Intel ISPC 1.28 Adds Optimized Support For AMD Zen 4 & Zen 5 CPUs
([Intel] 5 Hours Ago
Intel ISPC 1.28)
- Reference: 0001569242
- News link: https://www.phoronix.com/news/Intel-ISPC-1.28
- Source link:
A new version of the Intel Implicit SPMD Program Compiler "ISPC" was just published for supporting that C programming language variant optimized for single program. multiple data (SPMD) programming that is optimized for Intel's various hardware offerings. While catering to Intel hardware, ISPC 1.28 notably adds new AMD Zen 4 and Zen 5 processor targets.
ISPC 1.28 brings some minor language changes, now supports ISPC usage as a C++ library for embedding ISPC compilation directly into applications, Python integration enhancements, new standard library functions and more vector types, and various optimizations and fixes.
The ISPC-as-a-Library support includes experimental support for Just-In-Time (JIT) compilation for run-time code generation and execution.
On the performance front, optimized shuffle / shift / rotate and reduce_equal on AVX-512 targets is now seeing up to a 90% speed-up.
Also notable with the Intel ISPC 1.28 compiler is adding AMD Zen 4 and Zen 5 targets. [1]This pull from open-source developer Jérôme Richard adds that Zen 4 and Zen 5 target support for taking advantage of the processor's ISA capabilities. Notably with Zen 4 and Zen 5 is now having AVX-512 support and other enhancements over earlier AMD Ryzen and EPYC processors. AVX-512 in particular is quite beneficial for Intel ISPC.
More details on all of the changes to find with Intel ISPC 1.28 and Windows/Linux downloads via [2]GitHub .
[1] https://github.com/ispc/ispc/pull/3426
[2] https://github.com/ispc/ispc/releases/tag/v1.28.0
ISPC 1.28 brings some minor language changes, now supports ISPC usage as a C++ library for embedding ISPC compilation directly into applications, Python integration enhancements, new standard library functions and more vector types, and various optimizations and fixes.
The ISPC-as-a-Library support includes experimental support for Just-In-Time (JIT) compilation for run-time code generation and execution.
On the performance front, optimized shuffle / shift / rotate and reduce_equal on AVX-512 targets is now seeing up to a 90% speed-up.
Also notable with the Intel ISPC 1.28 compiler is adding AMD Zen 4 and Zen 5 targets. [1]This pull from open-source developer Jérôme Richard adds that Zen 4 and Zen 5 target support for taking advantage of the processor's ISA capabilities. Notably with Zen 4 and Zen 5 is now having AVX-512 support and other enhancements over earlier AMD Ryzen and EPYC processors. AVX-512 in particular is quite beneficial for Intel ISPC.
More details on all of the changes to find with Intel ISPC 1.28 and Windows/Linux downloads via [2]GitHub .
[1] https://github.com/ispc/ispc/pull/3426
[2] https://github.com/ispc/ispc/releases/tag/v1.28.0
sdack