summaryrefslogtreecommitdiff
path: root/src/arch/x86/memmove.c
diff options
context:
space:
mode:
authorJohnny Lin <johnny_lin@wiwynn.com>2020-12-18 16:47:57 +0800
committerHung-Te Lin <hungte@chromium.org>2020-12-24 08:15:43 +0000
commit749a78d1794661cfbd72121484eb2b9b80361af8 (patch)
treea7ab3f74b903775bf5170cfa6b77279609bc5b7e /src/arch/x86/memmove.c
parenta858ec86e45fe658086cf6abf55066ca1b8d4c40 (diff)
soc/intel/xeon_sp/cpx: Disable isoch operation for performance
Isochronous operation negatively impacts memory performance, as per Intel MLC (Memory Latency Checker) benchmark results. Thus, disable isochronous operation, like analogous UEFI firmware does. The MLC results after disabling isoch: "--max_bandwidth" ALL Reads : 106948.17 3:1 Reads-Writes : 101580.46 2:1 Reads-Writes : 100523.26 1:1 Reads-Writes : 99059.44 Stream-triad like : 97762.47 "--peak_injection_bandwidth" ALL Reads : 105724.3 3:1 Reads-Writes : 100655.8 2:1 Reads-Writes : 99463 1:1 Reads-Writes : 98708 Stream-triad like : 91515 The MLC results before disabling isoch: "--max_bandwidth" ALL Reads : 88824.96 3:1 Reads-Writes : 94820.81 2:1 Reads-Writes : 94867.53 1:1 Reads-Writes : 92567.36 Stream-triad like : 91900.43 "--peak_injection_bandwidth" ALL Reads : 88859.6 3:1 Reads-Writes : 94064 2:1 Reads-Writes : 94186.2 1:1 Reads-Writes : 92516.1 Stream-triad like : 85147.4 TEST=On OCP Delta Lake, verify that MLC benchmark results have improved. Change-Id: I08c22ee001b601e607452b3f23fad969ecb484b4 Signed-off-by: Johnny Lin <johnny_lin@wiwynn.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/48738 Reviewed-by: Marc Jones <marc@marcjonesconsulting.com> Reviewed-by: Jonathan Zhang <jonzhang@fb.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Diffstat (limited to 'src/arch/x86/memmove.c')
0 files changed, 0 insertions, 0 deletions