From: Nicolai Hähnle <[email protected]>
---
src/amd/common/ac_binary.c | 77 +++++++++++++++++++++-----------------
src/amd/common/ac_binary.h | 4 ++
2 files changed, 47 insertions(+), 34 deletions(-)
diff --git a/src/amd/common/ac_binary.c b/src/amd/common/ac_binary.c
index fabeb15a204..44251886b5f 100644
--- a/src/amd/common/ac_binary.c
+++ b/src/amd/common/ac_binary.c
@@ -199,57 +199,30 @@ const unsigned char *ac_shader_binary_config_start(
unsigned i;
for (i = 0; i < binary->global_symbol_count; ++i) {
if (binary->global_symbol_offsets[i] == symbol_offset) {
unsigned offset = i * binary->config_size_per_symbol;
return binary->config + offset;
}
}
return binary->config;
}
-
-static const char *scratch_rsrc_dword0_symbol =
- "SCRATCH_RSRC_DWORD0";
-
-static const char *scratch_rsrc_dword1_symbol =
- "SCRATCH_RSRC_DWORD1";
-
-void ac_shader_binary_read_config(struct ac_shader_binary *binary,
- struct ac_shader_config *conf,
- unsigned symbol_offset,
- bool supports_spill)
+/* Parse configuration data in .AMDGPU.config section format. */
+void ac_parse_shader_binary_config(const char *data, size_t nbytes,
+ bool really_needs_scratch,
+ struct ac_shader_config *conf)
{
- unsigned i;
- const unsigned char *config =
- ac_shader_binary_config_start(binary, symbol_offset);
- bool really_needs_scratch = false;
uint32_t wavesize = 0;
- /* LLVM adds SGPR spills to the scratch size.
- * Find out if we really need the scratch buffer.
- */
- if (supports_spill) {
- really_needs_scratch = true;
- } else {
- for (i = 0; i < binary->reloc_count; i++) {
- const struct ac_shader_reloc *reloc =
&binary->relocs[i];
- if (!strcmp(scratch_rsrc_dword0_symbol, reloc->name) ||
- !strcmp(scratch_rsrc_dword1_symbol, reloc->name)) {
- really_needs_scratch = true;
- break;
- }
- }
- }
-
- for (i = 0; i < binary->config_size_per_symbol; i+= 8) {
- unsigned reg = util_le32_to_cpu(*(uint32_t*)(config + i));
- unsigned value = util_le32_to_cpu(*(uint32_t*)(config + i + 4));
+ for (size_t i = 0; i < nbytes; i += 8) {
+ unsigned reg = util_le32_to_cpu(*(uint32_t*)(data + i));
+ unsigned value = util_le32_to_cpu(*(uint32_t*)(data + i + 4));
switch (reg) {
case R_00B028_SPI_SHADER_PGM_RSRC1_PS:
case R_00B128_SPI_SHADER_PGM_RSRC1_VS:
case R_00B228_SPI_SHADER_PGM_RSRC1_GS:
case R_00B848_COMPUTE_PGM_RSRC1:
case R_00B428_SPI_SHADER_PGM_RSRC1_HS:
conf->num_sgprs = MAX2(conf->num_sgprs,
(G_00B028_SGPRS(value) + 1) * 8);
conf->num_vgprs = MAX2(conf->num_vgprs,
(G_00B028_VGPRS(value) + 1) * 4);
conf->float_mode = G_00B028_FLOAT_MODE(value);
break;
@@ -292,20 +265,56 @@ void ac_shader_binary_read_config(struct ac_shader_binary
*binary,
if (!conf->spi_ps_input_addr)
conf->spi_ps_input_addr = conf->spi_ps_input_ena;
}
if (really_needs_scratch) {
/* sgprs spills aren't spilling */
conf->scratch_bytes_per_wave = G_00B860_WAVESIZE(wavesize) *
256 * 4;
}
}
+static const char *scratch_rsrc_dword0_symbol =
+ "SCRATCH_RSRC_DWORD0";
+
+static const char *scratch_rsrc_dword1_symbol =
+ "SCRATCH_RSRC_DWORD1";
+
+void ac_shader_binary_read_config(struct ac_shader_binary *binary,
+ struct ac_shader_config *conf,
+ unsigned symbol_offset,
+ bool supports_spill)
+{
+ unsigned i;
+ const char *config =
+ (const char *)ac_shader_binary_config_start(binary,
symbol_offset);
+ bool really_needs_scratch = false;
+ /* LLVM adds SGPR spills to the scratch size.
+ * Find out if we really need the scratch buffer.
+ */
+ if (supports_spill) {
+ really_needs_scratch = true;
+ } else {
+ for (i = 0; i < binary->reloc_count; i++) {
+ const struct ac_shader_reloc *reloc =
&binary->relocs[i];
+
+ if (!strcmp(scratch_rsrc_dword0_symbol, reloc->name) ||
+ !strcmp(scratch_rsrc_dword1_symbol, reloc->name)) {
+ really_needs_scratch = true;
+ break;
+ }
+ }
+ }
+
+ ac_parse_shader_binary_config(config, binary->config_size_per_symbol,
+ really_needs_scratch, conf);
+}
+
void ac_shader_binary_clean(struct ac_shader_binary *b)
{
if (!b)
return;
FREE(b->code);
FREE(b->config);
FREE(b->rodata);
FREE(b->global_symbol_offsets);
FREE(b->relocs);
FREE(b->disasm_string);
diff --git a/src/amd/common/ac_binary.h b/src/amd/common/ac_binary.h
index 735e3932055..febc4da7fed 100644
--- a/src/amd/common/ac_binary.h
+++ b/src/amd/common/ac_binary.h
@@ -17,20 +17,21 @@
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE
* SOFTWARE.
*/
#ifndef AC_BINARY_H
#define AC_BINARY_H
+#include <stddef.h>
#include <stdint.h>
#include <stdbool.h>
#ifdef __cplusplus
extern "C" {
#endif
struct ac_shader_reloc {
char name[32];
uint64_t offset;
@@ -89,20 +90,23 @@ bool ac_elf_read(const char *elf_data, unsigned elf_size,
struct ac_shader_binary *binary);
/**
* @returns A pointer to the start of the configuration information for
* the function starting at \p symbol_offset of the binary.
*/
const unsigned char *ac_shader_binary_config_start(
const struct ac_shader_binary *binary,
uint64_t symbol_offset);
+void ac_parse_shader_binary_config(const char *data, size_t nbytes,
+ bool really_needs_scratch,
+ struct ac_shader_config *conf);
void ac_shader_binary_read_config(struct ac_shader_binary *binary,
struct ac_shader_config *conf,
unsigned symbol_offset,
bool supports_spill);
void ac_shader_binary_clean(struct ac_shader_binary *b);
#ifdef __cplusplus
}
#endif
--
2.20.1
_______________________________________________
mesa-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
