[Mesa-dev] [PATCH v3 17/24] i965/vec4: consider subregister offset in live variables

[Samuel Iglesias Gonsálvez]

From: "Juan A. Suarez Romero" <jasuarez at igalia.com>

Take into account offset values less than a full register (32 bytes)
when getting the var from register.

This is required when dealing with an operation that writes half of the
register (like one d2x in IVB/BYT, which uses exec_size == 4).

- v2: take in account this offset < 32 in liveness analysis too (Curro)
---
 src/mesa/drivers/dri/i965/brw_vec4_live_variables.cpp | 12 ++++++++----
 src/mesa/drivers/dri/i965/brw_vec4_live_variables.h   |  6 ++++--
 2 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_live_variables.cpp b/src/mesa/drivers/dri/i965/brw_vec4_live_variables.cpp
index 73f658cd8fa..dc1ad21038c 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_live_variables.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_live_variables.cpp
@@ -78,7 +78,8 @@ vec4_live_variables::setup_def_use()
 	    if (inst->src[i].file == VGRF) {
                for (unsigned j = 0; j < DIV_ROUND_UP(inst->size_read(i), 16); j++) {
                   for (int c = 0; c < 4; c++) {
-                     const unsigned v = var_from_reg(alloc, inst->src[i], c, j);
+                     const unsigned v =
+                        var_from_reg(alloc, inst->src[i], c, j);
                      if (!BITSET_TEST(bd->def, v))
                         BITSET_SET(bd->use, v);
                   }
@@ -101,7 +102,8 @@ vec4_live_variables::setup_def_use()
             for (unsigned i = 0; i < DIV_ROUND_UP(inst->size_written, 16); i++) {
                for (int c = 0; c < 4; c++) {
                   if (inst->dst.writemask & (1 << c)) {
-                     const unsigned v = var_from_reg(alloc, inst->dst, c, i);
+                     const unsigned v =
+                        var_from_reg(alloc, inst->dst, c, i);
                      if (!BITSET_TEST(bd->use, v))
                         BITSET_SET(bd->def, v);
                   }
@@ -257,7 +259,8 @@ vec4_visitor::calculate_live_intervals()
 	 if (inst->src[i].file == VGRF) {
             for (unsigned j = 0; j < DIV_ROUND_UP(inst->size_read(i), 16); j++) {
                for (int c = 0; c < 4; c++) {
-                  const unsigned v = var_from_reg(alloc, inst->src[i], c, j);
+                  const unsigned v =
+                     var_from_reg(alloc, inst->src[i], c, j);
                   start[v] = MIN2(start[v], ip);
                   end[v] = ip;
                }
@@ -269,7 +272,8 @@ vec4_visitor::calculate_live_intervals()
          for (unsigned i = 0; i < DIV_ROUND_UP(inst->size_written, 16); i++) {
             for (int c = 0; c < 4; c++) {
                if (inst->dst.writemask & (1 << c)) {
-                  const unsigned v = var_from_reg(alloc, inst->dst, c, i);
+                  const unsigned v =
+                     var_from_reg(alloc, inst->dst, c, i);
                   start[v] = MIN2(start[v], ip);
                   end[v] = ip;
                }
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_live_variables.h b/src/mesa/drivers/dri/i965/brw_vec4_live_variables.h
index 8807c453743..b23df650c11 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_live_variables.h
+++ b/src/mesa/drivers/dri/i965/brw_vec4_live_variables.h
@@ -89,7 +89,8 @@ var_from_reg(const simple_allocator &alloc, const src_reg &reg,
    const unsigned csize = DIV_ROUND_UP(type_sz(reg.type), 4);
    unsigned result =
       8 * (alloc.offsets[reg.nr] + reg.offset / REG_SIZE) +
-      (BRW_GET_SWZ(reg.swizzle, c) + k / csize * 4) * csize + k % csize;
+      (BRW_GET_SWZ(reg.swizzle, c) + k / csize * 4) * csize + k % csize +
+      (reg.offset % REG_SIZE) / type_sz(reg.type);
    /* Do not exceed the limit for this register */
    assert(result < 8 * (alloc.offsets[reg.nr] + alloc.sizes[reg.nr]));
    return result;
@@ -103,7 +104,8 @@ var_from_reg(const simple_allocator &alloc, const dst_reg &reg,
    const unsigned csize = DIV_ROUND_UP(type_sz(reg.type), 4);
    unsigned result =
       8 * (alloc.offsets[reg.nr] + reg.offset / REG_SIZE) +
-      (c + k / csize * 4) * csize + k % csize;
+      (c + k / csize * 4) * csize + k % csize +
+      (reg.offset % REG_SIZE) / type_sz(reg.type);
    /* Do not exceed the limit for this register */
    assert(result < 8 * (alloc.offsets[reg.nr] + alloc.sizes[reg.nr]));
    return result;
-- 
2.11.0