Episode 5.2 - Scalar Tuning and General Optimization
Table of Contents:
00:15 - Scalar optimizations
00:32 - Strength reduction
00:44 - Devision vs Reciprocal Multiplication
01:19 - Algebraic transformation for Strength Reduction
01:32 - Strength reduction and transcendental math
01:37 - pow vs sqrt
01:58 - log/exp vs log2/exp2
02:18 - Interpolation of power law function approximation
02:46 - Precision control
02:52 - -fimf-precision
03:11 - -fimf-max-error and -fimf-accuracy-bits
03:30 - -fimf-domain-exclusion
03:54 - https://software.intel.com/en-us/arti...
03:58 - Floating-point semantics
04:29 - Default: -fp-model fast=1
04:46 - -fp-model srtict
05:03 - Rounding of intermediate results
05:23 - Precision of constants and functions
05:35 - Integer constants
05:51 - Avoid unnecessary type conversion
06:11 - Integer overflow example
06:29 - Floating point constants
06:51 - Performance penalties of unnecessary type conversion
07:26 - Precision of functions
07:33 - Precision of transcendental functions
08:17 - Overload of math functions in C++
09:26 - In the next episode...