cluster/ec: Add support for hardware acceleration

This patch implements functionalities for fast encoding/decoding
using hardware support. Currently optimized x86_64, SSE and AVX is
added.

Additionally this patch implements a caching mecanism for inverse
matrices to reduce computation time, as well as a new method for
computing the inverse that takes quadratic time instead of cubic.

Finally some unnecessary memory copies have been eliminated to
further increase performance.

Change-Id: I26c75f26fb4201bd22b51335448ea4357235065a
BUG: 1289922
Signed-off-by: Xavier Hernandez <xhernandez@datalab.es>
Reviewed-on: http://review.gluster.org/12837
Tested-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>

1 files changed, 150 insertions, 0 deletions

diff --git a/xlators/cluster/ec/src/ec-code-x64.c b/xlators/cluster/ec/src/ec-code-x64.c
new file mode 100644
index 00000000000..e94ddd4b155
--- /dev/null
+++ b/xlators/cluster/ec/src/ec-code-x64.c
@@ -0,0 +1,150 @@
+/*
+  Copyright (c) 2015 DataLab, s.l. <http://www.datalab.es>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#include <errno.h>
+
+#include "ec-code-intel.h"
+
+static ec_code_intel_reg_t ec_code_x64_regmap[] = {
+    REG_AX, REG_CX, REG_BP, REG_8,  REG_9, REG_10,
+    REG_11, REG_12, REG_13, REG_14, REG_15
+};
+
+static void
+ec_code_x64_prolog(ec_code_builder_t *builder)
+{
+    uint32_t i;
+
+    ec_code_intel_op_push_r(builder, REG_BP);
+    if (!builder->linear) {
+        ec_code_intel_op_push_r(builder, REG_BX);
+    }
+    if (builder->regs > 11) {
+        ec_code_error(builder, EINVAL);
+        return;
+    }
+    for (i = 7; i < builder->regs; i++) {
+        ec_code_intel_op_push_r(builder, ec_code_x64_regmap[i]);
+    }
+
+    builder->loop = builder->address;
+}
+
+static void
+ec_code_x64_epilog(ec_code_builder_t *builder)
+{
+    uint32_t i;
+
+    ec_code_intel_op_add_i2r(builder, 8, REG_DX);
+    ec_code_intel_op_add_i2r(builder, 8, REG_DI);
+    ec_code_intel_op_test_i2r(builder, builder->width - 1, REG_DX);
+    ec_code_intel_op_jne(builder, builder->loop);
+
+    if (builder->regs > 11) {
+        ec_code_error(builder, EINVAL);
+    }
+    for (i = builder->regs; i > 7; i--) {
+        ec_code_intel_op_pop_r(builder, ec_code_x64_regmap[i - 1]);
+    }
+    if (!builder->linear) {
+        ec_code_intel_op_pop_r(builder, REG_BX);
+    }
+    ec_code_intel_op_pop_r(builder, REG_BP);
+    ec_code_intel_op_ret(builder, 0);
+}
+
+static void
+ec_code_x64_load(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
+                 uint32_t bit)
+{
+    dst = ec_code_x64_regmap[dst];
+
+    if (builder->linear) {
+        ec_code_intel_op_mov_m2r(builder, REG_SI, REG_DX, 1,
+                                 idx * builder->width * builder->bits +
+                                 bit * builder->width,
+                                 dst);
+    } else {
+        if (builder->base != idx) {
+            ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
+                                     REG_BX);
+            builder->base = idx;
+        }
+        ec_code_intel_op_mov_m2r(builder, REG_BX, REG_DX, 1,
+                                 bit * builder->width, dst);
+    }
+}
+
+static void
+ec_code_x64_store(ec_code_builder_t *builder, uint32_t src, uint32_t bit)
+{
+    src = ec_code_x64_regmap[src];
+
+    ec_code_intel_op_mov_r2m(builder, src, REG_DI, REG_NULL, 0,
+                             bit * builder->width);
+}
+
+static void
+ec_code_x64_copy(ec_code_builder_t *builder, uint32_t dst, uint32_t src)
+{
+    dst = ec_code_x64_regmap[dst];
+    src = ec_code_x64_regmap[src];
+
+    ec_code_intel_op_mov_r2r(builder, src, dst);
+}
+
+static void
+ec_code_x64_xor2(ec_code_builder_t *builder, uint32_t dst, uint32_t src)
+{
+    dst = ec_code_x64_regmap[dst];
+    src = ec_code_x64_regmap[src];
+
+    ec_code_intel_op_xor_r2r(builder, src, dst);
+}
+
+static void
+ec_code_x64_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
+                 uint32_t bit)
+{
+    dst = ec_code_x64_regmap[dst];
+
+    if (builder->linear) {
+        ec_code_intel_op_xor_m2r(builder, REG_SI, REG_DX, 1,
+                                 idx * builder->width * builder->bits +
+                                 bit * builder->width,
+                                 dst);
+    } else {
+        if (builder->base != idx) {
+            ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
+                                     REG_BX);
+            builder->base = idx;
+        }
+        ec_code_intel_op_xor_m2r(builder, REG_BX, REG_DX, 1,
+                                 bit * builder->width, dst);
+    }
+}
+
+static char *ec_code_x64_needed_flags[] = {
+    NULL
+};
+
+ec_code_gen_t ec_code_gen_x64 = {
+    .name   = "x64",
+    .flags  = ec_code_x64_needed_flags,
+    .width  = sizeof(uint64_t),
+    .prolog = ec_code_x64_prolog,
+    .epilog = ec_code_x64_epilog,
+    .load   = ec_code_x64_load,
+    .store  = ec_code_x64_store,
+    .copy   = ec_code_x64_copy,
+    .xor2   = ec_code_x64_xor2,
+    .xor3   = NULL,
+    .xorm   = ec_code_x64_xorm
+};