Appendix¶

SIMD Mode¶

Many of the builtins for DSD operations have a SIMD (single instruction, multiple data) mode, in which multiple operations can be performed in a single cycle. Under appropriate conditions, these builtins will automatically execute in SIMD mode when operating on DSDs, if possible.

In particular, builtins can only operate at their full SIMD width if no bank conflicts occur when fetching the operands from memory. The 48 KB of memory in a PE are laid out into 8 banks of 6 KB each. Each successive 16 bits are located in successive banks. In a single cycle, the PE can perform two 32-bit reads and one 32-bit write. However, the reads must occur from separate banks. More specifically, if the 8 banks are numbered 0 to 7, then the bank IDs bank_1 and bank_2 of the two reads must be such that bank_1 % 4 != bank_2 % 4.

For best results in avoiding bank conflicts with SIMD operations, the operand addresses should be 32-bit aligned, and (src0_addr % 8) == ((src1_addr + 4) % 8), where src0_addr and src1_addr are the addresses of the operands. Dumping the ELF file’s symbol table with the --sym option of cs_readelf provides addresses and banking information for all symbols in a compiled CSL program, and can be useful for determining if bank conflicts may occur.

Additionally, if the DSD operands have non-contiguous strided accesses, the SIMD width may be limited:

Strides of 0 and 1 can operate at full SIMD width.
Strides such that stride % 8 is 2, 3, 5, or 6 can operate at full SIMD width.
Unless stride is 1, strides such that stride % 8 is 1 or 7 is limited to two operations per cycle, or a SIMD width of 2.
Unless stride is 0, strides such that stride % 8 is 0 is limited to one operation per cycle, or a SIMD width of 1.
Strides such that stride % 8 is 4 are limited to two operations per cycle, or a SIMD width of 4.

The maximum width of builtins which can operate in SIMD mode are given in the table below.

Builtin	WSE-2 SIMD Width	WSE-3 SIMD Width
@add16	4	8
@addc16	1	8
@and16	4	8
@fabsh	4	8
@fabss	2	4
@faddh	4	8
@faddhs	2	4
@fadds	2	4
@fnormh	4	8
@fnorms	2	4
@fh2s	1	4
@fh2xp16	1	8
@fmach	4	8
@fmachs	2	4
@fmaxh	1	8
@fmaxs	1	4
@fmovh	4	8
@fmovs	2	4
@fmulh	4	8
@fnegh	4	8
@fnegs	2	4
@fs2h	1	4
@fs2xp16	1	4
@fscaleh	4	8
@fscales	2	4
@fsubh	4	8
@fsubs	2	4
@mov16	4	8
@mov32	2	4
@or16	4	8
@sar16	1	4
@sll16	1	4
@slr16	1	4
@sub16	4	8
@xor16	4	8
@xp162fh	1	8
@xp162fs	1	4

SDK Documentation (1.2.0)

Appendix

Contents

Appendix¶

SIMD Mode¶