From 95f3230bcfbb4e68e2ba313648d288c0b9c0d42b Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Tue, 4 Feb 2020 15:19:53 +0100 Subject: [PATCH] Small speedup for -m 12500 (RAR3-hp) in optimized mode --- OpenCL/m12500-optimized.cl | 193 +++++++++---------------------------- tools/benchmark_deep.pl | 4 +- 2 files changed, 47 insertions(+), 150 deletions(-) diff --git a/OpenCL/m12500-optimized.cl b/OpenCL/m12500-optimized.cl index dbcd97121..d3e91121e 100644 --- a/OpenCL/m12500-optimized.cl +++ b/OpenCL/m12500-optimized.cl @@ -37,134 +37,6 @@ typedef struct rar3_tmp } rar3_tmp_t; -DECLSPEC void sha1_transform_intern (const u32 *w, u32 *digest) -{ - u32 A = digest[0]; - u32 B = digest[1]; - u32 C = digest[2]; - u32 D = digest[3]; - u32 E = digest[4]; - - u32 w0_t = w[ 0]; - u32 w1_t = w[ 1]; - u32 w2_t = w[ 2]; - u32 w3_t = w[ 3]; - u32 w4_t = w[ 4]; - u32 w5_t = w[ 5]; - u32 w6_t = w[ 6]; - u32 w7_t = w[ 7]; - u32 w8_t = w[ 8]; - u32 w9_t = w[ 9]; - u32 wa_t = w[10]; - u32 wb_t = w[11]; - u32 wc_t = w[12]; - u32 wd_t = w[13]; - u32 we_t = w[14]; - u32 wf_t = w[15]; - - #undef K - #define K SHA1C00 - - SHA1_STEP (SHA1_F0o, A, B, C, D, E, w0_t); - SHA1_STEP (SHA1_F0o, E, A, B, C, D, w1_t); - SHA1_STEP (SHA1_F0o, D, E, A, B, C, w2_t); - SHA1_STEP (SHA1_F0o, C, D, E, A, B, w3_t); - SHA1_STEP (SHA1_F0o, B, C, D, E, A, w4_t); - SHA1_STEP (SHA1_F0o, A, B, C, D, E, w5_t); - SHA1_STEP (SHA1_F0o, E, A, B, C, D, w6_t); - SHA1_STEP (SHA1_F0o, D, E, A, B, C, w7_t); - SHA1_STEP (SHA1_F0o, C, D, E, A, B, w8_t); - SHA1_STEP (SHA1_F0o, B, C, D, E, A, w9_t); - SHA1_STEP (SHA1_F0o, A, B, C, D, E, wa_t); - SHA1_STEP (SHA1_F0o, E, A, B, C, D, wb_t); - SHA1_STEP (SHA1_F0o, D, E, A, B, C, wc_t); - SHA1_STEP (SHA1_F0o, C, D, E, A, B, wd_t); - SHA1_STEP (SHA1_F0o, B, C, D, E, A, we_t); - SHA1_STEP (SHA1_F0o, A, B, C, D, E, wf_t); - w0_t = hc_rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F0o, E, A, B, C, D, w0_t); - w1_t = hc_rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F0o, D, E, A, B, C, w1_t); - w2_t = hc_rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F0o, C, D, E, A, B, w2_t); - w3_t = hc_rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F0o, B, C, D, E, A, w3_t); - - #undef K - #define K SHA1C01 - - w4_t = hc_rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, w4_t); - w5_t = hc_rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, w5_t); - w6_t = hc_rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, w6_t); - w7_t = hc_rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, w7_t); - w8_t = hc_rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, w8_t); - w9_t = hc_rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, w9_t); - wa_t = hc_rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, wa_t); - wb_t = hc_rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, wb_t); - wc_t = hc_rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, wc_t); - wd_t = hc_rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, wd_t); - we_t = hc_rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, we_t); - wf_t = hc_rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, wf_t); - w0_t = hc_rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, w0_t); - w1_t = hc_rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, w1_t); - w2_t = hc_rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, w2_t); - w3_t = hc_rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, w3_t); - w4_t = hc_rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, w4_t); - w5_t = hc_rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, w5_t); - w6_t = hc_rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, w6_t); - w7_t = hc_rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, w7_t); - - #undef K - #define K SHA1C02 - - w8_t = hc_rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F2o, A, B, C, D, E, w8_t); - w9_t = hc_rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F2o, E, A, B, C, D, w9_t); - wa_t = hc_rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F2o, D, E, A, B, C, wa_t); - wb_t = hc_rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F2o, C, D, E, A, B, wb_t); - wc_t = hc_rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F2o, B, C, D, E, A, wc_t); - wd_t = hc_rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F2o, A, B, C, D, E, wd_t); - we_t = hc_rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F2o, E, A, B, C, D, we_t); - wf_t = hc_rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F2o, D, E, A, B, C, wf_t); - w0_t = hc_rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F2o, C, D, E, A, B, w0_t); - w1_t = hc_rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F2o, B, C, D, E, A, w1_t); - w2_t = hc_rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F2o, A, B, C, D, E, w2_t); - w3_t = hc_rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F2o, E, A, B, C, D, w3_t); - w4_t = hc_rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F2o, D, E, A, B, C, w4_t); - w5_t = hc_rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F2o, C, D, E, A, B, w5_t); - w6_t = hc_rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F2o, B, C, D, E, A, w6_t); - w7_t = hc_rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F2o, A, B, C, D, E, w7_t); - w8_t = hc_rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F2o, E, A, B, C, D, w8_t); - w9_t = hc_rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F2o, D, E, A, B, C, w9_t); - wa_t = hc_rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F2o, C, D, E, A, B, wa_t); - wb_t = hc_rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F2o, B, C, D, E, A, wb_t); - - #undef K - #define K SHA1C03 - - wc_t = hc_rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, wc_t); - wd_t = hc_rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, wd_t); - we_t = hc_rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, we_t); - wf_t = hc_rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, wf_t); - w0_t = hc_rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, w0_t); - w1_t = hc_rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, w1_t); - w2_t = hc_rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, w2_t); - w3_t = hc_rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, w3_t); - w4_t = hc_rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, w4_t); - w5_t = hc_rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, w5_t); - w6_t = hc_rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, w6_t); - w7_t = hc_rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, w7_t); - w8_t = hc_rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, w8_t); - w9_t = hc_rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, w9_t); - wa_t = hc_rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, wa_t); - wb_t = hc_rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F1, A, B, C, D, E, wb_t); - wc_t = hc_rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, E, A, B, C, D, wc_t); - wd_t = hc_rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, D, E, A, B, C, wd_t); - we_t = hc_rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, C, D, E, A, B, we_t); - wf_t = hc_rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, B, C, D, E, A, wf_t); - - digest[0] += A; - digest[1] += B; - digest[2] += C; - digest[3] += D; - digest[4] += E; -} - KERNEL_FQ void m12500_init (KERN_ATTR_TMPS_ESALT (rar3_tmp_t, pbkdf2_sha1_t)) { /** @@ -268,7 +140,29 @@ KERNEL_FQ void m12500_loop (KERN_ATTR_TMPS_ESALT (rar3_tmp_t, pbkdf2_sha1_t)) { const u32 j16 = j * 16; - sha1_transform_intern (&largeblock[j16], dgst); + u32 w0[4]; + u32 w1[4]; + u32 w2[4]; + u32 w3[4]; + + w0[0] = largeblock[j16 + 0]; + w0[1] = largeblock[j16 + 1]; + w0[2] = largeblock[j16 + 2]; + w0[3] = largeblock[j16 + 3]; + w1[0] = largeblock[j16 + 4]; + w1[1] = largeblock[j16 + 5]; + w1[2] = largeblock[j16 + 6]; + w1[3] = largeblock[j16 + 7]; + w2[0] = largeblock[j16 + 8]; + w2[1] = largeblock[j16 + 9]; + w2[2] = largeblock[j16 + 10]; + w2[3] = largeblock[j16 + 11]; + w3[0] = largeblock[j16 + 12]; + w3[1] = largeblock[j16 + 13]; + w3[2] = largeblock[j16 + 14]; + w3[3] = largeblock[j16 + 15]; + + sha1_transform (w0, w1, w2, w3, dgst); } } } @@ -349,24 +243,27 @@ KERNEL_FQ void m12500_comp (KERN_ATTR_TMPS_ESALT (rar3_tmp_t, pbkdf2_sha1_t)) const u32 p3 = (pw_len * 2) + salt_len + 3; - u32 w_buf[16]; + u32 w0[4]; + u32 w1[4]; + u32 w2[4]; + u32 w3[4]; - w_buf[ 0] = 0x80000000; - w_buf[ 1] = 0; - w_buf[ 2] = 0; - w_buf[ 3] = 0; - w_buf[ 4] = 0; - w_buf[ 5] = 0; - w_buf[ 6] = 0; - w_buf[ 7] = 0; - w_buf[ 8] = 0; - w_buf[ 9] = 0; - w_buf[10] = 0; - w_buf[11] = 0; - w_buf[12] = 0; - w_buf[13] = 0; - w_buf[14] = 0; - w_buf[15] = (p3 * ROUNDS) * 8; + w0[0] = 0x80000000; + w0[1] = 0; + w0[2] = 0; + w0[3] = 0; + w1[0] = 0; + w1[1] = 0; + w1[2] = 0; + w1[3] = 0; + w2[0] = 0; + w2[1] = 0; + w2[2] = 0; + w2[3] = 0; + w3[0] = 0; + w3[1] = 0; + w3[2] = 0; + w3[3] = (p3 * ROUNDS) * 8; u32 dgst[5]; @@ -376,7 +273,7 @@ KERNEL_FQ void m12500_comp (KERN_ATTR_TMPS_ESALT (rar3_tmp_t, pbkdf2_sha1_t)) dgst[3] = tmps[gid].dgst[16][3]; dgst[4] = tmps[gid].dgst[16][4]; - sha1_transform_intern (w_buf, dgst); + sha1_transform (w0, w1, w2, w3, dgst); u32 ukey[4]; @@ -477,7 +374,7 @@ KERNEL_FQ void m12500_comp (KERN_ATTR_TMPS_ESALT (rar3_tmp_t, pbkdf2_sha1_t)) dgst[3] = tmps[gid].dgst[i][3]; dgst[4] = tmps[gid].dgst[i][4]; - sha1_transform_intern (w, dgst); + sha1_transform (w + 0, w + 4, w + 8, w + 12, dgst); PUTCHAR (iv, i, dgst[4] & 0xff); } diff --git a/tools/benchmark_deep.pl b/tools/benchmark_deep.pl index 01e898839..e75332dba 100755 --- a/tools/benchmark_deep.pl +++ b/tools/benchmark_deep.pl @@ -8,7 +8,7 @@ my $amd_cache = "~/.AMD"; my $hashcat_path = "."; my $kernels_cache = "$hashcat_path/kernels"; my $hashcat_bin = "$hashcat_path/hashcat"; -my $device = 3; +my $device = 1; my $workload_profile = 3; my $runtime = 24; my $sleep_sec = 12; @@ -16,7 +16,7 @@ my $default_mask = "?b?b?b?b?b?b?b"; my $result = "result.txt"; my $old_hashcat = 0; # requires to have ran with new hashcat before to create the hashfiles my $repeats = 1; -my $cpu_benchmark = 1; +my $cpu_benchmark = 0; print "\nHardware preparations... You may need to adjust some settings and probably can ignore some of the error\n\n";