* NE10 Library : ReleaseNote.txt
*/
NE10 SIMD LIBRARY - Release Note
-LAST UPDATED ON: 1 / MAR / 2012
+LAST UPDATED ON: 10 / APR / 2012
========
Contents
3. Installation
3-a. Requirements
3-b. Alternative Approach
+ 4. Changelog
+ 4-a. r1.0_beta
==========
2-a. Product release status
---------------------------
-Version 0.9 dev
+Version 1.0 beta
-Implemented functions have been tested and are believed to be working, but the
-complete set of functions planned for the initial release are not in this
-release.
+The set of functions planned for this release are in place. However some issues
+remain where their intended behaviour diverges from the planned specification.
+
+The aim is for all functions to operate such that it is valid for the source
+parameter to match the destination. This requirement may be relaxed where it
+is impractical to implement or it is not possible to provide that behaviour
+efficiently. Currently some functions are not capable of fulfilling that
+requirement:
2-b. Functions included
---------------------------
c) Operations on Vectors:
- len_vec2f, len_vec3f, len_vec4f
- normalize_vec2f, normalize_vec3f, normalize_vec4f
-
+ abs_vec2f, abs_vec3f, abs_vec4f,
+ addc_vec2f, addc_vec3f, addc_vec4f,
+ add_vec2f, add_vec3f, add_vec4f,
+ divc_vec2f, divc_vec3f, divc_vec4f,
+ dot_vec2f, dot_vec3f, dot_vec4f
+ len_vec2f, len_vec3f, len_vec4f,
+ mlac_vec2f, mlac_vec3f, mlac_vec4f,
+ mulc_vec2f, mulc_vec3f, mulc_vec4f,
+ normalize_vec2f, normalize_vec3f, normalize_vec4f,
+ rsbc_vec2f, rsbc_vec3f, rsbc_vec4f,
+ setc_vec2f, setc_vec3f, setc_vec4f,
+ subc_vec2f, subc_vec3f, subc_vec4f,
+ sub_vec2f, sub_vec3f, sub_vec4f,
+ vdiv_vec2f, vdiv_vec3f, vdiv_vec4f,
+ vmla_vec2f, vmla_vec3f, vmla_vec4f,
+ vmul_vec2f, vmul_vec3f, vmul_vec4f,
+ cross_vec3f
+
+ d) Matrix operations:
+
+ addmat_2x2f, addmat_3x3f, addmat_4x4f,
+ detmat_2x2f, detmat_3x3f, detmat_4x4f,
+ divmat_2x2f, divmat_3x3f, divmat_4x4f,
+ identitymat_2x2f, identitymat_3x3f, identitymat_4x4f,
+ invmat_2x2f, invmat_3x3f, invmat_4x4f,
+ mulcmatvec_2x2f, mulcmatvec_3x3f, mulcmatvec_4x4f,
+ mulmat_2x2f, mulmat_3x3f, mulmat_4x4f,
+ multrans_mat2x2f, multrans_mat3x3f, multrans_mat4x4f,
+ setmat_2x2f, setmat_3x3f, setmat_4x4f,
+ submat_2x2f, submat_3x3f, submat_4x4f,
+ transmat_2x2f, transmat_3x3f, transmat_4x4f,
2-c. Test cases and results
---------------------------
as part of this library and can be used to verify and benchmark these functions
on a target platform.
-
-
===============
3. Installation
===============
incorporated (licensing conflicts permitting) within other projects as is.
Details of how to do this are too project specific to detail here.
+============
+4. Changelog
+============
+
+4-a. r1.0_beta
+
+ * Updated AOSP Makefile, cleaned native Makefile
+ * Adding new files to the AOSP build
+ * Made the default makefile a little more readable
+ * New functions: Matrix transpose and identity matrix routines.
+ * New functions: Matrix inversion routines.
+ * New functions: Matrix determinant routines.
+ * New functions: Matrix-vector multiplication routines.
+ * New functions: Matrix multiplication routines.
+ * New functions: Matrix multiplication routines.
+ * New functions: Matrix addition and subtraction.
+ * New functions: Cross product routine.
+ * New functions: Dot product routines.
+ * New functions: vectorized mla routines.
+ * New functions: vectorized division routines.
+ * New functions: vectorized abs routine.
+ * New functions: vector-sub routines.
+ * New functions: vector-add routines.
+ * Added the disclaimer:
+ Each function is implemented in C, ARM Assembly and NEON code as a
+ basis for comparison. Assembly versions, while efficient, are not
+ intended as best-practice examples. NEON versions are intended as
+ best-practice examples.