{"schema":"libjg2-1", "vpath":"/git/", "avatar":"/git/avatar/", "alang":"", "gen_ut":1750829395, "reponame":"openssl", "desc":"OpenSSL", "owner": { "name": "Andy Green", "email": "andy@warmcat.com", "md5": "c50933ca2aa61e0fe2c43d46bb6b59cb" },"url":"https://warmcat.com/repo/openssl", "f":3, "items": [ {"schema":"libjg2-1", "cid":"5d905c09f993a2fdd58e05501380005a", "commit": {"type":"commit", "time": 1528053606, "time_ofs": 120, "oid_tree": { "oid": "51fae99bc0bb6a7cb78500bc5109f2f86cd03d83", "alias": []}, "oid":{ "oid": "41013cd63c068e2f271fabc92702ee67d800f0cb", "alias": []}, "msg": "PPC assembly pack: correct POWER9 results.", "sig_commit": { "git_time": { "time": 1528053606, "offset": 120 }, "name": "Andy Polyakov", "email": "appro@openssl.org", "md5": "50bd64fa2a792cbbf679fa16213a3b2a" }, "sig_author": { "git_time": { "time": 1527941007, "offset": 120 }, "name": "Andy Polyakov", "email": "appro@openssl.org", "md5": "50bd64fa2a792cbbf679fa16213a3b2a" }}, "body": "PPC assembly pack: correct POWER9 results.\n\nAs it turns out originally published results were skewed by \u0022turbo\u0022\nmode. VM apparently remains oblivious to dynamic frequency scaling,\nand reports that processor operates at \u0022base\u0022 frequency at all times.\nWhile actual frequency gets increased under load.\n\nReviewed-by: Rich Salz \u003crsalz@openssl.org\u003e\n(Merged from https://github.com/openssl/openssl/pull/6406)\n" , "diff": "diff --git a/crypto/aes/asm/aesp8-ppc.pl b/crypto/aes/asm/aesp8-ppc.pl\nindex 8670940..488b133 100755\n--- a/crypto/aes/asm/aesp8-ppc.pl\n+++ b/crypto/aes/asm/aesp8-ppc.pl\n@@ -40,7 +40,8 @@\n #\t\tCBC en-/decrypt\tCTR\tXTS\n # POWER8[le]\t3.96/0.72\t0.74\t1.1\n # POWER8[be]\t3.75/0.65\t0.66\t1.0\n-# POWER9[le]\t3.05/0.65\t0.65\t0.80\n+# POWER9[le]\t4.02/0.86\t0.84\t1.05\n+# POWER9[be]\t3.99/0.78\t0.79\t0.97\n \n $flavour \u003d shift;\n \ndiff --git a/crypto/chacha/asm/chacha-ppc.pl b/crypto/chacha/asm/chacha-ppc.pl\nindex 350d5fa..6dd0581 100755\n--- a/crypto/chacha/asm/chacha-ppc.pl\n+++ b/crypto/chacha/asm/chacha-ppc.pl\n@@ -27,7 +27,7 @@\n # PPC970/G5\t\t9.29/+160%\t?\n # POWER7\t\t8.62/+61%\t3.38\n # POWER8\t\t8.70/+51%\t3.36\n-# POWER9\t\t6.61/+29%\t3.30(*)\n+# POWER9\t\t8.80/+29%\t4.50(*)\n #\n # (*)\tthis is trade-off result, it's possible to improve it, but\n #\tthen it would negatively affect all others;\ndiff --git a/crypto/modes/asm/ghashp8-ppc.pl b/crypto/modes/asm/ghashp8-ppc.pl\nindex 6df485e..6a2ac71 100755\n--- a/crypto/modes/asm/ghashp8-ppc.pl\n+++ b/crypto/modes/asm/ghashp8-ppc.pl\n@@ -30,7 +30,7 @@\n # 2x aggregated reduction improves performance by 50% (resulting\n # performance on POWER8 is 1 cycle per processed byte), and 4x\n # aggregated reduction - by 170% or 2.7x (resulting in 0.55 cpb).\n-# POWER9 delivers 0.40 cpb.\n+# POWER9 delivers 0.51 cpb.\n \n $flavour\u003dshift;\n $output \u003dshift;\ndiff --git a/crypto/poly1305/asm/poly1305-ppc.pl b/crypto/poly1305/asm/poly1305-ppc.pl\nindex cb4ae23..0c6d015 100755\n--- a/crypto/poly1305/asm/poly1305-ppc.pl\n+++ b/crypto/poly1305/asm/poly1305-ppc.pl\n@@ -28,7 +28,7 @@\n # PPC970\t\t7.00/+114%\t3.51/+205%\n # POWER7\t\t3.75/+260%\t1.93/+100%\n # POWER8\t\t-\t\t2.03/+200%\n-# POWER9\t\t-\t\t1.56/+150%\n+# POWER9\t\t-\t\t2.00/+150%\n #\n # Do we need floating-point implementation for PPC? Results presented\n # in poly1305_ieee754.c are tricky to compare to, because they are for\ndiff --git a/crypto/poly1305/asm/poly1305-ppcfp.pl b/crypto/poly1305/asm/poly1305-ppcfp.pl\nindex 2abb8e2..09f8185 100755\n--- a/crypto/poly1305/asm/poly1305-ppcfp.pl\n+++ b/crypto/poly1305/asm/poly1305-ppcfp.pl\n@@ -26,7 +26,6 @@\n # PPC970\t\t6.03/+80%\n # POWER7\t\t3.50/+30%\n # POWER8\t\t3.75/+10%\n-# POWER9\t\t2.80/+12%\n \n $flavour \u003d shift;\n \ndiff --git a/crypto/poly1305/poly1305_ieee754.c b/crypto/poly1305/poly1305_ieee754.c\nindex 1a06e03..995a02e 100644\n--- a/crypto/poly1305/poly1305_ieee754.c\n+++ b/crypto/poly1305/poly1305_ieee754.c\n@@ -38,7 +38,6 @@\n * POWER6 4.92\n * POWER7 4.50\n * POWER8 4.10\n- * POWER9 3.14\n *\n * z10 11.2\n * z196+ 7.30\ndiff --git a/crypto/sha/asm/keccak1600-ppc64.pl b/crypto/sha/asm/keccak1600-ppc64.pl\nindex bc1023e..30e70c5 100755\n--- a/crypto/sha/asm/keccak1600-ppc64.pl\n+++ b/crypto/sha/asm/keccak1600-ppc64.pl\n@@ -30,7 +30,7 @@\n # PPC970/G5\t14.6/+120%\n # POWER7\t10.3/+100%\n # POWER8\t11.5/+85%\n-# POWER9\t7.2/+45%\n+# POWER9\t9.4/+45%\n #\n # (*)\tCorresponds to SHA3-256. Percentage after slash is improvement\n #\tover gcc-4.x-generated KECCAK_1X_ALT code. Newer compilers do\ndiff --git a/crypto/sha/asm/keccak1600p8-ppc.pl b/crypto/sha/asm/keccak1600p8-ppc.pl\nindex a0aeeb0..de2bcd6 100755\n--- a/crypto/sha/asm/keccak1600p8-ppc.pl\n+++ b/crypto/sha/asm/keccak1600p8-ppc.pl\n@@ -23,7 +23,7 @@\n # buffer for r\u003d1088, which matches SHA3-256. This is 17% better than\n # scalar PPC64 code. It probably should be noted that if POWER8's\n # successor can achieve higher scalar instruction issue rate, then\n-# this module will loose... And it does on POWER9 with 8.8 vs. 7.2.\n+# this module will loose... And it does on POWER9 with 12.0 vs. 9.4.\n \n $flavour \u003d shift;\n \ndiff --git a/crypto/sha/asm/sha512p8-ppc.pl b/crypto/sha/asm/sha512p8-ppc.pl\nindex a33ae4d..7a8d435 100755\n--- a/crypto/sha/asm/sha512p8-ppc.pl\n+++ b/crypto/sha/asm/sha512p8-ppc.pl\n@@ -37,8 +37,8 @@\n # build of sha512-ppc.pl, presented for reference.\n #\n #\t\tPOWER8\t\tPOWER9\n-# SHA256\t9.9 [15.8]\t9.2 [9.3]\n-# SHA512\t6.3 [10.3]\t5.8 [5.9]\n+# SHA256\t9.9 [15.8]\t12.2 [12.5]\n+# SHA512\t6.3 [10.3]\t7.7 [7.9]\n \n $flavour\u003dshift;\n $output \u003dshift;\n","s":{"c":1750829395,"u": 45415}} ],"g": 47366,"chitpc": 0,"ehitpc": 0,"indexed":0 , "ab": 0, "si": 0, "db":0, "di":0, "sat":0, "lfc": "0000"}