Mesa (master): pan/midgard: Use better heuristic for shader termination

[GitLab Mirror]

Module: Mesa
Branch: master
Commit: 66f98ffab095347d93e0310844e3037ab036c1b3
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=66f98ffab095347d93e0310844e3037ab036c1b3

Author: Alyssa Rosenzweig <alyssa.rosenzweig at collabora.com>
Date:   Tue Dec 31 14:21:26 2019 -0500

pan/midgard: Use better heuristic for shader termination

This still may not be perfect (in the sense that legal shaders might
still get cut off) but this fits how writeout is done with both Panfrost
and the blob, so it's good enough for what we need and allows MRT
shaders to be sanely disassembled.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig at collabora.com>

---

 src/panfrost/midgard/disassemble.c | 41 ++++++++++++++++----------------------
 1 file changed, 17 insertions(+), 24 deletions(-)

diff --git a/src/panfrost/midgard/disassemble.c b/src/panfrost/midgard/disassemble.c
index 1b7d4689ef7..115dfc5e676 100644
--- a/src/panfrost/midgard/disassemble.c
+++ b/src/panfrost/midgard/disassemble.c
@@ -736,10 +736,11 @@ print_branch_cond(int cond)
         }
 }
 
-static void
+static bool
 print_compact_branch_writeout_field(uint16_t word)
 {
         midgard_jmp_writeout_op op = word & 0x7;
+        midg_stats.instruction_count++;
 
         switch (op) {
         case midgard_jmp_writeout_op_branch_uncond: {
@@ -757,7 +758,7 @@ print_compact_branch_writeout_field(uint16_t word)
                 print_tag_short(br_uncond.dest_tag);
                 printf("\n");
 
-                break;
+                return br_uncond.offset >= 0;
         }
 
         case midgard_jmp_writeout_op_branch_cond:
@@ -781,14 +782,14 @@ print_compact_branch_writeout_field(uint16_t word)
                 print_tag_short(br_cond.dest_tag);
                 printf("\n");
 
-                break;
+                return br_cond.offset >= 0;
         }
         }
 
-        midg_stats.instruction_count++;
+        return false;
 }
 
-static void
+static bool
 print_extended_branch_writeout_field(uint8_t *words, unsigned next)
 {
         midgard_branch_extended br;
@@ -836,6 +837,7 @@ print_extended_branch_writeout_field(uint8_t *words, unsigned next)
         midg_tags[I] = br.dest_tag;
 
         midg_stats.instruction_count++;
+        return br.offset >= 0;
 }
 
 static unsigned
@@ -873,7 +875,7 @@ float_bitcast(uint32_t integer)
         return v.f;
 }
 
-static void
+static bool
 print_alu_word(uint32_t *words, unsigned num_quad_words,
                unsigned tabs, unsigned next)
 {
@@ -882,6 +884,7 @@ print_alu_word(uint32_t *words, unsigned num_quad_words,
         unsigned num_fields = num_alu_fields_enabled(control_word);
         uint16_t *word_ptr = beginning_ptr + num_fields;
         unsigned num_words = 2 + num_fields;
+        bool branch_forward = false;
 
         if ((control_word >> 16) & 1)
                 printf("unknown bit 16 enabled\n");
@@ -933,13 +936,13 @@ print_alu_word(uint32_t *words, unsigned num_quad_words,
         }
 
         if ((control_word >> 26) & 1) {
-                print_compact_branch_writeout_field(*word_ptr);
+                branch_forward |= print_compact_branch_writeout_field(*word_ptr);
                 word_ptr += 1;
                 num_words += 1;
         }
 
         if ((control_word >> 27) & 1) {
-                print_extended_branch_writeout_field((uint8_t *) word_ptr, next);
+                branch_forward |= print_extended_branch_writeout_field((uint8_t *) word_ptr, next);
                 word_ptr += 3;
                 num_words += 3;
         }
@@ -984,6 +987,8 @@ print_alu_word(uint32_t *words, unsigned num_quad_words,
 
                 }
         }
+
+        return branch_forward;
 }
 
 static void
@@ -1454,9 +1459,8 @@ disassemble_midgard(uint8_t *code, size_t size, unsigned gpu_id, gl_shader_stage
         uint32_t *words = (uint32_t *) code;
         unsigned num_words = size / 4;
         int tabs = 0;
-        num_words = MIN2(num_words, 100);
 
-        bool prefetch_flag = false;
+        bool branch_forward = false;
 
         int last_next_tag = -1;
 
@@ -1515,7 +1519,7 @@ disassemble_midgard(uint8_t *code, size_t size, unsigned gpu_id, gl_shader_stage
                         break;
 
                 case midgard_word_type_alu:
-                        print_alu_word(&words[i], num_quad_words, tabs, i + 4*num_quad_words);
+                        branch_forward = print_alu_word(&words[i], num_quad_words, tabs, i + 4*num_quad_words);
 
                         /* Reset word static analysis state */
                         is_embedded_constant_half = false;
@@ -1537,23 +1541,12 @@ disassemble_midgard(uint8_t *code, size_t size, unsigned gpu_id, gl_shader_stage
                 midg_stats.bundle_count++;
                 midg_stats.quadword_count += num_quad_words;
 
-                if (prefetch_flag && midgard_word_types[tag] == midgard_word_type_alu)
-                        break;
-
                 printf("\n");
 
                 unsigned next = (words[i] & 0xF0) >> 4;
 
-                /* Break based on instruction prefetch flag */
-
-#if 0
-                if (i < num_words && next == 1) {
-                        prefetch_flag = true;
-
-                        if (midgard_word_types[words[i] & 0xF] != midgard_word_type_alu)
-                                break;
-                }
-#endif
+                if (i < num_words && next == 1 && !branch_forward)
+                        break;
 
                 i += 4 * num_quad_words;
         }