1.582112e+04   	5.725785e+07   	4.049511e+13   	81
+4220e23d	29491200       	2.097152e+09   	2.135531e+05   	4.787239e+03   	1.729780e+07   	3.695855e+12   	81
 
 ####################
 # COMB_2
@@ -67,9 +67,9 @@ a904baa5	33587200       	1.310720e+09   	1.389650e+05   	2.221518e+03   	1.11172
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.219321e+03   	1.052627e+02   	5.375900e+05   	2.806996e+09   	103
-19d944cc	302284800      	3.538944e+10   	1.296586e+05   	9.187674e+03   	1.102098e+07   	1.436140e+12   	85
-9e87ad36	134348800      	1.048576e+10   	3.558879e+04   	1.880737e+02   	3.736823e+06   	1.329927e+11   	105
+0b0b0ce8	7372800        	2.621440e+08   	1.040745e+03   	1.710737e+01   	1.040745e+05   	1.083442e+08   	100
+492beed5	66355200       	7.077888e+09   	2.322675e+04   	6.514638e+01   	2.438809e+06   	5.664606e+10   	105
+4220e23d	29491200       	2.097152e+09   	7.042883e+03   	4.736092e+01   	7.395027e+05   	5.208467e+09   	105
 
 ####################
 # COMB_0
@@ -99,9 +99,9 @@ a904baa5	33587200       	1.310720e+09   	5.219321e+03   	1.052627e+02   	5.37590
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.294629e+03   	1.366974e+02   	5.506414e+05   	2.917386e+09   	104
-19d944cc	302284800      	3.538944e+10   	1.293071e+05   	8.164654e+03   	1.137902e+07   	1.477255e+12   	88
-9e87ad36	134348800      	1.048576e+10   	3.559057e+04   	1.877150e+02   	3.772600e+06   	1.342727e+11   	106
+0b0b0ce8	7372800        	2.621440e+08   	1.057967e+03   	4.209841e+01   	1.057967e+05   	1.121067e+08   	100
+492beed5	66355200       	7.077888e+09   	2.322865e+04   	8.861437e+01   	2.439008e+06   	5.665569e+10   	105
+4220e23d	29491200       	2.097152e+09   	7.053091e+03   	5.410169e+01   	7.405746e+05   	5.223647e+09   	105
 
 ####################
 # COMB_1
@@ -131,7 +131,7 @@ a904baa5	33587200       	1.310720e+09   	5.294629e+03   	1.366974e+02   	5.50641
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	33587200       	1.310720e+09   	5.223975e+03   	8.128456e+01   	5.380694e+05   	2.811542e+09   	103
-19d944cc	302284800      	3.538944e+10   	1.298085e+05   	9.721517e+03   	1.129334e+07   	1.474194e+12   	87
-9e87ad36	134348800      	1.048576e+10   	3.556314e+04   	1.481994e+02   	3.734130e+06   	1.327997e+11   	105
+0b0b0ce8	7372800        	2.621440e+08   	1.050834e+03   	7.708100e+01   	1.019309e+05   	1.076889e+08   	97
+492beed5	66355200       	7.077888e+09   	2.323864e+04   	5.619683e+01   	2.440057e+06   	5.670394e+10   	105
+4220e23d	29491200       	2.097152e+09   	7.040571e+03   	3.296604e+01   	7.392600e+05   	5.204926e+09   	105
 

+ 12 - 12
tools/perfmodels/sampling/codelets/44/starpu_sgemm_gemm.attila

@@ -35,9 +35,9 @@
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	6.437897e+04   	1.392917e+03   	4.699665e+06   	3.027012e+11   	73
-19d944cc	151142400      	3.538944e+10   	1.651467e+06   	2.533878e+04   	1.321174e+08   	2.182388e+14   	80
-9e87ad36	67174400       	1.048576e+10   	4.988315e+05   	1.083517e+04   	3.990652e+07   	1.991602e+13   	80
+492beed5	33177600       	7.077888e+09   	3.328725e+05   	1.185902e+04   	2.563119e+07   	8.542747e+12   	77
+0b0b0ce8	3686400        	2.621440e+08   	1.421718e+04   	3.409134e+02   	9.098993e+05   	1.294364e+10   	64
+4220e23d	14745600       	2.097152e+09   	1.008105e+05   	2.361630e+03   	8.064841e+06   	8.134670e+11   	80
 
 ####################
 # COMB_0
@@ -67,9 +67,9 @@ a904baa5	16793600       	1.310720e+09   	6.437897e+04   	1.392917e+03   	4.69966
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.518696e+03   	2.456600e+02   	3.659444e+05   	1.293924e+09   	104
-19d944cc	151142400      	3.538944e+10   	5.651725e+04   	3.211294e+03   	4.973518e+06   	2.819970e+11   	88
-9e87ad36	67174400       	1.048576e+10   	2.754677e+04   	3.854974e+02   	2.919957e+06   	8.045113e+10   	106
+492beed5	33177600       	7.077888e+09   	1.123499e+04   	6.785566e+01   	1.190909e+06   	1.338033e+10   	106
+0b0b0ce8	3686400        	2.621440e+08   	6.738679e+02   	4.393713e+01   	6.873452e+04   	4.651489e+07   	102
+4220e23d	14745600       	2.097152e+09   	5.557425e+03   	3.241733e+02   	5.835297e+05   	3.253957e+09   	105
 
 ####################
 # COMB_2
@@ -99,9 +99,9 @@ a904baa5	16793600       	1.310720e+09   	3.518696e+03   	2.456600e+02   	3.65944
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.519726e+03   	2.643171e+02   	3.660515e+05   	1.295666e+09   	104
-19d944cc	151142400      	3.538944e+10   	5.631482e+04   	2.073911e+03   	4.955704e+06   	2.794581e+11   	88
-9e87ad36	67174400       	1.048576e+10   	2.756588e+04   	3.850160e+02   	2.894417e+06   	7.980271e+10   	105
+492beed5	33177600       	7.077888e+09   	1.123077e+04   	9.504466e+01   	1.179231e+06   	1.324463e+10   	105
+0b0b0ce8	3686400        	2.621440e+08   	6.672056e+02   	3.376608e+01   	6.805497e+04   	4.552295e+07   	102
+4220e23d	14745600       	2.097152e+09   	5.553764e+03   	3.500896e+02   	5.831453e+05   	3.251521e+09   	105
 
 ####################
 # COMB_1
@@ -131,7 +131,7 @@ a904baa5	16793600       	1.310720e+09   	3.519726e+03   	2.643171e+02   	3.66051
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.490259e+03   	2.149136e+02   	3.699675e+05   	1.296178e+09   	106
-19d944cc	151142400      	3.538944e+10   	5.827658e+04   	5.717882e+03   	5.477999e+06   	3.223123e+11   	94
-9e87ad36	67174400       	1.048576e+10   	2.758960e+04   	5.859871e+01   	2.896907e+06   	7.992487e+10   	105
+492beed5	33177600       	7.077888e+09   	1.124174e+04   	2.629960e+01   	1.180383e+06   	1.326963e+10   	105
+0b0b0ce8	3686400        	2.621440e+08   	6.002221e+02   	2.259043e+01   	6.242310e+04   	3.752080e+07   	104
+4220e23d	14745600       	2.097152e+09   	5.577722e+03   	1.615194e+02   	5.912385e+05   	3.300529e+09   	106
 

+ 12 - 12
tools/perfmodels/sampling/codelets/44/starpu_sgemm_gemm.mirage

@@ -35,9 +35,9 @@
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	6.707933e+04   	7.224956e+02   	5.433426e+06   	3.645129e+11   	81
-19d944cc	151142400      	3.538944e+10   	1.796711e+06   	3.314021e+04   	1.455336e+08   	2.615707e+14   	81
-9e87ad36	67174400       	1.048576e+10   	5.407828e+05   	9.552098e+03   	4.326263e+07   	2.340299e+13   	80
+0b0b0ce8	3686400        	2.621440e+08   	1.352609e+04   	3.616534e+02   	1.082087e+06   	1.464687e+10   	80
+492beed5	33177600       	7.077888e+09   	3.550396e+05   	8.949994e+03   	2.840317e+07   	1.009066e+13   	80
+4220e23d	14745600       	2.097152e+09   	1.078112e+05   	1.983800e+03   	8.624897e+06   	9.301755e+11   	80
 
 ####################
 # COMB_1
@@ -67,9 +67,9 @@ a904baa5	16793600       	1.310720e+09   	6.707933e+04   	7.224956e+02   	5.43342
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.517309e+03   	1.972716e+02   	3.658001e+05   	1.290679e+09   	104
-19d944cc	151142400      	3.538944e+10   	5.914394e+04   	3.899693e+02   	6.210114e+06   	3.673066e+11   	105
-9e87ad36	67174400       	1.048576e+10   	2.761071e+04   	3.567319e+02   	2.899125e+06   	8.006027e+10   	105
+0b0b0ce8	3686400        	2.621440e+08   	6.589631e+02   	8.406511e+00   	6.787320e+04   	4.473321e+07   	103
+492beed5	33177600       	7.077888e+09   	1.151398e+04   	9.050114e+01   	1.220482e+06   	1.405348e+10   	106
+4220e23d	14745600       	2.097152e+09   	5.574713e+03   	3.353004e+02   	5.909196e+05   	3.306125e+09   	106
 
 ####################
 # COMB_0
@@ -99,9 +99,9 @@ a904baa5	16793600       	1.310720e+09   	3.517309e+03   	1.972716e+02   	3.65800
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.518071e+03   	2.387327e+02   	3.658794e+05   	1.293117e+09   	104
-19d944cc	151142400      	3.538944e+10   	5.916943e+04   	4.366031e+02   	6.212790e+06   	3.676273e+11   	105
-9e87ad36	67174400       	1.048576e+10   	2.760580e+04   	5.909500e+01   	2.926215e+06   	8.078087e+10   	106
+0b0b0ce8	3686400        	2.621440e+08   	6.663664e+02   	8.616537e+01   	6.796937e+04   	4.604980e+07   	102
+492beed5	33177600       	7.077888e+09   	1.150036e+04   	8.404527e+01   	1.207538e+06   	1.388786e+10   	105
+4220e23d	14745600       	2.097152e+09   	5.579034e+03   	3.672012e+02   	5.857985e+05   	3.282348e+09   	105
 
 ####################
 # COMB_2
@@ -131,7 +131,7 @@ a904baa5	16793600       	1.310720e+09   	3.518071e+03   	2.387327e+02   	3.65879
 # a		b		c
 nan            	nan            	nan            
 # hash		size		flops		mean (us)	dev (us)	sum		sum2		n
-a904baa5	16793600       	1.310720e+09   	3.497709e+03   	1.653655e+02   	3.672595e+05   	1.287438e+09   	105
-19d944cc	151142400      	3.538944e+10   	5.899970e+04   	4.129676e+02   	6.194969e+06   	3.655192e+11   	105
-9e87ad36	67174400       	1.048576e+10   	2.761335e+04   	3.211454e+02   	2.899402e+06   	8.007303e+10   	105
+0b0b0ce8	3686400        	2.621440e+08   	6.181769e+02   	5.174143e+01   	6.181769e+04   	3.848198e+07   	100
+492beed5	33177600       	7.077888e+09   	1.148096e+04   	7.289415e+01   	1.205501e+06   	1.384086e+10   	105
+4220e23d	14745600       	2.097152e+09   	5.580581e+03   	3.970717e+02   	5.859610e+05   	3.286558e+09   	105